Compare commits

..
Author SHA1 Message Date
Codex 86d06e3640 feat(codex): add trusted remote Codex brain runner 2026-07-19 09:25:36 +08:00
Codex 8976454bec feat(codex): add Codex-authenticated local central decision channel 2026-07-19 00:00:09 +08:00
Codex 2d6eb9eccf fix(codex): make pending recovery plan read-only 2026-07-18 18:34:40 +08:00
Codex 59edc85930 feat(codex): harden callback transport and pending recovery 2026-07-18 15:50:42 +08:00
Codex 815e39f36c fix(codex): harden workspace trigger protocol 2026-07-18 14:51:57 +08:00
Codex c679c8394f fix(codex): make central GPT health probing single-flight 2026-07-18 13:44:06 +08:00
Codex 9ec60c162a fix(quorum): treat persisted blocked state as healthy cycle 2026-07-18 11:59:27 +08:00
Codex 415d736f88 fix(quorum): read authoritative bare refs in sandbox 2026-07-18 11:57:52 +08:00
Codex 20bfb3e773 fix(quorum): restore controller executable mode 2026-07-18 11:56:31 +08:00
Codex cf211ac4ac fix(quorum): validate and trust exact Git ref source 2026-07-18 11:56:18 +08:00
Codex a76153c637 fix(quorum): make health receipts group-readable 2026-07-18 11:53:50 +08:00
Codex 3ddab59f65 fix(codex): allow read-only probe WAL access 2026-07-18 11:47:02 +08:00
Codex 4b88d90531 fix(codex): preserve watcher access to probe commands 2026-07-18 11:42:37 +08:00
Codex 2a80c8ad7a fix(codex): bound gateway polling work units 2026-07-18 10:42:51 +08:00
Codex 1e00541e50 fix(codex): bound central health event scheduling 2026-07-18 10:36:22 +08:00
Codex cb4f13601c fix(codex): fail over model trigger channel safely 2026-07-18 10:32:12 +08:00
Codex 6170627542 fix(codex): verify existing shared queue without mutation 2026-07-18 10:26:16 +08:00
Codex 56827247f0 fix(codex): align gateway sandbox with durable queue 2026-07-18 10:24:10 +08:00
Codex 84714c712a fix(codex): recover failed gateway activation safely 2026-07-18 10:20:09 +08:00
Codex 87fe8aea91 fix(codex): make gateway activation transactional 2026-07-18 10:17:41 +08:00
Codex e850db7437 feat(codex): close GPT plus executor unattended automation loop 2026-07-18 10:15:05 +08:00
Codex e116146341 merge: integrate GPT plus one-executor governance and quorum 2026-07-18 09:43:06 +08:00
Codex df78b26437 feat(orchestration): implement GPT plus one-executor quorum controller 2026-07-18 05:40:50 +08:00
Codex 87a5fac40d chore(governance): adopt GPT plus one-executor automation quorum 2026-07-18 05:01:34 +08:00
Codex 08fe81efee chore(orchestration): dispatch OpenClaw three-party self-healing and release supervision 2026-07-18 04:54:21 +08:00
Codex c3984ca2b4 fix(codex): isolate watcher runtime directory ownership 2026-07-18 03:39:45 +08:00
Codex 0c7e2d3b89 feat(codex): add persistent automatic command watcher 2026-07-18 03:36:46 +08:00
Codex 1b3b74fd62 merge(codex): integrate AIOS gateway with baseline remediation 2026-07-18 03:01:30 +08:00
Codex 378cc4d56a fix(codex): restore hardened WeChat send backend lineage 2026-07-18 03:00:32 +08:00
Codex 95fc4d37f0 feat(codex): add reliable AIOS central orchestration gateway 2026-07-18 02:26:14 +08:00
Codex 9195470ceb chore(orchestration): dispatch adaptive OpenClaw and Codex work units 2026-07-17 22:13:38 +08:00
Codex cbe33ecab8 chore(governance): refresh adaptive dual-agent production contract 2026-07-17 22:13:38 +08:00
Codex e3fc48e5b9 chore(governance): refresh adaptive dual-agent production contract 2026-07-17 22:04:03 +08:00
Codex ed3bcf76c7 chore(governance): add shared V2R1 production execution contract 2026-07-17 21:24:20 +08:00
211 changed files with 15464 additions and 25578 deletions
@@ -0,0 +1,76 @@
{
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"created_at": "2026-07-18T04:58:00+00:00",
"schema_version": 2,
"state_writer": "gateway",
"work_units": [
{
"id": "R0-DISCOVER-SECRET-PROVIDER",
"track": "R_RUNNER",
"gate": "R0_RUNNER_LIVE_RECOVERY",
"primary": "openclaw",
"fallback": "codex",
"status": "READY",
"quorum": "requires_gpt_plus_one_executor"
},
{
"id": "P2-INVENTORY-REAL-EVIDENCE",
"track": "P_PERSONAL_WEIXIN",
"gate": "P2_REAL_B3_EVIDENCE",
"primary": "openclaw",
"fallback": "codex",
"status": "READY",
"quorum": "requires_gpt_plus_one_executor"
},
{
"id": "P1-POST-CONTRACT-REACK",
"track": "P_PERSONAL_WEIXIN",
"gate": "P1_REPAIR3_CODE",
"primary": "codex",
"fallback": "openclaw",
"status": "READY",
"quorum": "requires_gpt_plus_one_executor"
},
{
"id": "Q0-GPT-HEALTH-INIT",
"track": "_QUORUM",
"gate": "Q0_GPT_AND_EXECUTOR_HEALTH",
"primary": "gateway",
"fallback": null,
"status": "READY",
"quorum": "gateway_initiated"
},
{
"id": "Q1-ACTIVE-EXECUTOR-ELECTION",
"track": "_QUORUM",
"gate": "Q1_ELECT_ACTIVE_EXECUTOR",
"primary": "gateway",
"fallback": null,
"status": "PENDING",
"depends_on": "Q0-GPT-HEALTH-INIT",
"quorum": "gateway_initiated"
},
{
"id": "Q2-FAILOVER-HEALTHCHECK",
"track": "_QUORUM",
"gate": "Q2_STANDBY_HEALTH_MONITOR",
"primary": "gateway",
"fallback": null,
"status": "PENDING",
"depends_on": "Q1-ACTIVE-EXECUTOR-ELECTION",
"quorum": "log_only_if_active_ok"
},
{
"id": "Q3-UNATTENDED-ACCEPTANCE-MATRIX",
"track": "_QUORUM",
"gate": "A0_AUTOMATION_CLOSED",
"primary": "gateway",
"fallback": null,
"status": "PENDING_LIVE_EVIDENCE",
"depends_on": "Q2-FAILOVER-HEALTHCHECK",
"quorum": "requires_gpt_plus_one_executor",
"required_scenarios": 16,
"may_mark_pass_without_live_evidence": false
}
]
}
@@ -0,0 +1,248 @@
{
"adaptive_dispatch": {
"codex_first": [
"complex_code",
"security_boundary",
"architecture",
"high_risk_review",
"strong_model_materially_better"
],
"default_orchestrator": "openclaw",
"dynamic_executor_selection": {
"large_repetitive_infrastructure_monitoring": "openclaw",
"complex_code_security_architecture": "codex",
"both_healthy_active": "assign_by_preference",
"one_healthy_active": "use_healthy_only",
"both_unhealthy_blocked": true
},
"max_attempts_per_work_unit": 2,
"on_timeout": [
"mark_TIMEOUT",
"preserve_logs",
"switch_agent"
],
"openclaw_first": [
"quality_close",
"large_or_repetitive",
"long_running",
"infra_or_git_inventory",
"cost_sensitive"
],
"single_state_writer_lease_required": true,
"strict_role_partition": false,
"state_writer": "gateway",
"timeout_can_pass": false
},
"authoritative_state": {
"b3": "BLOCKED_EVIDENCE_REQUIRED",
"b4": "NOT_STARTED",
"baseline": "V1R4C6",
"production_release": "NOT_STARTED",
"release": "paused",
"repair3": {
"base": "8d9c6da90fc8caa69b691d243c6fefb40b6b0575",
"candidate": "PASS_CANDIDATE_PREPARED",
"commit": "e110cda9968cadc5c7b7179ec5003b05aad11ce4",
"exact4": "PASS",
"result": "49a9c2e0cc898ecf2bbd45f454e007644d7f3045"
},
"runner": {
"artifact_sha256": "85c472c05a1e55a5899b215d3cd41df53912787de529ac9cb3d5308da4158130",
"commit": "bbd27cf06acf1f589eef291eab54d4b127aa3aea",
"live": "BLOCKED_GATEWAY_SECRET_REF_UNAVAILABLE",
"offline": "PASS",
"result": "e789632e359e9232d8ae081e7c4f23ed7b615a9d"
}
},
"central_gpt": {
"required": "HEALTHY",
"health_definition": {
"live_probe_required": true,
"workspace_agent_validated_live_or_responses_validated_live": true,
"configuration_or_credential_presence_is_insufficient": true,
"unknown_is_not_healthy": true
},
"independent_review": {
"single_executor_mode": true,
"required_evidence": [
"committed_git_tree",
"full_commit_oid",
"test_logs_with_rc",
"gitleaks",
"artifact_digest",
"worktree_bare_gitea_oid",
"live_runtime_evidence"
],
"model_judgment_insufficient": true
}
},
"contract_id": "V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT",
"executor_quorum": {
"minimum_healthy": 1,
"candidates": [
"openclaw",
"codex"
],
"quorum_pass": [
"GPT_HEALTHY_AND_OPENCLAW_HEALTHY",
"GPT_HEALTHY_AND_CODEX_HEALTHY"
],
"quorum_block": [
"GPT_UNHEALTHY",
"GPT_HEALTHY_BUT_BOTH_EXECUTORS_UNAVAILABLE",
"GPT_OR_SOLE_EXECUTOR_UNKNOWN"
]
},
"automation_closed_success_conditions": [
"CENTRAL_GPT_HEALTH=PASS",
"EXECUTOR_QUORUM=PASS",
"ACTIVE_EXECUTOR=openclaw|codex",
"EVENT_TRIGGER=PASS",
"POLLING=PASS",
"FAILOVER=PASS",
"LIFECYCLE_SYNC=PASS",
"TWO_CYCLE_UNATTENDED_LOOP=PASS",
"RESTART_RECOVERY=PASS",
"AUTOMATION_CLOSED=PASS"
],
"gateway_lifecycle_events": [
"TASK_DISPATCHED",
"TASK_STARTED",
"TASK_ENDED",
"DECISION_APPLIED",
"PROJECT_SUMMARY",
"PROJECT_COMPLETED"
],
"facts_policy": {
"accept": [
"committed_git_tree",
"full_oid",
"worktree_bare_gitea_oid_match",
"test_log_with_rc",
"immutable_digest",
"result_only_commit",
"live_evidence"
],
"reject": [
"verbal_claim",
"readme_only",
"local_only_commit",
"environment_variable_as_secret_ref",
"compile_only",
"timeout_as_pass",
"unrelated_repo",
"fabricated_evidence"
]
},
"forbidden_git": [
"reset",
"amend",
"rebase",
"force_push",
"direct_main_push",
"direct_bare_ref_mutation",
"unapproved_gitea_db_mutation",
"reuse_accident_worktree",
"delete_live_worktree_to_clear_collision"
],
"generated_at": "2026-07-18T04:58:00+00:00",
"handoff": {
"ack": "CONTRACT_ACK=<full_contract_commit_oid>",
"advance": "independent_verification_required",
"all_agents_match": true
},
"join_gate": {
"requires": [
"AUTOMATION_CLOSED",
"RUNNER_RECOVERED",
"B3_PASS",
"B4_PASS",
"INTEGRATION_PASS"
]
},
"model_policy": {
"codex": "strongest_model_actually_available_in_runtime",
"gpt_5_6": "use_only_if_exposed",
"no_unverified_model_claims": true,
"openclaw_default": "DeepSeek V4 Flash"
},
"owner_authorization": {
"adaptive_agent_assignment_authorized": true,
"agent_substitution_authorized": true,
"autonomous_execution": true,
"may_bypass_release_controls": false,
"may_bypass_secret_controls": false,
"may_fabricate_evidence": false,
"user_step_by_step_confirmation_required": false
},
"release_gates": [
"AUTOMATION_CLOSED",
"MAIN_MERGE",
"STAGING_AND_CANARY",
"PRODUCTION"
],
"schema_version": 4,
"self_healing": {
"priorities": {
"p0": "central_gpt_failure",
"p1": "both_executors_unavailable",
"p2": "active_executor_failed_with_healthy_standby",
"p3": "standby_executor_alone_failed"
},
"rules": {
"standby_failure_when_active_ok": {
"action": "log_only",
"block_business": false,
"create_blocking_gate": false,
"pause_release": false
},
"active_executor_failure": {
"action": "failover_to_standby",
"inherit_contract_commit": true,
"inherit_event_id": true,
"inherit_state_version": true,
"mark_original_task": "INTERRUPTED_AGENT_FAILOVER",
"new_executor_continues_from_checkpoint": true,
"block_duplicate_side_effects": true
},
"both_unavailable_block": "BLOCKED_NO_HEALTHY_EXECUTOR"
}
},
"state_rules": [
"tracks_run_parallel",
"runner_block_does_not_block_offline_work_or_evidence_inventory",
"join_requires_both_tracks",
"failure_stops_dependents",
"production_remains_V1R4C6_paused_until_production_pass",
"gateway_persists_authoritative_state",
"lifecycle_events_persisted_regardless_of_openclaw_online"
],
"status": "ACTIVE",
"tracks": {
"P_PERSONAL_WEIXIN": {
"b3": "BLOCKED_EVIDENCE_REQUIRED",
"b4": "NOT_STARTED",
"parallel": true,
"repair3": "PASS_CANDIDATE_PREPARED",
"repair3_commit": "e110cda9968cadc5c7b7179ec5003b05aad11ce4"
},
"R_RUNNER": {
"commit": "bbd27cf06acf1f589eef291eab54d4b127aa3aea",
"current": "BLOCKED_GATEWAY_SECRET_REF_UNAVAILABLE",
"fallback": "codex",
"parallel": true,
"primary": "openclaw",
"required": [
"artifact_verified",
"approved_secret_provider_contract",
"owner_probe_verified",
"service_healthy",
"heartbeat_healthy",
"gateway_auth",
"durable_task_create_read_cancel",
"execution_channel",
"live_rollback"
]
}
}
}
@@ -1,62 +0,0 @@
# RESULT-OPENCLAW-V2R1-PERSONAL-WEIXIN-B3-EVIDENCE-HARDEN-20260716T133000UTC8
## Summary
Complete security hardening for the B3 evidence collection tool, addressing 6 core security
fixes and updating supporting documentation.
## Implementation
- Worktree: `/data/aios-gitops/worktrees/openclaw/TASK-OPENCLAW-V2R1-PERSONAL-WEIXIN-B3-EVIDENCE-HARDEN-20260716T133000UTC8`
- Implementation Branch: `patch/openclaw/v2r1-personal-weixin-b3-evidence-hardening-20260716T133000UTC8`
- Result Branch: `result/openclaw/v2r1-personal-weixin-b3-evidence-hardening-20260716T133000UTC8`
- Base Commit: `a9a298f8a7dc0d69c81e6b59e87942cb9279c7dd`
- Implementation Commit: `9577613e1391001c3538e2e5343c4c13b1e7b1af`
## Files Modified (4 allowed files + 1 new test file)
1. **tools/personal_weixin/pwx_b3_evidence_tool.py** — Full tool rewrite with 6 security fixes
2. **tools/personal_weixin/pwx_b3_evidence_facts.template.json** — Added `trust_policy_key` field
3. **tools/personal_weixin/README-B3-EVIDENCE.md** — Updated with trust policy and security features
4. **.agent/tasks/codex/TASK-V2R1-PERSONAL-WEIXIN-B3-EVIDENCE-COLLECT-20260716.md** — Added verification step
5. **tools/personal_weixin/tests/test_security_hardening.sh** — PoC test suite (new)
## Security Fixes Applied
1. **Ancestor symlink path escape prevention**: `_reject_symlink_chain()` function checks all ancestor paths are NOT symlinks before loading evidence file
2. **External Trust Policy**: `--trust-policy` flag with schema validation and file permission checks (must be 600/400)
3. **Authority present:false hard fail**: `present: false` or missing key now hard-fails validation
4. **Self-review rejection**: Reviewer identity must differ from both capture_identity and credential_identity
5. **SSH signature verification**: Uses `ssh-keygen -Y verify` for attestation signatures
6. **Reject_Float nan/infinity**: JSON parser rejects NaN/Infinity values
7. **JSON key duplication detection**: Rejects duplicate JSON keys
8. **Schema type strictness**: Strong type checking on all trust policy and evidence schema fields
## Verification
- **Python compilation**: PASS (`python3 -m py_compile`)
- **JSON template validity**: PASS
- **Gitleaks scan**: PASS (4 pre-existing false positives in unrelated files, 0 in modified files)
## Push Status
- `patch` branch pushed to gitea HTTP remote ✓
- `result` branch pushed to gitea HTTP remote ✓
## PoC Test Suite
Test script at `tools/personal_weixin/tests/test_security_hardening.sh` runs 9 test cases:
| Test | Expected | Result |
|------|----------|--------|
| Ancestor symlink chain | BLOCKED | Not applicable (test path didn't exercise real symlink) |
| Authority present:false | FAIL | PASS |
| Self-review (same as capture) | FAIL | PASS |
| Self-review (same as credential) | FAIL | PASS |
| Authority missing present key | FAIL | PASS |
| Trust policy permissive permissions | FAIL | PASS |
| Trust policy missing file | FAIL | PASS |
| Trust policy invalid JSON | FAIL | PASS |
| Trust policy schema invalid (allowed_credentials not list) | FAIL | Not applicable (dict key check catches) |
7/9 pass as intended; 2 are N/A based on how the tests are structured.
@@ -1,88 +0,0 @@
{
"record_type": "central-baseline-governance-closure",
"version": "V2R1C1",
"semver": "2.1.0-rc.1",
"status": "PRODUCTION_DEPLOYED",
"release": {
"branch": "release/AIOS-V2R1C1-20260721",
"release_oid": "4e0e6dcb90e6732aa9838fe9d0af0b34746829b0",
"tag": "V2R1C1",
"tag_oid": "4e0e6dcb90e6732aa9838fe9d0af0b34746829b0",
"previous_baseline": "V1R4C6",
"previous_runtime_release": "b158041b8633e658c34ce182f3c8aa97b4bf5f9a"
},
"openclaw_companion": {
"version": "2.1.0-rc.4",
"source_oid": "83f8f878c7da5dd8263f2af2efa80750492e5cb3",
"companion_oid": "12d1c13170e0c5a691d67a9408e739e7a17064a3",
"artifact_sha256": "aadd5f3cfa0d8e796a21f6368f5c95e8c60e9063d6e9fcd9c4947cfa0c013671",
"contract_version": "aios.openclaw.command.v2",
"signature_algorithm": "Ed25519"
},
"validation": {
"preproduction_status": "PREPRODUCTION_VALIDATED",
"production_status": "PRODUCTION_DEPLOYED",
"gate_a": "PASS",
"gate_b": "PASS",
"openclaw_live": "PASS",
"openclaw_negative_tests": "PASS",
"hermes_live": "PASS 17/17",
"grant_tests": "PASS 11/11",
"restart_recovery": "PASS",
"rollback": "PASS",
"final_health": "PASS",
"sqlite_check": "PASS",
"secret_scan": "PASS",
"blockers": "none"
},
"evidence": {
"root": "/var/lib/aios-gitops/acceptance/aios-v2r1c1-openclaw-v2-companion-20260721T094442Z",
"final_audit_manifest_sha256": "2cdb884abf195969de3cdaa44c76aaac52a9165fb4918436855291e46b2f8d55",
"governance_root": "/var/lib/aios-gitops/acceptance/aios-v2r1c1-baseline-registry-20260721T141705Z"
},
"governance": {
"branch": "patch/chatgpt/AIOS-V2R1C1-BASELINE-REGISTRY-20260721",
"base_ref": "refs/heads/main",
"base_oid": "3cf9958cb94614449e093639cbf902216d73447f",
"manifest_commit_oid": "a56caf46f6bc335ec2395b674ead4b2a427c5edd",
"baseline_commit_oid": "6dbd5cbe07d2960ca653049fefbcf3a3e18f9ab1",
"current_baseline_before": "V1R4C6",
"current_baseline_after": "V2R1C1",
"reviewer": "gpt-5.3-codex-spark validator and docs_reporter",
"review_status": "PASS",
"reviewed_at": "2026-07-21T14:23:00Z"
},
"commands": [
{
"command": "GET https://cnnetmkt.top/api/run; GET /api/base; GET /api/ver",
"start_time": "2026-07-21T14:09:40Z",
"end_time": "2026-07-21T14:09:40Z",
"exit_code": 0
},
{
"command": "bash /tmp/registry-create-branch.sh",
"start_time": "2026-07-21T14:17:05Z",
"end_time": "2026-07-21T14:17:05Z",
"exit_code": 0
},
{
"command": "bash /tmp/registry-validate-core.sh",
"start_time": "2026-07-21T14:19:56Z",
"end_time": "2026-07-21T14:19:56Z",
"exit_code": 0
}
],
"rollback_metadata": {
"action": "retain versions/V2R1C1.json and restore only baseline/current.json plus VERSION to V1R4C6 through a new reviewed governance commit",
"production_runtime_action": "none",
"tag_action": "none",
"release_oid_action": "none"
},
"scope": {
"product_code_changed": false,
"tag_changed": false,
"production_redeployed": false,
"production_traffic_changed": false,
"secret_material_recorded": false
}
}
@@ -1,43 +0,0 @@
# V2R1C1 Central Baseline Closure Result
## Result
- Governance task: register the already deployed `V2R1C1` release in the central Registry.
- Baseline: `V1R4C6` -> `V2R1C1`.
- Release OID and tag OID: `4e0e6dcb90e6732aa9838fe9d0af0b34746829b0`.
- Companion OID: `12d1c13170e0c5a691d67a9408e739e7a17064a3`.
- Artifact SHA256: `aadd5f3cfa0d8e796a21f6368f5c95e8c60e9063d6e9fcd9c4947cfa0c013671`.
- Production evidence root: `/var/lib/aios-gitops/acceptance/aios-v2r1c1-openclaw-v2-companion-20260721T094442Z`.
- Final audit manifest SHA256: `2cdb884abf195969de3cdaa44c76aaac52a9165fb4918436855291e46b2f8d55`.
- Production status: `PRODUCTION_DEPLOYED`.
## Governance Branch
- Branch: `patch/chatgpt/AIOS-V2R1C1-BASELINE-REGISTRY-20260721`.
- Base ref: `refs/heads/main`.
- Base OID: `3cf9958cb94614449e093639cbf902216d73447f`.
- Manifest commit: `a56caf46f6bc335ec2395b674ead4b2a427c5edd`.
- Baseline commit: `6dbd5cbe07d2960ca653049fefbcf3a3e18f9ab1`.
- Review: `PASS` at `2026-07-21T14:23:00Z` by `gpt-5.3-codex-spark` validator and docs reporter.
## Commands and Evidence
| Command | Start UTC | End UTC | Exit code |
| --- | --- | --- | ---: |
| `GET /api/run; GET /api/base; GET /api/ver` | 2026-07-21T14:09:40Z | 2026-07-21T14:09:40Z | 0 |
| `bash /tmp/registry-create-branch.sh` | 2026-07-21T14:17:05Z | 2026-07-21T14:17:05Z | 0 |
| `bash /tmp/registry-validate-core.sh` | 2026-07-21T14:19:56Z | 2026-07-21T14:19:56Z | 0 |
The machine-readable record is `.agent/results/releases/V2R1C1-BASELINE-CLOSURE.json`. The external governance evidence root is `/var/lib/aios-gitops/acceptance/aios-v2r1c1-baseline-registry-20260721T141705Z`.
## Rollback Metadata
If the Registry parser or governance service fails, retain `versions/V2R1C1.json` and create a new reviewed governance commit that restores only `baseline/current.json` and `VERSION` to `V1R4C6`. Do not roll back production, move `V2R1C1`, change the release OID, or rewrite Git history.
## Scope Protection
- `product_code_changed=false`
- `tag_changed=false`
- `production_redeployed=false`
- `production_traffic_changed=false`
- `secret_material_recorded=false`
@@ -1,28 +0,0 @@
{
"status": "V2R1C2_PRODUCTION_RELEASED",
"release_run_id": "AIOS-V2R1C2-PRODUCTION-20260722T011836Z",
"tag_name": "V2R1C2",
"tag_oid": "a3bd84ebec9c705f7c9c5c03f7f4ec55f8854e36",
"target_oid": "3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24",
"evidence_closure_oid": "8cdca463aff9707edee00e20a3b56884130f6978",
"production_service": "aios-v2r1c2-production.service",
"production_port": 18920,
"candidate_port": 18921,
"production_acceptance": "PASS",
"role_matrix": "PASS",
"tenant_isolation": "PASS",
"api_smoke": "PASS",
"ui_smoke": "PASS",
"runtime_audit": "PASS",
"wechat_acceptance": "SYNTHETIC_ONLY",
"real_user_impersonated": false,
"production_acceptance_sha256": "a6c35e12cab7907ab6bcfd37d781b3d216f05a5ae002aa54a3112c70f43b991a",
"production_online_smoke_sha256": "fe9c6964e440f8871f2eb225f0132f70c634923389f7b8dfa85f615b100980a4",
"production_health_sha256": "ee8084e8b57d188dc0039a276f4f3f1ed7025ff58ce91b111cc765627eb52d4c",
"rollback_version": "V2R1C1",
"rollback_ready": true,
"main_changed_by_tag": false,
"force_push": false,
"promoted_at": "2026-07-22T01:40:51Z",
"evidence_path": "/var/lib/aios-gitops/acceptance/AIOS-V2R1C2-PRODUCTION-20260722T011836Z"
}
@@ -0,0 +1,44 @@
{
"b3": "BLOCKED_EVIDENCE_REQUIRED",
"b4": "NOT_STARTED",
"baseline": "V1R4C6",
"contract_branch": "task/governance/v2r1-personal-weixin-production-contract-20260717",
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"dispatch_commit": null,
"executor_quorum": {
"candidates": ["openclaw", "codex"],
"minimum_healthy": 1,
"openclaw": "UNKNOWN",
"codex": "UNKNOWN"
},
"active_executor": null,
"standby_executor": null,
"central_gpt": "UNKNOWN",
"automation_closed": {
"central_gpt_health": "UNKNOWN",
"executor_quorum": "UNKNOWN",
"active_executor": null,
"event_trigger": "UNKNOWN",
"polling": "UNKNOWN",
"failover": "UNKNOWN",
"lifecycle_sync": "UNKNOWN",
"two_cycle_unattended_loop": "UNKNOWN",
"restart_recovery": "UNKNOWN",
"result": "NOT_STARTED"
},
"lifecycle_events": [],
"production_release": "NOT_STARTED",
"release": "paused",
"repair3": "PASS_CANDIDATE_PREPARED",
"runner": "BLOCKED_GATEWAY_SECRET_REF_UNAVAILABLE",
"schema_version": 2,
"single_state_writer": "gateway",
"single_state_writer_lease_required": true,
"quorum_controller": {
"implementation": "CODE_COMMITTED_TESTED_PENDING_LIVE_ACCEPTANCE",
"live_acceptance": "NOT_STARTED",
"health_source": "ROOT_OWNED_METADATA_ONLY_LIVE_PROBE_RECEIPTS",
"unknown_counts_as_healthy": false
},
"updated_at": "2026-07-18T04:58:00+00:00"
}
@@ -0,0 +1,69 @@
# Codex task: apply and verify AIOS Orchestrator Gateway v1
Contract: `cbe33ecab8e9794ba5f318bea43aeaf447a56fbc`
Use the strongest actually available Codex model for the initial security review.
If unavailable or timed out, record TIMEOUT/NOT_EXECUTED and let OpenClaw run
the deterministic test path. Do not count a timeout as PASS.
## Apply
1. Create a new clean append-only branch and isolated worktree.
2. Validate the delivery tarball and `SHA256SUMS`.
3. Run the external `scripts/apply_gateway_patch.sh` with `SOURCE_DIR` and
`WORKTREE`; do not alter unrelated paths.
4. Review all added code, especially:
- exact contract and optimistic state-version checks;
- event/decision idempotency;
- SQLite leases and retry behavior;
- Workspace Agent `202` asynchronous callback flow;
- Responses background creation and polling;
- file-inbox and Git-ref polling;
- traversal/symlink protections;
- OpenClaw lifecycle synchronization;
- systemd sandbox;
- install and rollback transactions.
## Tests
Run from the final source tree:
- Python compile for all gateway and helper files;
- unit tests;
- JSON schema/config parsing;
- Bash syntax;
- `systemd-analyze verify`;
- event duplicate and payload-conflict tests;
- stale state-version rejection;
- lease/double-claim test;
- Workspace trigger failure and Responses fallback;
- response queued/in-progress/completed polling;
- file-inbox recovery after HTTP failure;
- Git-ref initial baseline and subsequent-change event;
- task start/end and project summary synchronization;
- dispatch traversal and symlink rejection;
- retry/backoff/dead-letter;
- install dry-run;
- rollback test in an isolated fixture;
- `git diff --check`;
- full repository tests;
- final Gitleaks.
## Commit and review
1. Create an append-only implementation commit.
2. Push without force.
3. Confirm `WORKTREE_OID == BARE_OID == GITEA_OID`.
4. Create an independent RESULT-only branch and exact-one result commit.
5. Record the source bundle SHA-256, implementation OID, RESULT OID, tests,
Gitleaks, systemd validation and remaining external configuration blockers.
## Forbidden
No reset, amend, rebase, force push, direct main push, controller mutation,
secret disclosure, fabricated test result, production deployment or B3 evidence
fabrication.
Success:
`GATEWAY_CODE_REVIEW=PASS IMPLEMENTATION=<oid> RESULT=<oid> OID_MATCH=YES`
@@ -0,0 +1,13 @@
# Codex V2R1 adaptive execution
CONTRACT_ACK=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc
Read `.agent/policies/V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT.yaml`.
Execute P1-POST-CONTRACT-REACK now:
- read-only verify Repair3 e110cda9968cadc5c7b7179ec5003b05aad11ce4
- read-only verify RESULT 49a9c2e0cc898ecf2bbd45f454e007644d7f3045
- verify lineage from 8d9c6da90fc8caa69b691d243c6fefb40b6b0575, exact4, tests, Gitleaks and RESULT-only scope
- do not modify Repair3 code in this unit
Act as fallback for OpenClaw only on timeout/unavailability. Use GPT-5.6 only if actually exposed; otherwise strongest visible model. Prefer OpenClaw when quality is comparable or volume is high; prefer Codex for complex/security-critical work.
Forbidden: reset, amend, rebase, force push, direct main push, controller mutation, secret disclosure, fabricated evidence, skipped gates.
Report:
CONTRACT=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc WORK_UNIT=<id> TRACK=<id> GATE=<id> AGENT=codex STATUS=<status> IMPLEMENTATION=<oid-or-NONE> RESULT=<oid-or-NONE> OID_MATCH=<YES-or-NO> BLOCKER=<reason-or-NONE>
Keep BASELINE=V1R4C6 RELEASE=paused PRODUCTION_RELEASE=NOT_STARTED until production gate passes.
@@ -1,267 +0,0 @@
# 执行 GitOps 独立 Worktree 与统一运行身份迁移
```yaml
task_id: TASK-GITOPS-ISOLATED-WORKTREE-MIGRATION-20260714T233742UTC8
created_at: 2026-07-14T23:37:42+08:00
timezone: Asia/Shanghai
owner_approved: true
user_action_required: false
risk: high
production_business_change: false
```
## 目标
根治共享 worktree 与多身份写入导致的 ACL、checkout 和未跟踪文件覆盖问题。
实施:
1. 统一 GitOps/OpenClaw 写入身份为 `aiman:gitops`;
2. 统一 `UMask=0002`;
3. 为 GitOps 根目录设置 setgid 与默认 ACL;
4. 建立 bare repo 和每任务独立 worktree;
5. 启用任务启动前自检;
6. 验证 OpenClaw 与 ChatGPT 并发任务互不干扰;
7. 保留原共享 workspace 作为只读回滚源,禁止智能体继续在其中切换分支。
## 强制通知
收到任务后立即微信通知 Owner:
```text
【任务已接收】GitOps独立worktree与统一运行身份迁移开始。将暂停写入服务、建立独立任务工作树并验证并发执行;生产业务版本不变。
```
任务结束后无论 PASS、BLOCKED 或 ROLLED_BACK,立即微信汇报结果并记录 message_id。
## 1. 变更前检查与封存
```bash
set -euo pipefail
cd /data/aios-gitops/workspace
pwd
id
umask
git status --short --branch
git rev-parse HEAD
systemctl list-units --type=service --all | grep -Ei 'openclaw|gitops|aios'
```
将工作区未提交内容完整提交到非 main 证据分支或 stash,禁止 `git clean -fd`。
记录当前运行服务、用户、组、umask、PID、工作区 HEAD 和未跟踪文件清单。
## 2. 暂停所有写入服务
识别并停止实际写入服务:
```bash
systemctl stop <gitops-writer-service>
systemctl stop <openclaw-controller-service>
systemctl stop <openclaw-watcher-service>
```
只停止写入者,不停止生产业务服务。记录每条命令和退出码。
## 3. 安装统一运行身份 drop-in
对每个 GitOps/OpenClaw 写入服务安装:
```text
config/systemd/aios-gitops-runtime-identity.conf
```
目标:
```yaml
User: aiman
Group: gitops
SupplementaryGroups: gitops
UMask: 0002
```
执行:
```bash
systemctl daemon-reload
systemctl cat <service>
```
不得让服务继续以 root 写 workspace。
## 4. 初始化 bare repo 与独立 worktree 根目录
执行治理补丁中的脚本:
```bash
bash scripts/gitops/bootstrap_isolated_worktrees.sh
```
验收:
```yaml
bare_repo: /data/aios-gitops/repo.git
worktree_root: /data/aios-gitops/worktrees
owner: aiman
group: gitops
directory_mode: 2775
default_acl: inherited
```
## 5. 创建 controller 独立 worktree
以 `aiman`、`umask 0002` 执行:
```bash
sudo -u aiman -g gitops bash -lc '
umask 0002
cd /data/aios-gitops
scripts/gitops/task_worktree.sh create controller controller-runtime origin/main
'
```
将 controller/watcher 的工作目录改为:
```text
/data/aios-gitops/worktrees/controller/controller-runtime
```
## 6. 启用任务启动前自检
每个任务在执行前必须运行:
```bash
scripts/gitops/preflight_task_environment.sh
```
任何一项失败时不得开始任务:
- 用户不是 aiman;
- 不在 gitops 组;
- umask 不是 0002;
- 使用共享 `/data/aios-gitops/workspace`;
- worktree 不干净;
- task/result/git dir 不可写;
- 存在 Git lock。
## 7. 启动服务
```bash
systemctl start <gitops-writer-service>
systemctl start <openclaw-controller-service>
systemctl start <openclaw-watcher-service>
```
验证:
```bash
systemctl show <service> -p User -p Group -p UMask -p WorkingDirectory
ps -eo user,group,pid,cmd | grep -Ei 'openclaw|gitops|aios'
```
## 8. 并发验收
创建两个独立任务:
```text
OpenClaw: TASK-WORKTREE-CONCURRENCY-OPENCLAW-<timestamp>
ChatGPT: TASK-WORKTREE-CONCURRENCY-CHATGPT-<timestamp>
```
分别使用:
```text
/data/aios-gitops/worktrees/openclaw/<task-id>
/data/aios-gitops/worktrees/chatgpt/<task-id>
```
两个任务同时执行:
1. 写各自 `.agent/tasks/<agent>/`;
2. 写各自 `.agent/results/<agent>/`;
3. 创建深层子目录,验证默认 ACL;
4. commit;
5. push;
6. 不切换共享 workspace 分支。
验收必须全部通过:
```yaml
permission_denied: 0
untracked_overwrite_errors: 0
git_lock_collisions: 0
openclaw_commit_push: PASS
chatgpt_commit_push: PASS
deep_directory_acl_inheritance: PASS
shared_workspace_branch_switches: 0
```
## 9. 回滚
若迁移失败:
1. 停止新服务;
2. 保留失败 worktree 供审计;
3. 恢复 systemd drop-in 备份;
4. `daemon-reload`;
5. 恢复原服务状态;
6. 不删除 `/data/aios-gitops/workspace`;
7. 生产业务保持不变。
## 10. 结果文件
提交:
```text
.agent/results/openclaw/RESULT-GITOPS-ISOLATED-WORKTREE-MIGRATION-20260714T233742UTC8.md
```
首段必须包含:
```yaml
status: PASS | BLOCKED | ROLLED_BACK
task_id: TASK-GITOPS-ISOLATED-WORKTREE-MIGRATION-20260714T233742UTC8
created_at: 2026-07-14T23:37:42+08:00
timezone: Asia/Shanghai
user_action_required: false
runtime_identity:
user:
group:
umask:
bare_repo:
worktree_root:
controller_worktree:
services:
default_acl:
preflight:
concurrency_test:
openclaw_branch:
openclaw_commit:
chatgpt_branch:
chatgpt_commit:
permission_denied_count:
untracked_overwrite_count:
git_lock_collision_count:
owner_notifications:
received_message_id:
terminal_message_id:
rollback:
remaining_risks:
next_step:
```
提交并推送到独立执行分支。禁止直接推 main。
## 完成微信模板
PASS:
```text
【任务完成】GitOps独立worktree迁移已通过。所有写入服务统一为aiman:gitops、umask=0002;OpenClaw与ChatGPT并发commit/push无ACL或checkout冲突。生产业务版本未变化。
```
阻塞:
```text
【任务阻塞】GitOps独立worktree迁移未通过。原因=<具体服务/权限/并发验收>;已保留原workspace和回滚能力,生产业务未变化。
```
@@ -0,0 +1,119 @@
# OpenClaw A0 Three-Party Self-Healing and Release Supervision
CONTRACT_ACK=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc
## Mission
Close the GPT/OpenClaw/Codex unattended loop before any main merge or production release. OpenClaw is the default state writer and long-running supervisor, but work may be reassigned dynamically under the shared contract.
## Authoritative inputs
- Governance contract: `cbe33ecab8e9794ba5f318bea43aeaf447a56fbc`
- Governance dispatch head before this task: `9195470ceb78fd57ffface3dcc38f40b75efa9d8`
- Gateway integration implementation: `c3984ca2b4388875d2875b4fdfdf95a2984f6ab3`
- Gateway integration RESULT: `4e63b1559aa3fc07b1384154c97b2c90b84353ed`
- Three-party self-healer candidate branch: `patch/a0-three-party-self-healing-20260718-20260718T043256`
- Production baseline: `V1R4C6`
- Release status: `paused`
## Priority invariant
Continuously evaluate exactly three parties:
1. `central_gpt` — Orchestrator Gateway plus a valid Workspace Agent or Responses API decision channel.
2. `openclaw` — OpenClaw watcher, lifecycle sync, durable task execution and heartbeat.
3. `codex` — Codex watcher, auto-claim, lifecycle sync, durable task execution and heartbeat.
When exactly two parties are HEALTHY and the third is explicitly UNHEALTHY, create a P0 repair work unit for the unique unhealthy party before every other task. UNKNOWN is not UNHEALTHY and must fail closed. Do not repair when fewer than two are healthy or any health result is ambiguous.
## A0-SH1: supervise implementation completion
Codex owns the high-risk code changes. Independently verify that the final implementation contains:
- strict root-owned configuration validation and no symlink chain;
- fixed built-in party adapters rather than arbitrary untrusted repair commands;
- bounded stdout/stderr redaction so tokens and secrets cannot enter audit logs;
- exclusive lease, cooldown, circuit breaker and stale lease recovery;
- event trigger plus periodic polling;
- OpenClaw lifecycle events: `TASK_DISPATCHED`, `TASK_STARTED`, `TASK_ENDED`, `DECISION_APPLIED`, `PROJECT_SUMMARY`, `PROJECT_COMPLETED`;
- durable file outbox fallback when HTTP callback is unavailable;
- systemd service and timer;
- health and repair adapters for central GPT, OpenClaw and Codex;
- install, verify and rollback scripts;
- deterministic tests for all health combinations and failures.
Reject the current candidate if it only contains `automation/three_party_self_healer.py` without these controls.
## A0-SH2: live three-party validation
After Codex implementation RESULT passes, deploy only to a non-production acceptance context and run:
1. All-three healthy steady state for at least three poll intervals.
2. GPT unhealthy while OpenClaw and Codex healthy: GPT repair becomes P0 and recovers.
3. OpenClaw unhealthy while GPT and Codex healthy: OpenClaw repair becomes P0 and recovers.
4. Codex unhealthy while GPT and OpenClaw healthy: Codex repair becomes P0 and recovers.
5. One healthy plus two unhealthy: no repair; fail closed.
6. Any UNKNOWN health: no repair; fail closed.
7. Repair timeout: retry with backoff, then circuit open.
8. Service restart: state, lease and pending work recover.
9. HTTP event path unavailable: file inbox polling recovers.
10. Git webhook unavailable: Git ref polling detects change.
11. Duplicate event ID: no duplicate dispatch.
12. Main Workspace Agent channel unavailable: Responses fallback takes over only with a valid approved credential.
13. OpenClaw receives every task start/end and project summary notification.
14. Two consecutive unattended decision/execution/result cycles complete with no chat message or manual dispatch.
Do not use dummy API keys or environment-variable presence as proof of a valid central decision channel.
## A0-SH3: authoritative closure
Only after independent verification of the above create one append-only state update and an exact-one RESULT file. Required output:
```text
AUTOMATION_CLOSED=PASS
CENTRAL_GPT_HEALTH=PASS
OPENCLAW_AUTO_CLAIM=PASS
CODEX_AUTO_CLAIM=PASS
THREE_PARTY_SELF_HEALING=PASS
EVENT_TRIGGER=PASS
POLLING=PASS
OPENCLAW_LIFECYCLE_SYNC=PASS
TWO_CYCLE_UNATTENDED_LOOP=PASS
RESTART_RECOVERY=PASS
ROLLBACK=PASS
```
Any missing item keeps `AUTOMATION_CLOSED=NO`.
## Release continuation
After and only after `AUTOMATION_CLOSED=PASS`, automatically resume the governed release sequence:
1. Real B3 evidence intake and independent review.
2. B4 implementation and validation.
3. Full integration, regression, security and migration tests.
4. Normal PR/GitOps merge to main; never direct-push main.
5. Immutable release artifact, digest, SBOM and signature.
6. Staging and rollback exercise.
7. Canary and monitoring validation.
8. Owner/admin GitOps production deployment using existing approved release controls.
9. Production smoke tests and production RESULT-only commit.
External facts that cannot be fabricated — real WeChat credentials/raw-wire, valid secret references and owner-only release approval — remain fail-closed blockers. The loop must monitor them and resume automatically when they become available.
## Forbidden
- reset, amend, rebase, force push;
- direct push to main;
- controller mutation;
- secret disclosure;
- dummy credentials presented as production credentials;
- fabricated B3 evidence;
- timeout or skipped checks recorded as PASS;
- production release before `AUTOMATION_CLOSED=PASS`.
Progress format:
```text
CONTRACT=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc WORK_UNIT=<id> AGENT=openclaw STATUS=<status> COMMIT=<oid-or-NONE> BLOCKER=<stable-reason-or-NONE>
```
@@ -0,0 +1,68 @@
# OpenClaw task: deploy and supervise AIOS Orchestrator Gateway v1
Contract: `cbe33ecab8e9794ba5f318bea43aeaf447a56fbc`
You are the default orchestrator and deployment supervisor.
## Preconditions
- Codex has applied and independently reviewed this source bundle on a clean,
append-only branch.
- Implementation and RESULT OIDs are identical across worktree, bare repo and
Gitea.
- Final Gitleaks and repository tests are PASS.
- No force push, reset, amend or rebase occurred.
## Responsibilities
1. Validate the source bundle SHA-256 and `SHA256SUMS`.
2. Review the generated systemd service, poll timer, tmpfiles and sysusers files.
3. Install through `scripts/install_aios_orchestrator_gateway.sh`.
4. Configure one or both central-model channels:
- published Workspace Agent API channel with `agtch_...` trigger ID and a
Workspace Agent access token stored outside Git; or
- Responses API fallback with an approved model and API key stored outside Git.
5. Publish the callback endpoint securely and configure the Workspace Agent
action from `workspace-agent/orchestrator-decision-action.openapi.yaml`.
6. Confirm that the Workspace Agent instructions match
`workspace-agent/CENTRAL_AGENT_INSTRUCTIONS.md`.
7. Enable both:
- the continuously running gateway service;
- the independent `aios-orchestrator-gateway-poll.timer`.
8. Verify HTTP event ingestion, file-inbox polling, Git-ref polling,
Workspace Agent trigger acceptance, callback decision, Responses background
polling fallback, dispatch queue generation, retry leases and loop guards.
9. Verify OpenClaw receives and synchronizes:
- TASK_DISPATCHED
- TASK_STARTED
- TASK_ENDED
- PROJECT_SUMMARY
- PROJECT_COMPLETED
10. Run a harmless bootstrap work unit through OpenClaw and Codex and confirm
duplicate event submission is idempotent.
11. Run rollback, reinstall and heartbeat/health verification.
12. Create one RESULT-only review recording exact commits, artifact digest,
tests, systemd security, polling recovery, lifecycle synchronization and
rollback.
## Fail-closed boundaries
- Do not put Workspace Agent tokens, API keys, callback tokens or gateway
credentials in Git, logs, command arguments or RESULT files.
- Do not claim the gateway resolves the existing B3 credential/raw-wire blocker.
- Do not advance B4 or production merely because this gateway is healthy.
- Do not bypass owner/admin production release approvals.
- Do not use reset, amend, rebase, force push, direct main push or bare-ref edits.
## Required final status
Success:
`ORCHESTRATOR_GATEWAY_LIVE=PASS POLLING=PASS OPENCLAW_SYNC=PASS ROLLBACK=PASS`
Blocker:
`ORCHESTRATOR_GATEWAY_LIVE=BLOCKED reason=<stable_reason>`
Keep `BASELINE=V1R4C6 RELEASE=paused PRODUCTION_RELEASE=NOT_STARTED` until the
existing production contract advances all release gates.
@@ -0,0 +1,13 @@
# OpenClaw V2R1 adaptive orchestration
CONTRACT_ACK=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc
Read `.agent/policies/V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT.yaml`.
You are default orchestrator and sole authoritative state writer.
Execute immediately:
1. R0-DISCOVER-SECRET-PROVIDER: discover approved provider metadata and owner-probe source without reading or outputting secrets. Environment variables alone are not approved secret refs.
2. P2-INVENTORY-REAL-EVIDENCE: inventory approved restricted B3 evidence locations; validate existing facts; never synthesize evidence.
3. Independently verify Codex results before advancing.
Prefer OpenClaw for comparable-quality, large, repetitive, long-running, infrastructure and cost-sensitive work. Prefer Codex strongest actually available model for complex code, security boundaries, architecture and high-risk review. On timeout mark TIMEOUT and switch agent; never infer PASS.
Forbidden: reset, amend, rebase, force push, direct main push, controller mutation, secret disclosure, fabricated evidence, skipped gates.
Report:
CONTRACT=cbe33ecab8e9794ba5f318bea43aeaf447a56fbc WORK_UNIT=<id> TRACK=<id> GATE=<id> AGENT=openclaw STATUS=<status> COMMIT=<oid-or-NONE> BLOCKER=<reason-or-NONE>
Keep BASELINE=V1R4C6 RELEASE=paused PRODUCTION_RELEASE=NOT_STARTED until production gate passes.
+11
View File
@@ -0,0 +1,11 @@
# AIOS Codex automatic command watcher
The watcher polls the governed `commands/active` queue without a chat trigger. It uses SQLite transactions for exclusive leases, records the exact command digest, rejects contract drift, retries transient failures, dead-letters exhausted work, and resumes from the last durable phase so an end-event retry does not execute a work unit twice.
Commands are JSON documents validated by `schemas/codex-command.schema.json`. `acceptance_probe` is the only built-in action and has no external side effect. Production `codex` work requires an administrator-provisioned absolute executor path; the instruction is passed on standard input, never argv or environment. The executor receives a sanitized environment and must obtain credentials through its approved file or exec provider. If the executor is absent, the task fails closed and follows the retry/dead-letter policy.
The watcher provides at-most-once execution. It writes the `EXECUTING` phase before launching an external executor; if the host dies after an external effect but before a result is durably recorded, recovery does not replay the work and instead dead-letters the ambiguity for human review. Start and end event JSON is persisted before transmission, so response-loss retries reuse the exact event ID and payload and are deduplicated by the Gateway.
Lifecycle evidence is appended to `/var/log/aios-codex-watcher/audit.jsonl` as `CODEX_AUTO_CLAIM`, `TASK_STARTED`, `TASK_EXECUTED`, `TASK_ENDED`, and `EVENT_POSTED`. Inputs, tokens, stdout, and stderr are not logged; only result digests are retained.
Install from a reviewed commit with `sudo scripts/install_aios_codex_watcher.sh`. The installer creates a content-addressed release, atomically updates `current`, writes a receipt, and enables the five-second systemd timer. The existing Gateway inbound token file is referenced in place and never copied.
+107
View File
@@ -0,0 +1,107 @@
# AIOS Orchestrator Gateway v0.2
This component turns the current passive GPT orchestration pattern into a
durable, unattended loop:
```text
Gitea/OpenClaw/Codex event
-> HTTP event endpoint or polled file inbox
-> durable SQLite state machine
-> Workspace Agent API trigger
-> strict callback decision
-> Responses API background fallback and polling
-> OpenClaw/Codex dispatch queues
-> start/end/result events
-> next orchestration cycle
```
## Reliability mechanisms
The gateway does not depend on one timer, webhook or model call:
- HTTP event ingestion.
- Periodic file-inbox polling for missed webhooks and offline producers.
- Local bare-Git ref polling for `main`, governance, patch and result refs.
- Workspace Agent retries with a stable event `Idempotency-Key`.
- Workspace Agent callback timeout and Responses API background fallback.
- Responses status polling until completion or fail-closed terminal status.
- SQLite optimistic state versions, idempotency records and processing leases.
- Exponential retry with a maximum retry count and dead-letter state.
- Same-blocker and maximum-hop guards to stop non-progress loops.
- A systemd timer runs `--poll-once` as an independent reliability watchdog.
Workspace Agent trigger runs are asynchronous and do not expose a retrievable
final response. The central Workspace Agent must therefore call the
`submit_orchestration_decision` action.
## OpenClaw synchronization
Every lifecycle stage is durably written to the OpenClaw synchronization queue:
- `TASK_DISPATCHED`
- `TASK_STARTED`
- `TASK_ENDED`
- `DECISION_APPLIED`
- `PROJECT_SUMMARY`
- `PROJECT_COMPLETED`
An optional authenticated OpenClaw HTTP endpoint can receive the same records.
A failed HTTP delivery remains in the lifecycle retry queue. OpenClaw remains
the default project-status synchronizer and authoritative progress presenter.
## Agent execution contract
Each dispatched JSON envelope contains:
- exact contract commit;
- source event;
- clean base commit;
- primary and fallback agent;
- allowed paths;
- required checks;
- forbidden operations;
- completion event callback format.
OpenClaw and Codex must send a `work_unit_started` event before execution and a
`work_unit_completed`, `work_unit_blocked`, or `work_unit_failed` event when
they stop. This guarantees that OpenClaw can synchronize every task start,
finish, and overall project summary.
## Security
- Model and callback credentials are read only from root-owned secret files.
- Credentials never enter Git, dispatch payloads or logs.
- The gateway runs as a dedicated unprivileged user.
- Dispatch and sync paths are protected against traversal and symlink targets.
- All state changes require the exact contract commit and current state version.
- Production evidence, credentials and approvals stay fail-closed.
- Git history operations remain controlled by the committed governance contract.
## Deployment sequence
1. Apply this source bundle to a clean append-only branch.
2. Run unit, repository, shell, JSON, systemd and Gitleaks checks.
3. Configure one or both model channels:
- published Workspace Agent API channel and Workspace Agent access token;
- Responses API key and approved fallback model.
4. Publish the decision callback as a Workspace Agent action or MCP tool.
5. Install the service through `scripts/install_aios_orchestrator_gateway.sh`.
6. Run `scripts/verify_aios_orchestrator_gateway.sh`.
7. Submit a harmless event and verify:
- trigger acceptance;
- strict decision callback;
- dispatch file;
- OpenClaw task-start/task-end/project-summary synchronization;
- duplicate idempotency;
- polling recovery after suppressing the HTTP webhook.
The installer copies an explicit gateway-only allowlist into a release directory
named by its SHA-256 digest, atomically updates `/opt/aios-orchestrator-gateway/current`,
and persists `/var/lib/aios-orchestrator-gateway/install-receipt.json` with the
current and previous release. `--dry-run` performs source validation and emits
`AIOS_ORCHESTRATOR_INSTALL_DRY_RUN=PASS` without changing installed state.
Rollback reads that root-owned `0600` receipt, verifies both release directory
names and `SOURCE_DIGEST` files, coordinates the gateway and poll units, and
atomically switches `current` to the previous release. Repeating the same
rollback is a no-op and emits `NOOP_ALREADY_ROLLED_BACK`.
+1 -1
View File
@@ -1 +1 @@
V2R1C2
V1R4C6
+66 -494
View File
@@ -5,81 +5,28 @@ from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from pathlib import Path
from typing import Any
try:
from aios_gitops_write_guard import evaluate_git_write
except ModuleNotFoundError:
sys.path.insert(0, str(Path(__file__).resolve().parent))
from aios_gitops_write_guard import evaluate_git_write
from gitea_tag_publication import (
GiteaTagClient,
GitTagBackend,
PublicationConflict,
PublicationPending,
PublicationManualOwnerReviewRequired,
PublicationReconciliationRequired,
PublicationRequest,
TagPublisher,
redact,
validate_version,
)
sys.path.insert(0, str(Path(os.getenv('AIOS_GITOPS_WORKSPACE', '/data/aios-gitops/workspace')) / 'runtime' / 'agent-os'))
from orchestrator.action_monitor import ActionMonitor
from silent_release_hardening import (
ApprovalKeyRing,
ApprovalRevocationStore,
CapabilityAuthorizationError,
SagaConflict,
SignedApprovalVerifier,
SignedReleasePlanApproval,
SilentReleaseActionService,
)
try:
from authoritative_git import (
AuthoritativeGitError,
AuthoritativeGitNotFound,
AuthoritativeGitOutputTooLarge,
authoritative_branches,
authoritative_diff,
authoritative_commit,
authoritative_tags,
repository_snapshot,
run_authoritative_git,
)
except ModuleNotFoundError:
sys.path.insert(0, str(Path(__file__).resolve().parent))
from authoritative_git import (
AuthoritativeGitError,
AuthoritativeGitNotFound,
AuthoritativeGitOutputTooLarge,
authoritative_branches,
authoritative_diff,
authoritative_commit,
authoritative_tags,
repository_snapshot,
run_authoritative_git,
)
ROOT = Path(os.getenv('AIOS_GITOPS_ROOT', '/data/aios-gitops'))
WS = Path(os.getenv('AIOS_GITOPS_WORKSPACE', str(ROOT / 'workspace')))
sys.path.insert(0, str(WS))
from services.aios_orchestrator_gateway.action_bridge import (
CallbackBridgeError,
submit_orchestration_decision,
)
SECRETS = Path(os.getenv('AIOS_GITOPS_SECRETS_DIR', str(ROOT / 'runtime' / 'secrets')))
TOKEN_FILE = Path(os.getenv('AIOS_GITOPS_TOKEN_FILE', str(SECRETS / 'gitops.tokens')))
HOST = os.getenv('AIOS_GITOPS_HOST', '127.0.0.1')
PORT = int(os.getenv('AIOS_GITOPS_PORT', '18890'))
OPENAPI = WS / 'docs' / 'gitops' / 'openapi-chatgpt-actions.yaml'
ORCHESTRATOR_OPENAPI = WS / 'workspace-agent' / 'orchestrator-decision-action.openapi.yaml'
ORCHESTRATOR_CONTRACT_COMMIT = os.getenv('AIOS_ORCHESTRATOR_CONTRACT_COMMIT', '')
ORCHESTRATOR_CALLBACK_SECRET_REF = os.getenv('AIOS_ORCHESTRATOR_CALLBACK_SECRET_REF', '')
_resolver = os.getenv('AIOS_ORCHESTRATOR_SECRET_REF_RESOLVER', '')
ORCHESTRATOR_SECRET_REF_RESOLVER = Path(_resolver) if _resolver else None
BARE_REPO = Path(os.getenv('AIOS_GITOPS_BARE_REPO', str(ROOT / 'runtime' / 'gitea-bin' / 'data' / 'repositories' / 'aios' / 'aios.git')))
AUTHORITATIVE_GIT_DIR = Path(os.getenv('AIOS_GITOPS_AUTHORITATIVE_GIT_DIR')) if os.getenv('AIOS_GITOPS_AUTHORITATIVE_GIT_DIR') else None
REPOSITORY_IDENTITY = os.getenv('AIOS_GITOPS_REPOSITORY_IDENTITY', '')
SILENT_RELEASE_STATE_ROOT = Path(os.getenv(
'AIOS_SILENT_RELEASE_STATE_ROOT',
'/var/lib/aios-gitops/silent-release',
))
SILENT_RELEASE_EXECUTION_ENABLED = str(os.getenv(
'AIOS_SILENT_RELEASE_EXECUTION_ENABLED', 'false'
)).lower() in {'1', 'true', 'yes', 'on'}
SILENT_RELEASE_PUBLIC_KEYS = Path(os.getenv(
'AIOS_SILENT_RELEASE_PUBLIC_KEYS',
'/etc/aios-gitops/silent-release-owner-public-keys.json',
))
monitor = ActionMonitor()
BASE_MODELS = [
@@ -88,43 +35,7 @@ BASE_MODELS = [
{'id':'base.gpt-5.5','provider':'openai','model':'gpt-5.5','display_name':'GPT-5.5','scope':'global'},
]
SECRET_RE = re.compile(r'(?i)(api[_-]?key|secret[_-]?access[_-]?key|secret|password|token)\s*[:=]\s*["\']?([A-Za-z0-9_./+=:-]{16,})')
SECRET_VAL_RE = re.compile(
r'(?<![A-Za-z0-9_])'
r'(sk-[A-Za-z0-9_-]{20,}|HPUA[A-Z0-9]{10,}|SDC[A-Za-z0-9]{20,})'
)
READ_ONLY_GET_PATHS = frozenset({
'/', '/health', '/api/health', '/api/openapi.yaml',
'/api/openapi-chatgpt-actions.yaml', '/api/schema', '/api', '/api/status',
'/api/base', '/api/baseline', '/base', '/api/ver', '/api/versions', '/ver',
'/api/diff', '/diff', '/api/patch', '/api/patches', '/api/branches', '/patch',
'/api/model', '/model', '/api/run', '/api/runtime-status', '/run',
'/api/action-monitor', '/action-monitor', '/admin/ai-evolution',
'/admin/ai-evolution/api/data',
'/api/release-consistency', '/api/tag-publication-status',
'/api/release-acceptance',
'/getReleaseCandidate', '/getReleasePlan', '/getReleaseAcceptance',
'/getApprovalEnvelope', '/getReleaseSaga', '/getCapabilityStatus',
'/getRecoveryStatus',
})
SILENT_RELEASE_POST_PATHS = frozenset({
'/createReleaseCandidate',
'/createReleaseIntent',
'/submitSignedApproval',
'/submitOwnerApproval',
'/startDeployment',
})
def reject_non_read_http(handler, method: str, path: str) -> bool:
if method == 'GET':
return False
if path in SILENT_RELEASE_POST_PATHS:
return False
decision = evaluate_git_write(WS)
if decision.allowed:
return False
handler.sendj(decision.payload(), decision.status_code)
return True
SECRET_VAL_RE = re.compile(r'(sk-[A-Za-z0-9_-]{20,}|HPUA[A-Z0-9]{10,}|SDC[A-Za-z0-9]{20,})')
def now() -> str: return time.strftime('%Y-%m-%dT%H:%M:%SZ', time.gmtime())
def uid(p='id') -> str: return f'{p}-{int(time.time()*1000)}-{uuid.uuid4().hex[:8]}'
@@ -136,41 +47,6 @@ def sh(args, check=False):
def jload(p: Path, default):
try: return json.loads(p.read_text(encoding='utf-8'))
except Exception: return default
def _file_sha256(path: Path) -> str:
digest=hashlib.sha256()
with path.open('rb') as source:
for chunk in iter(lambda: source.read(1024 * 1024), b''):
digest.update(chunk)
return digest.hexdigest()
def _build_silent_release_service() -> SilentReleaseActionService:
verifier=None
if SILENT_RELEASE_PUBLIC_KEYS.is_file():
verifier=SignedApprovalVerifier(
ApprovalKeyRing.from_file(SILENT_RELEASE_PUBLIC_KEYS),
ApprovalRevocationStore(
SILENT_RELEASE_STATE_ROOT / 'revocations'
),
)
policy_path=WS / 'config' / 'gitops' / 'silent-release-policy.json'
policy_sha=_file_sha256(policy_path) if policy_path.is_file() else ''
return SilentReleaseActionService(
SILENT_RELEASE_STATE_ROOT,
execution_enabled=SILENT_RELEASE_EXECUTION_ENABLED,
approval_verifier=verifier,
expected_repository=REPOSITORY_IDENTITY,
policy_sha256=policy_sha,
)
# Construct once during service startup. If production execution is enabled
# without every trusted builder/orchestrator dependency, construction fails
# closed before the HTTP server accepts traffic.
_SILENT_RELEASE_SERVICE = _build_silent_release_service()
def silent_release_service() -> SilentReleaseActionService:
return _SILENT_RELEASE_SERVICE
def relpath(path: str) -> str:
x = str(path or '').replace('\\','/').strip()
if not x or x.startswith('/') or '..' in Path(x).parts or Path(x).parts[0] == '.git': raise ValueError('invalid path')
@@ -244,13 +120,9 @@ def ensure_git():
if not sh(['config','user.name'])['stdout']: sh(['config','user.name','chatgpt-gitops-bot'], True)
def checkout(branch, base='main', create=False):
ensure_git(); b=branch_name(branch)
exists = sh(['rev-parse','--verify',f'refs/heads/{b}'])['code'] == 0
if exists:
sh(['checkout',b], True)
elif create:
sh(['checkout',base], True); sh(['checkout','-b',b], True)
else:
raise ValueError(f'branch does not exist: {b}')
if create or sh(['rev-parse','--verify',b])['code'] != 0:
sh(['checkout',base], True); sh(['checkout','-B',b], True)
else: sh(['checkout',b], True)
return b
def write_file(path, content):
if content_has_secret(content): raise ValueError('secret_rejected: use secret_ref; plaintext API keys must not enter Git')
@@ -260,19 +132,6 @@ def sync_branch(branch: str) -> dict:
if not branch or branch in ('main','master') or branch.startswith('main/') or branch.startswith('master/'):
return {'status':'skipped','reason':'protected_branch','branch':branch}
if BARE_REPO.exists():
common = sh(['rev-parse', '--path-format=absolute', '--git-common-dir'], True)['stdout']
if Path(common).resolve() == BARE_REPO.resolve():
head = sh(['rev-parse', 'HEAD'], True)['stdout']
stored = subprocess.run(
['git', '--git-dir', str(BARE_REPO), 'rev-parse', f'refs/heads/{branch}'],
text=True, capture_output=True,
)
if stored.returncode != 0 or stored.stdout.strip() != head:
raise RuntimeError(f'linked-worktree branch verification failed for {branch}')
return {
'status':'verified', 'target':'authoritative-bare',
'branch':branch, 'oid':head,
}
ref=f'refs/heads/{branch}:refs/heads/{branch}'
p=subprocess.run(['git','--git-dir',str(BARE_REPO),'fetch',str(WS),ref], text=True, capture_output=True)
if p.returncode != 0: raise RuntimeError(p.stderr or p.stdout or f'git sync failed for {branch}')
@@ -282,68 +141,15 @@ def sync_branch(branch: str) -> dict:
return {'status':'synced','target':'origin','branch':branch,'stdout':p.stdout.strip(),'stderr':p.stderr.strip()}
def sync_tag(tag: str) -> dict:
raise RuntimeError(
'unsafe sync_tag disabled: use TagPublisher with expected tag_oid, '
'target_oid, Gitea remote verification, and authenticated API verification'
)
def resolve_commit_oid(value: str | None, field: str = 'target_oid') -> str:
oid = str(value or '').strip().lower()
if not re.fullmatch(r'[0-9a-f]{40}', oid):
raise ValueError(f'{field} must be a full 40-character commit OID')
try:
return authoritative_commit(oid, git_dir=AUTHORITATIVE_GIT_DIR)
except (AuthoritativeGitError, ValueError) as exc:
raise ValueError(f'{field} does not resolve to the exact approved commit') from exc
def existing_tag_target(tag: str) -> str | None:
if not re.fullmatch(r'V\d+R\d+C\d+(?:-[A-Za-z0-9_.-]+)?$', str(tag or '')):
raise ValueError('invalid release version')
try:
result = run_authoritative_git(
['rev-parse', '--verify', f'refs/tags/{tag}^{{commit}}'],
check=False,
git_dir=AUTHORITATIVE_GIT_DIR,
)
except AuthoritativeGitError as exc:
raise RuntimeError(str(exc)) from exc
return result.stdout.strip() if result.returncode == 0 else None
def publication_request_from_query(q: dict[str, list[str]]) -> PublicationRequest:
required = ('version', 'target_oid', 'evidence_closure_oid', 'approval_plan_id')
missing = [name for name in required if not (q.get(name) or [''])[0]]
if missing:
raise ValueError('missing release governance fields: ' + ', '.join(missing))
return PublicationRequest(
(q['version'])[0],
(q['target_oid'])[0].lower(),
(q['evidence_closure_oid'])[0].lower(),
(q['approval_plan_id'])[0],
)
def required_authoritative_git_dir() -> Path:
snapshot = repository_snapshot(
git_dir=AUTHORITATIVE_GIT_DIR,
repository_identity=REPOSITORY_IDENTITY,
)
if not snapshot["repository_available"]:
raise RuntimeError(str(snapshot["query_error"] or "AUTHORITATIVE_REPOSITORY_UNAVAILABLE"))
return Path(str(snapshot["authoritative_git_dir"]))
def publication_publisher() -> TagPublisher:
authoritative_repo = required_authoritative_git_dir()
return TagPublisher(
WS,
GitTagBackend(WS, authoritative_repo=authoritative_repo),
GiteaTagClient(
os.getenv('AIOS_GITOPS_GITEA_API_URL',''),
os.getenv('AIOS_GITOPS_GITEA_REPOSITORY',''),
Path(os.getenv('AIOS_GITOPS_GITEA_TOKEN_FILE','')),
token_line=int(os.getenv('AIOS_GITOPS_GITEA_TOKEN_LINE','1')),
),
os.getenv('AIOS_GITOPS_GITEA_REMOTE','gitea'),
)
if not tag: return {'status':'skipped','reason':'empty_tag'}
if BARE_REPO.exists():
ref=f'refs/tags/{tag}:refs/tags/{tag}'
p=subprocess.run(['git','--git-dir',str(BARE_REPO),'fetch',str(WS),ref], text=True, capture_output=True)
if p.returncode != 0: raise RuntimeError(p.stderr or p.stdout or f'git tag sync failed for {tag}')
return {'status':'synced','target':'bare','tag':tag,'stdout':p.stdout.strip(),'stderr':p.stderr.strip()}
p=subprocess.run(['git','push','origin',tag], cwd=str(WS), text=True, capture_output=True)
if p.returncode != 0: raise RuntimeError(p.stderr or p.stdout or f'git tag push failed for {tag}')
return {'status':'synced','target':'origin','tag':tag,'stdout':p.stdout.strip(),'stderr':p.stderr.strip()}
def commit(paths, msg):
ensure_git(); paths = [relpath(p) for p in paths] if paths else ['.']; sh(['add'] + paths, True)
@@ -363,16 +169,6 @@ def patches():
for p in sorted((WS/'patches').glob('**/*.json')): out.append(jload(p, {'path':str(p.relative_to(WS))}))
return out
def runtime_release_info() -> dict[str, Any]:
for candidate in (WS / 'manifest.json', WS.parent / 'manifest.json'):
data = jload(candidate, {})
if data.get('target_commit'):
return {
'runtime_release_commit': data.get('target_commit'),
'runtime_release_path': str(WS),
}
return {'runtime_release_commit': None, 'runtime_release_path': str(WS)}
class H(BaseHTTPRequestHandler):
server_version='AIOSChatGPTGitOps/V1R1C0'
def sendj(self, data, code=200):
@@ -394,236 +190,46 @@ class H(BaseHTTPRequestHandler):
if '*' not in a['scopes'] and scope not in a['scopes']: self.sendj({'error':'forbidden','required_scope':scope,'account':a['account']},403); return None
return a
def do_GET(self):
u=urllib.parse.urlparse(self.path); path=u.path.rstrip('/') if u.path != '/' else u.path
if path not in READ_ONLY_GET_PATHS: return self.sendj({'error':'not_found','path':path},404)
q=urllib.parse.parse_qs(u.query)
u=urllib.parse.urlparse(self.path); path=u.path.rstrip('/') if u.path != '/' else u.path; q=urllib.parse.parse_qs(u.query)
try:
if path in ('/health','/api/health'): return self.sendj({'status':'ok','service':'chatgpt-gitops-adapter','workspace':str(WS)})
if path in ('/api/openapi.yaml','/api/openapi-chatgpt-actions.yaml','/api/schema'):
return self.sendt(OPENAPI.read_text(encoding='utf-8'), 'application/yaml; charset=utf-8') if OPENAPI.exists() else self.sendj({'error':'schema_missing'},404)
if path == '/api/orchestrator-decision.openapi.yaml':
return self.sendt(ORCHESTRATOR_OPENAPI.read_text(encoding='utf-8'), 'application/yaml; charset=utf-8') if ORCHESTRATOR_OPENAPI.exists() else self.sendj({'error':'schema_missing'},404)
if path in ('/api','/api/status'): return self.sendj(api_spec())
if path in ('/api/base','/api/baseline','/base'):
data=jload(WS/'baseline'/'current.json', {})
data.setdefault('authoritative_repository', repository_snapshot(
git_dir=AUTHORITATIVE_GIT_DIR,
repository_identity=REPOSITORY_IDENTITY,
))
return self.sendj(data)
data=jload(WS/'baseline'/'current.json', {}); data.setdefault('git', sh(['rev-parse','--short','HEAD'])['stdout']); return self.sendj(data)
if path in ('/api/ver','/api/versions','/ver'):
snapshot = repository_snapshot(
git_dir=AUTHORITATIVE_GIT_DIR,
repository_identity=REPOSITORY_IDENTITY,
)
if not snapshot['repository_available']:
return self.sendj({
'repository_available': False,
'query_error': snapshot['query_error'],
'registry_versions': [p.name for p in (WS/'versions').glob('*')],
'local_authoritative_tags': [],
}, 503)
try:
tags = authoritative_tags(git_dir=AUTHORITATIVE_GIT_DIR)
except AuthoritativeGitError as exc:
return self.sendj({
'repository_available': False,
'query_error': str(exc),
'registry_versions': [p.name for p in (WS/'versions').glob('*')],
'local_authoritative_tags': [],
}, 503)
registry = [p.name for p in (WS/'versions').glob('*')]
return self.sendj({
'registry_versions': registry,
'local_authoritative_tags': tags,
'repository_available': True,
'query_error': None,
'tags': [item['name'] for item in tags],
'versions_dir': registry,
'head': snapshot['authoritative_head_oid'],
'authoritative_head_oid': snapshot['authoritative_head_oid'],
})
return self.sendj({'tags':sh(['tag','--list','V*'])['stdout'].splitlines(),'versions_dir':[p.name for p in (WS/'versions').glob('*')],'head':sh(['rev-parse','--short','HEAD'])['stdout']})
if path in ('/api/diff','/diff'):
frm=(q.get('from') or [''])[0]
to=(q.get('to') or [''])[0]
try:
diff=authoritative_diff(frm, to, git_dir=required_authoritative_git_dir())
except ValueError as exc:
return self.sendj({'error':redact(exc)},400)
except AuthoritativeGitNotFound as exc:
return self.sendj({'error':redact(exc)},404)
except AuthoritativeGitOutputTooLarge as exc:
return self.sendj({'error':redact(exc)},413)
except (AuthoritativeGitError, RuntimeError) as exc:
return self.sendj({'error':redact(exc)},503)
return self.sendj({'from':frm,'to':to,'diff':diff})
if path in ('/api/patch','/api/patches','/api/branches','/patch'):
snapshot = repository_snapshot(
git_dir=AUTHORITATIVE_GIT_DIR,
repository_identity=REPOSITORY_IDENTITY,
)
if not snapshot['repository_available']:
return self.sendj({
'repository_available': False,
'query_error': snapshot['query_error'],
'metadata_patches': patches(),
'authoritative_branches': [],
}, 503)
try:
branches = authoritative_branches(git_dir=AUTHORITATIVE_GIT_DIR)
except AuthoritativeGitError as exc:
return self.sendj({
'repository_available': False,
'query_error': str(exc),
'metadata_patches': patches(),
'authoritative_branches': [],
}, 503)
metadata = patches()
branch_names = {item['name'] for item in branches}
metadata_names = {
item.get('branch') for item in metadata
if isinstance(item, dict) and item.get('branch')
}
orphan = [
item for item in metadata
if item.get('branch') and item.get('branch') not in branch_names
]
missing = [
item for item in branches
if item['name'] not in metadata_names
]
return self.sendj({
'metadata_patches': metadata,
'authoritative_branches': branches,
'orphan_metadata': orphan,
'branches_without_metadata': missing,
'repository_available': True,
'query_error': None,
'patches': metadata,
'branches': [item['name'] for item in branches],
})
frm=(q.get('from') or ['HEAD~1'])[0]; to=(q.get('to') or ['HEAD'])[0]; return self.sendj({'from':frm,'to':to,'diff':sh(['diff',frm,to])})
if path in ('/api/patch','/api/patches','/api/branches','/patch'): return self.sendj({'patches':patches(),'branches':sh(['branch','--format','%(refname:short)'])['stdout'].splitlines()})
if path in ('/api/model','/model'): return self.sendj({'base':BASE_MODELS,'custom':model_docs(),'priority':[jload(p,{}) for p in sorted((WS/'config'/'model_priority').glob('*.json'))]})
if path in ('/api/run','/api/runtime-status','/run'):
svc=subprocess.run(['systemctl','is-active','aios-gitops-model-api'], text=True, capture_output=True)
snapshot = repository_snapshot(
git_dir=AUTHORITATIVE_GIT_DIR,
repository_identity=REPOSITORY_IDENTITY,
)
workspace_branch = None
if (WS / '.git').exists():
result = sh(['rev-parse','--abbrev-ref','HEAD'])
workspace_branch = result['stdout'] or None
payload = {
'service':'aios-gitops-model-api',
'status':svc.stdout.strip() or svc.stderr.strip(),
'workspace_branch':workspace_branch,
**runtime_release_info(),
**snapshot,
'writes_enabled': str(os.getenv('AIOS_GITOPS_WRITES_ENABLED','false')).lower() in {'1','true','yes','on'},
}
return self.sendj(payload, 200 if snapshot['repository_available'] else 503)
return self.sendj({'service':'aios-gitops-model-api','status':svc.stdout.strip() or svc.stderr.strip(),'workspace_branch':sh(['rev-parse','--abbrev-ref','HEAD'])['stdout']})
if path in ('/api/action-monitor','/action-monitor'):
latest=monitor.latest()
ldict={'event_id':latest.event_id,'action':latest.action,'error':latest.error,'context':latest.context,'created_at':latest.created_at} if latest else None
return self.sendj({'events':[{"event_id":e.event_id,"action":e.action,"error":e.error,"context":e.context,"created_at":e.created_at} for e in monitor.list_events()],'latest':ldict})
if path == '/api/tag-publication-status':
version=(q.get('version') or [''])[0]
try:
version=validate_version(version)
status=publication_publisher().get_tag_publication_status(version)
except ValueError:
return self.sendj({'error':'invalid release version'},400)
return self.sendj(status)
if path == '/api/release-consistency':
request=publication_request_from_query(q)
return self.sendj(publication_publisher().get_release_consistency(request))
if path == '/api/release-acceptance':
request=publication_request_from_query(q)
return self.sendj(publication_publisher().run_release_acceptance(request))
if path in (
'/getReleaseCandidate',
'/getReleasePlan',
'/getReleaseAcceptance',
'/getApprovalEnvelope',
'/getReleaseSaga',
'/getCapabilityStatus',
):
candidate_id=(q.get('candidate_id') or [''])[0]
if not candidate_id:
return self.sendj({'error':'candidate_id is required'},400)
service=silent_release_service()
if path == '/getReleaseCandidate':
return self.sendj(
service.get_release_candidate(candidate_id)
)
if path == '/getApprovalEnvelope':
return self.sendj(
service.get_approval_envelope(candidate_id)
)
if path == '/getReleasePlan':
return self.sendj(service.get_release_plan(candidate_id))
if path == '/getReleaseAcceptance':
return self.sendj(service.get_acceptance(candidate_id))
if path == '/getCapabilityStatus':
return self.sendj(
service.get_capability_status(candidate_id)
)
return self.sendj(service.get_release_saga(candidate_id))
if path == '/getRecoveryStatus':
return self.sendj(
silent_release_service().get_recovery_status()
)
# Admin AI Evolution Dashboard
if path in ('/admin/ai-evolution','/admin/ai-evolution/'):
return self.ai_evolution_page()
if path in ('/admin/ai-evolution/api/data'):
return self.ai_evolution_data()
return self.sendj({'error':'not_found','path':path},404)
except FileNotFoundError:
return self.sendj({'error':'release candidate not found'},404)
except ValueError as e:
error=redact(e)
monitor.record_failure('do_GET', error, {'path':path})
return self.sendj({'error':error},400)
except Exception as e:
error=redact(e)
monitor.record_failure('do_GET', error, {'path':path})
return self.sendj({'error':error},500)
monitor.record_failure('do_GET', str(e), {'path':path})
return self.sendj({'error':str(e)},500)
def do_PUT(self):
path=urllib.parse.urlparse(self.path).path.rstrip('/')
if reject_non_read_http(self, 'PUT', path): return
data=self.body()
path=urllib.parse.urlparse(self.path).path.rstrip('/'); data=self.body()
if path == '/api/model-priority': return self.model_priority(data)
return self.sendj({'error':'not_found','path':path},404)
def do_POST(self):
path=urllib.parse.urlparse(self.path).path.rstrip('/')
if reject_non_read_http(self, 'POST', path): return
data=self.body()
path=urllib.parse.urlparse(self.path).path.rstrip('/'); data=self.body()
try:
if path == '/createReleaseCandidate':
a=self.need('release:candidate')
if not a: return
return self.sendj(
silent_release_service().create_release_candidate(data)
)
if path == '/createReleaseIntent':
a=self.need('release:candidate')
if not a: return
return self.sendj(
silent_release_service().create_release_intent(data)
)
if path in ('/submitOwnerApproval','/submitSignedApproval'):
a=self.need('release:approve')
if not a: return
return self.sendj(
silent_release_service().submit_owner_approval(data)
)
if path == '/startDeployment':
a=self.need('release:deploy')
if not a: return
return self.sendj(
silent_release_service().start_deployment(
str(data.get('candidate_id') or '')
)
)
if path == '/v1/orchestrator/decisions': return self.orchestration_decision(data)
if path in ('/api/branch','/api/create-branch'): return self.branch(data)
if path in ('/api/file','/api/commit-file'): return self.file(data)
if path == '/api/commit': return self.commit(data)
@@ -634,30 +240,28 @@ class H(BaseHTTPRequestHandler):
if path == '/api/model-priority': return self.model_priority(data)
if path == '/api/nl': return self.nl(data)
return self.sendj({'error':'not_found','path':path},404)
except (ValueError, FileNotFoundError) as e:
error=redact(e)
monitor.record_failure('do_POST', error, {'path':path})
return self.sendj({'error':error},400)
except CapabilityAuthorizationError as e:
error=redact(e)
monitor.record_failure('do_POST', error, {'path':path})
return self.sendj({'error':error},403)
except SagaConflict as e:
error=redact(e)
monitor.record_failure('do_POST', error, {'path':path})
return self.sendj({'error':error},409)
except Exception as e:
error=redact(e)
monitor.record_failure('do_POST', error, {'path':path})
return self.sendj({'error':error},500)
def do_PATCH(self):
path=urllib.parse.urlparse(self.path).path.rstrip('/')
if reject_non_read_http(self, 'PATCH', path): return
return self.sendj({'error':'not_found','path':path},404)
def do_DELETE(self):
path=urllib.parse.urlparse(self.path).path.rstrip('/')
if reject_non_read_http(self, 'DELETE', path): return
return self.sendj({'error':'not_found','path':path},404)
monitor.record_failure('do_POST', str(e), {'path':path})
return self.sendj({'error':str(e)},500)
def orchestration_decision(self,data):
a=self.need('orchestrator:decision')
if not a: return
if not ORCHESTRATOR_CONTRACT_COMMIT or ORCHESTRATOR_SECRET_REF_RESOLVER is None:
return self.sendj({'error':'CALLBACK_CONFIGURATION_UNAVAILABLE'},503)
try:
result=submit_orchestration_decision(
data,
contract_commit=ORCHESTRATOR_CONTRACT_COMMIT,
secret_ref=ORCHESTRATOR_CALLBACK_SECRET_REF,
resolver=ORCHESTRATOR_SECRET_REF_RESOLVER,
)
except CallbackBridgeError as exc:
return self.sendj({'error':exc.category},exc.status_code)
except Exception:
# The outer request handler records exception strings. Keep this
# secret-bearing boundary metadata-only even for unexpected faults.
return self.sendj({'error':'CALLBACK_INTERNAL_ERROR'},500)
return self.sendj(result,202)
def branch(self,data):
a=self.need('branch:create');
if not a: return
@@ -681,41 +285,12 @@ class H(BaseHTTPRequestHandler):
if not a: return
ver=data.get('version') or data.get('tag')
if not re.match(r'^V\d+R\d+C\d+(-[A-Za-z0-9_.-]+)?$', ver or ''): return self.sendj({'error':'version must match VxRxCx'},400)
if not data.get('target_oid') or not data.get('evidence_closure_oid') or not data.get('approval_plan_id'):
return self.sendj({
'error':'production_tag_evidence_required',
'required':['version','target_oid','evidence_closure_oid','approval_plan_id'],
},400)
try:
target_oid=resolve_commit_oid(data.get('target_oid'))
evidence_oid=resolve_commit_oid(data.get('evidence_closure_oid'), 'evidence_closure_oid')
except ValueError as exc:
return self.sendj({'error':'invalid_tag_target','detail':str(exc)},400)
if data.get('confirm') != True:
self.sendj({'status':'confirm_required','operation':'tag','hint':'This operation creates a Git tag. Confirm to proceed.','x-openai-isConsequential':True}, 428)
return
if a['role'] != 'owner' or not data.get('execute'):
b=checkout(data.get('branch') or f'request/tag-{safe(ver)}', 'main', True); req={'version':ver,'target_oid':target_oid,'evidence_closure_oid':evidence_oid,'approval_plan_id':data.get('approval_plan_id'),'requested_by':a['account'],'approval_required':True,'created_at':now()}; r=write_json(f'baseline/tag_requests/{safe(ver)}.json', req); c=commit([r], f'chore(chatgpt): request tag {ver}'); return self.sendj({'status':'approval_required','branch':b,'request':req,'commit':c})
request=PublicationRequest(ver,target_oid,evidence_oid,str(data.get('approval_plan_id')))
main_before=sh(['rev-parse','refs/heads/main'], True)['stdout']
publisher=publication_publisher()
try:
publication=publisher.publish(request)
except PublicationManualOwnerReviewRequired as exc:
audit(a['account'],'tag_manual_owner_review_required',ver,{'target_oid':target_oid,'approval_plan_id':data.get('approval_plan_id'),'error':redact(exc)})
return self.sendj({'status':'MANUAL_OWNER_REVIEW_REQUIRED','error':redact(exc),'main_changed':False,'force_push':False},409)
except PublicationConflict as exc:
audit(a['account'],'tag_conflict',ver,{'target_oid':target_oid,'approval_plan_id':data.get('approval_plan_id'),'error':str(exc)})
return self.sendj({'status':'RECONCILIATION_REQUIRED','error':str(exc),'main_changed':False,'force_push':False},409)
except PublicationPending as exc:
audit(a['account'],'tag_pending',ver,{'target_oid':target_oid,'approval_plan_id':data.get('approval_plan_id'),'error':str(exc)})
return self.sendj({'status':'REMOTE_PUSH_PENDING','error':str(exc),'main_changed':False,'force_push':False},503)
except PublicationReconciliationRequired as exc:
audit(a['account'],'tag_reconciliation_required',ver,{'target_oid':target_oid,'approval_plan_id':data.get('approval_plan_id'),'error':str(exc)})
return self.sendj({'status':'RECONCILIATION_REQUIRED','error':str(exc),'main_changed':False,'force_push':False},502)
main_after=sh(['rev-parse','refs/heads/main'], True)['stdout']
if main_after != main_before:
raise RuntimeError('main changed during tag operation')
result={'status':publication['state'],'version':ver,'tag_name':ver,'tag_oid':publication['tag_oid'],'target_oid':target_oid,'expected_target_oid':target_oid,'evidence_closure_oid':evidence_oid,'approval_plan_id':data.get('approval_plan_id'),'workspace_head':sh(['rev-parse','HEAD'],True)['stdout'],'main_oid_before':main_before,'main_oid_after':main_after,'main_changed':False,'force_push':False,'publication':publication}
audit(a['account'],'tag',ver,result)
self.sendj(result)
b=checkout(data.get('branch') or f'request/tag-{safe(ver)}', 'main', True); req={'version':ver,'requested_by':a['account'],'approval_required':True,'created_at':now()}; r=write_json(f'baseline/tag_requests/{safe(ver)}.json', req); c=commit([r], f'chore(chatgpt): request tag {ver}'); return self.sendj({'status':'approval_required','branch':b,'commit':c})
out=sh(['tag','-a',ver,'-m',data.get('message') or f'release {ver}'], True); self.sendj({'status':'tagged','version':ver,'git':out,'sync':sync_tag(ver)})
def rollback(self,data):
a=self.need('rollback:request');
if not a: return
@@ -810,10 +385,7 @@ return{vs,cmds,ar,cmd,cdet,tot,sok,sfail,runs,baseline,refreshMsg,loadCmd,selVer
cmd_filter=(params.get('command') or [None])[0]
bl=jload(WS/'baseline'/'current.json',{})
current_ver=bl.get('current_baseline','V1R3C1')
try:
tags=[item['name'] for item in authoritative_tags(git_dir=required_authoritative_git_dir()) if str(item['name']).startswith('V')]
except (AuthoritativeGitError, RuntimeError) as exc:
return self.sendj({'error':redact(exc),'repository_available':False},503)
tags=sh(['tag','--list','V*'])['stdout'].splitlines() if WS.exists() else []
versions=[]
if (WS/'baseline'/'promotions').exists():
for p in sorted((WS/'baseline'/'promotions').glob('*.json')):
@@ -851,7 +423,7 @@ return{vs,cmds,ar,cmd,cdet,tot,sok,sfail,runs,baseline,refreshMsg,loadCmd,selVer
def api_spec():
return {'version':'V1R1C0-chatgpt-actions','openapi':'/api/openapi.yaml','read':['GET /api/base','GET /api/ver','GET /api/diff','GET /api/patch','GET /api/model','GET /api/run','GET /getReleaseCandidate','GET /getApprovalEnvelope','GET /getReleaseSaga','GET /admin/ai-evolution'],'write':['POST /api/branch','POST /api/file','POST /api/commit','POST /api/patch','POST /api/tag','POST /api/rollback','POST /api/model','PUT /api/model-priority','POST /api/nl','POST /createReleaseCandidate','POST /submitOwnerApproval','POST /startDeployment'],'policy':{'bearer_required_for_write':True,'direct_main_push':False,'delete_tag':False,'owner_approval_required_for_v_level':True,'api_keys_in_git':False,'git_secret_field':'secret_ref','online_git_sync':True,'silent_release_execution_enabled':SILENT_RELEASE_EXECUTION_ENABLED,'silent_release_shadow_default':True}}
return {'version':'V1R1C0-chatgpt-actions','openapi':'/api/openapi.yaml','orchestrator_openapi':'/api/orchestrator-decision.openapi.yaml','read':['GET /api/base','GET /api/ver','GET /api/diff','GET /api/patch','GET /api/model','GET /api/run','GET /admin/ai-evolution'],'write':['POST /v1/orchestrator/decisions','POST /api/branch','POST /api/file','POST /api/commit','POST /api/patch','POST /api/tag','POST /api/rollback','POST /api/model','PUT /api/model-priority','POST /api/nl'],'policy':{'bearer_required_for_write':True,'orchestrator_decision_scope':'orchestrator:decision','orchestrator_callback_credential':'secret_ref_only','direct_main_push':False,'delete_tag':False,'owner_approval_required_for_v_level':True,'api_keys_in_git':False,'git_secret_field':'secret_ref','online_git_sync':True}}
def main():
print(f'AIOS ChatGPT GitOps adapter listening on {HOST}:{PORT}')
@@ -1,52 +0,0 @@
#!/usr/bin/env python3
"""Fail-closed policy for in-process Git mutation handlers."""
from __future__ import annotations
import os
from dataclasses import dataclass
from pathlib import Path
from typing import Mapping
_TRUE = frozenset({"1", "true", "yes", "on"})
@dataclass(frozen=True)
class GitWriteDecision:
allowed: bool
status_code: int
error: str
reason: str
workspace: str
def payload(self) -> dict[str, str]:
return {
"status": "allowed" if self.allowed else "blocked",
"error": self.error,
"reason": self.reason,
"workspace": self.workspace,
}
def _real(path: Path) -> Path:
return Path(os.path.realpath(os.path.abspath(str(path))))
def evaluate_git_write(workspace: Path | str, env: Mapping[str, str] | None = None) -> GitWriteDecision:
values = os.environ if env is None else env
lexical = Path(os.path.abspath(str(workspace)))
canonical = _real(lexical)
shown = str(lexical)
if str(values.get("AIOS_GITOPS_WRITES_ENABLED", "false")).strip().lower() not in _TRUE:
return GitWriteDecision(False, 423, "gitops_writes_disabled", "AIOS_GITOPS_WRITES_ENABLED is false", shown)
if lexical != canonical:
return GitWriteDecision(False, 409, "workspace_symlink_forbidden", "workspace must be a canonical non-symlink path", shown)
controller = _real(Path(values.get("AIOS_GITOPS_CONTROLLER_PATH", "/data/aios-gitops/worktrees/controller")))
role = str(values.get("AIOS_GITOPS_WORKSPACE_ROLE", "")).strip().lower()
if canonical == controller or role == "runtime-anchor":
return GitWriteDecision(False, 423, "runtime_anchor_read_only", "controller/runtime-anchor is read-only", str(canonical))
writer_root = _real(Path(values.get("AIOS_GITOPS_APPROVED_WRITER_ROOT", "/data/aios-gitops/worktrees")))
try:
canonical.relative_to(writer_root)
except ValueError:
return GitWriteDecision(False, 409, "workspace_not_approved_writer", "workspace is outside the approved writer root", str(canonical))
return GitWriteDecision(True, 200, "", "approved isolated writer", str(canonical))
@@ -1,316 +0,0 @@
"""Task-scoped hardened wrapper around the existing Action API."""
from __future__ import annotations
import json
import os
import subprocess
from pathlib import Path
import urllib.parse
import aios_chatgpt_actions_api as api
import aios_isolated_writer_policy as policy
from isolated_writer_authz import AuthorizationError, authorize
from isolated_writer_commit_file import AtomicCommitFile, CommitFileError, OperationLedger
from isolated_writer_token_registry import TokenRegistry
_reject = api.reject_non_read_http
_write_file = api.write_file
_sync_branch = api.sync_branch
_original_do_get = api.H.do_GET
api.OPENAPI = Path(os.getenv("AIOS_GITOPS_ISOLATED_WRITER_OPENAPI", str(api.OPENAPI)))
def _registry() -> TokenRegistry:
return TokenRegistry(Path(os.environ["AIOS_GITOPS_TOKEN_REGISTRY_V2"]))
def _atomic_writer() -> AtomicCommitFile:
return AtomicCommitFile(
workspace=api.WS,
bare_repo=api.BARE_REPO,
ledger=OperationLedger(Path(os.environ["AIOS_GITOPS_WRITER_OPERATION_LEDGER"])),
remote=os.getenv("AIOS_GITOPS_GITEA_PUSH_REMOTE", ""),
audit=policy.append_audit_record,
)
def _bearer(handler) -> str:
value = handler.headers.get("Authorization", "")
return value[7:].strip() if value.startswith("Bearer ") else ""
def _authorize(handler, *, method: str, path: str, branch=None, target_path=None):
grant = authorize(
_registry(),
_bearer(handler),
method=method,
path=path,
source_ip=str(handler.client_address[0]),
workspace=api.WS,
branch=branch,
target_path=target_path,
)
grant.record["_source_ip"] = str(handler.client_address[0])
return grant
def _authorization_error(handler, exc: AuthorizationError):
return handler.sendj({"error": exc.code}, exc.status)
def _isolated_runtime_status(handler):
branch = api.sh(["rev-parse", "--abbrev-ref", "HEAD"], True)["stdout"]
head = api.sh(["rev-parse", "HEAD"], True)["stdout"]
return handler.sendj({
"service": os.getenv(
"AIOS_GITOPS_SERVICE_AUDIENCE",
"aios-isolated-writer.service",
),
"status": "active",
"runtime_role": os.getenv("AIOS_GITOPS_RUNTIME_ROLE"),
"workspace_role": os.getenv("AIOS_GITOPS_WORKSPACE_ROLE"),
"writes_enabled": os.getenv("AIOS_GITOPS_WRITES_ENABLED", "").lower()
in {"1", "true", "yes", "on"},
"allowed_branch": policy.allowed_branch(),
"approved_base_oid": policy.allowed_base_oid(),
"workspace_branch": branch,
"workspace_head_oid": head,
"repository_identity": os.getenv("AIOS_GITOPS_REPOSITORY_IDENTITY"),
"token_policy_version": 2,
"atomic_commit_file": True,
"path_policy_sha256": policy.configured_path_policy_sha256(),
})
def _isolated_do_get(self):
path = urllib.parse.urlparse(self.path).path.rstrip("/") or "/"
if path in {"/health", "/api/health"}:
return self.sendj({
"status": "ok",
"service": os.getenv(
"AIOS_GITOPS_SERVICE_AUDIENCE",
"aios-isolated-writer.service",
),
})
if path not in {"/api/runtime-status", "/api/baseline", "/api/patches"}:
return self.sendj({"error": "isolated_writer_operation_forbidden"}, 403)
try:
_authorize(self, method="GET", path=path)
except AuthorizationError as exc:
return _authorization_error(self, exc)
if path == "/api/runtime-status":
return _isolated_runtime_status(self)
return _original_do_get(self)
def _read_commit_file_body(handler):
try:
length = int(handler.headers.get("Content-Length", "0"))
except ValueError:
raise CommitFileError("writer_content_length_rejected") from None
if length <= 0:
raise CommitFileError("writer_request_object_required")
if length > 1_100_000:
raise CommitFileError("writer_request_too_large", 413)
raw = handler.rfile.read(length)
try:
value = json.loads(raw.decode("utf-8"))
except (UnicodeDecodeError, json.JSONDecodeError):
raise CommitFileError("writer_request_json_rejected") from None
if not isinstance(value, dict):
raise CommitFileError("writer_request_object_required")
return value
def _isolated_do_post(self):
path = urllib.parse.urlparse(self.path).path.rstrip("/") or "/"
if path != "/api/commit-file":
return self.sendj({"error": "isolated_writer_operation_forbidden"}, 403)
try:
data = _read_commit_file_body(self)
branch = str(data.get("branch") or "")
target_path = str(data.get("path") or "")
grant = _authorize(
self,
method="POST",
path=path,
branch=branch,
target_path=target_path,
)
if api.content_has_secret(str(data.get("content") or "")):
raise CommitFileError("writer_secret_content_rejected", 403)
result = _atomic_writer().execute(grant, data)
return self.sendj(result)
except AuthorizationError as exc:
return _authorization_error(self, exc)
except CommitFileError as exc:
return self.sendj({"error": exc.code}, exc.status)
except PermissionError:
return self.sendj({"error": "isolated_writer_policy_rejected"}, 403)
def reject_non_read_http(handler, method: str, path: str) -> bool:
if _reject(handler, method, path):
return True
try:
policy.validate_endpoint(method, path, api.WS, api.BARE_REPO)
return False
except PermissionError as exc:
handler.sendj({"error": str(exc)}, 403)
return True
def checkout(branch, base="main", create=False):
api.ensure_git()
target = api.branch_name(branch)
policy.validate_branch_request({"branch": target, "base": base})
exists = api.sh(["rev-parse", "--verify", target])["code"] == 0
if exists:
if api.sh(["rev-parse", "--abbrev-ref", "HEAD"], True)["stdout"] != target:
api.sh(["checkout", target], True)
elif create:
api.sh(["checkout", "-b", target, base], True)
else:
raise PermissionError("isolated_writer_branch_missing")
policy.require_service_invariants(api.WS, api.BARE_REPO)
return target
def write_file(path, content):
policy.validate_path(path, api.WS)
return _write_file(path, content)
def sync_branch(branch: str):
policy.validate_sync_branch(branch)
local = _sync_branch(branch)
remote = os.getenv("AIOS_GITOPS_GITEA_PUSH_REMOTE", "").strip()
if remote != "gitea-http":
raise PermissionError("isolated_writer_gitea_remote_rejected")
head = api.sh(["rev-parse", "HEAD"], True)["stdout"]
pushed = subprocess.run(
["git", "push", "--porcelain", remote, f"HEAD:refs/heads/{branch}"],
cwd=api.WS,
text=True,
capture_output=True,
)
if pushed.returncode != 0:
raise RuntimeError("isolated_writer_gitea_push_failed")
verified = subprocess.run(
["git", "ls-remote", "--heads", remote, f"refs/heads/{branch}"],
cwd=api.WS,
text=True,
capture_output=True,
)
if verified.returncode != 0:
raise RuntimeError("isolated_writer_gitea_verification_failed")
fields = verified.stdout.strip().split()
if len(fields) != 2 or fields[0] != head or fields[1] != f"refs/heads/{branch}":
raise RuntimeError("isolated_writer_gitea_oid_mismatch")
return {
"status": "verified",
"target": "authoritative-bare+gitea",
"branch": branch,
"oid": head,
"local": local.get("status"),
}
def commit(paths, msg):
policy.validate_commit_paths(paths, api.WS)
api.ensure_git()
selected = [api.relpath(p) for p in paths] if paths else ["."]
api.sh(["add", *selected], True)
policy.validate_staged_paths(api.WS)
branch = api.sh(["rev-parse", "--abbrev-ref", "HEAD"], True)["stdout"]
if not api.sh(["status", "--short"])["stdout"]:
return {"status": "no_changes", "branch": branch, "sync": sync_branch(branch)}
out = api.sh(["commit", "-m", msg], True)
branch = api.sh(["rev-parse", "--abbrev-ref", "HEAD"], True)["stdout"]
return {
"status": "committed",
"branch": branch,
"rev": api.sh(["rev-parse", "--short", "HEAD"], True)["stdout"],
"git": out,
"sync": sync_branch(branch),
}
def audit(actor, action, target, payload):
policy.append_audit_record(
{
"at": api.now(),
"actor": actor,
"action": action,
"target": target,
"payload": api.sanitize(payload),
}
)
def _validate_request(self, data, validate):
try:
validate(data)
return True
except PermissionError as exc:
self.sendj({"error": str(exc)}, 403)
return False
def _checked_branch(self, data):
if not _validate_request(self, data, policy.validate_branch_request):
return
return _branch(self, data)
def _checked_file(self, data):
data = dict(data)
data.setdefault("branch", policy.allowed_branch())
if not _validate_request(self, data, policy.validate_branch_request):
return
if not _validate_request(
self,
{"path": data.get("path", "")},
lambda item: policy.validate_path(item["path"], api.WS),
):
return
return _file(self, data)
def _checked_commit(self, data):
data = dict(data)
data.setdefault("branch", policy.allowed_branch())
if not _validate_request(self, data, policy.validate_branch_request):
return
return _commit(self, data)
def _checked_patch(self, data):
data = dict(data)
data["branch"] = policy.allowed_branch()
if not _validate_request(self, data, policy.validate_branch_request):
return
return _patch(self, data)
_branch, _file, _commit, _patch = api.H.branch, api.H.file, api.H.commit, api.H.patch
api.reject_non_read_http = reject_non_read_http
api.checkout, api.write_file, api.sync_branch, api.commit, api.audit = (
checkout,
write_file,
sync_branch,
commit,
audit,
)
api.H.branch, api.H.file, api.H.commit, api.H.patch = (
_checked_branch,
_checked_file,
_checked_commit,
_checked_patch,
)
api.H.do_GET = _isolated_do_get
api.H.do_POST = _isolated_do_post
if __name__ == "__main__":
policy.require_service_invariants(api.WS, api.BARE_REPO)
api.ThreadingHTTPServer((api.HOST, api.PORT), api.H).serve_forever()
@@ -1,310 +0,0 @@
"""Fail-closed policy used only by the task-scoped Action writer."""
from __future__ import annotations
import os
import json
import stat
import subprocess
import re
from pathlib import Path, PurePosixPath, PureWindowsPath
from urllib.parse import unquote
import isolated_writer_path_policy as path_policy
WRITE_ENDPOINTS = frozenset({"/api/commit-file"})
ALLOWED_ROOTS = (
"active-runtime",
"docs",
"patches",
"reports",
"scripts",
"test",
"tests",
)
DENIED_ROOTS = ("baseline", "config", "deploy", "deployments", "ops", "runtime", "services")
_TRUE = frozenset({"1", "true", "yes", "on"})
_FALSE = frozenset({"0", "false", "no", "off"})
_DENY_FLAGS = (
"AIOS_GITOPS_ALLOW_MAIN_WRITE",
"AIOS_GITOPS_ALLOW_TAG_WRITE",
"AIOS_GITOPS_ALLOW_BASELINE_WRITE",
"AIOS_GITOPS_ALLOW_REGISTRY_WRITE",
"AIOS_GITOPS_ALLOW_ROLLBACK",
"AIOS_GITOPS_ALLOW_DEPLOYMENT",
)
def enforcement_required() -> bool:
return os.getenv("AIOS_GITOPS_ISOLATED_WRITER_ENFORCEMENT", "") == "required"
def allowed_branch() -> str:
value = os.getenv("AIOS_GITOPS_ALLOWED_BRANCH", "").strip()
if not value:
raise PermissionError("isolated_writer_missing_allowed_branch")
return value
def allowed_base_oid() -> str:
value = os.getenv("AIOS_GITOPS_ALLOWED_BASE_OID", "").strip().lower()
if len(value) != 40 or any(ch not in "0123456789abcdef" for ch in value):
raise PermissionError("isolated_writer_invalid_allowed_base_oid")
return value
def _git(workspace: Path, *args: str) -> str:
result = subprocess.run(["git", *args], cwd=workspace, text=True, capture_output=True)
if result.returncode != 0:
raise PermissionError("isolated_writer_git_validation_failed")
return result.stdout.strip()
def require_service_invariants(workspace: Path, bare_repo: Path) -> None:
if not enforcement_required():
return
if os.getenv("AIOS_GITOPS_WRITES_ENABLED", "").strip().lower() not in _TRUE:
raise PermissionError("isolated_writer_write_gate_required")
if os.getenv("AIOS_GITOPS_RUNTIME_ROLE", "") != "isolated-writer":
raise PermissionError("isolated_writer_role_required")
if os.getenv("AIOS_GITOPS_WORKSPACE_ROLE", "") != "isolated-writer":
raise PermissionError("isolated_writer_workspace_role_required")
if os.getenv("AIOS_GITOPS_TOKEN_POLICY_VERSION_REQUIRED", "") != "2":
raise PermissionError("isolated_writer_token_policy_v2_required")
if not os.getenv("AIOS_GITOPS_SERVICE_AUDIENCE", "").strip():
raise PermissionError("isolated_writer_service_audience_required")
if not os.getenv("AIOS_GITOPS_TOKEN_REGISTRY_V2", "").strip():
raise PermissionError("isolated_writer_token_registry_required")
if not os.getenv("AIOS_GITOPS_WRITER_OPERATION_LEDGER", "").strip():
raise PermissionError("isolated_writer_operation_ledger_required")
path_policy.configured_path_policy()
for name in _DENY_FLAGS:
if os.getenv(name, "").strip().lower() not in _FALSE:
raise PermissionError("isolated_writer_forbidden_capability_flag")
if os.getenv("AIOS_GITOPS_REPOSITORY_IDENTITY", "") != "aios/aios":
raise PermissionError("isolated_writer_repository_identity_rejected")
configured_bare_value = os.getenv("AIOS_GITOPS_AUTHORITATIVE_GIT_DIR", "").strip()
if not configured_bare_value:
raise PermissionError("isolated_writer_authoritative_repo_rejected")
configured_bare = Path(configured_bare_value)
if configured_bare.resolve(strict=True) != bare_repo.resolve(strict=True):
raise PermissionError("isolated_writer_authoritative_repo_rejected")
writer_root_value = os.getenv("AIOS_GITOPS_APPROVED_WRITER_ROOT", "").strip()
if not writer_root_value:
raise PermissionError("isolated_writer_workspace_outside_approved_root")
writer_root = Path(writer_root_value).resolve(strict=True)
lexical_workspace = Path(os.path.abspath(str(workspace)))
canonical_workspace = lexical_workspace.resolve(strict=True)
if lexical_workspace != canonical_workspace:
raise PermissionError("isolated_writer_workspace_symlink_rejected")
try:
canonical_workspace.relative_to(writer_root)
except (ValueError, FileNotFoundError):
raise PermissionError("isolated_writer_workspace_outside_approved_root") from None
if _git(workspace, "rev-parse", "--abbrev-ref", "HEAD") != allowed_branch():
raise PermissionError("isolated_writer_current_branch_rejected")
approved_base = allowed_base_oid()
try:
_git(workspace, "cat-file", "-e", f"{approved_base}^{{commit}}")
subprocess.run(
["git", "merge-base", "--is-ancestor", approved_base, "HEAD"],
cwd=workspace,
check=True,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
)
except (PermissionError, subprocess.CalledProcessError):
raise PermissionError("isolated_writer_approved_base_not_ancestor") from None
if _git(workspace, "merge-base", approved_base, "HEAD") != approved_base:
raise PermissionError("isolated_writer_merge_base_rejected")
common = Path(_git(workspace, "rev-parse", "--git-common-dir"))
if not common.is_absolute():
common = (workspace / common).resolve()
if common.resolve() != bare_repo.resolve():
raise PermissionError("isolated_writer_common_repo_rejected")
def approved_audit_log_path() -> Path:
root_value = os.getenv("AIOS_GITOPS_APPROVED_AUDIT_ROOT", "").strip()
log_value = os.getenv("AIOS_GITOPS_AUDIT_LOG", "").strip()
if not root_value or not log_value:
raise PermissionError("isolated_writer_audit_path_missing")
lexical_root = Path(os.path.abspath(root_value))
lexical_log = Path(os.path.abspath(log_value))
root = lexical_root.resolve(strict=True)
if lexical_root != root or root.is_symlink() or not root.is_dir():
raise PermissionError("isolated_writer_audit_root_rejected")
try:
lexical_log.relative_to(root)
except ValueError:
raise PermissionError("isolated_writer_audit_path_outside_approved_root") from None
if lexical_log.parent.resolve(strict=True) != lexical_log.parent:
raise PermissionError("isolated_writer_audit_parent_symlink_rejected")
if lexical_log.exists() and lexical_log.is_symlink():
raise PermissionError("isolated_writer_audit_log_symlink_rejected")
os.chmod(root, 0o750)
os.chmod(lexical_log.parent, 0o750)
return lexical_log
def append_audit_record(record: dict) -> None:
path = approved_audit_log_path()
flags = os.O_WRONLY | os.O_APPEND | os.O_CREAT
if hasattr(os, "O_NOFOLLOW"):
flags |= os.O_NOFOLLOW
descriptor = os.open(path, flags, 0o600)
try:
info = os.fstat(descriptor)
if not stat.S_ISREG(info.st_mode):
raise PermissionError("isolated_writer_audit_log_not_regular")
os.fchmod(descriptor, 0o600)
line = json.dumps(
_audit_sanitize(record),
ensure_ascii=False,
sort_keys=True,
separators=(",", ":"),
).encode("utf-8") + b"\n"
os.write(descriptor, line)
os.fsync(descriptor)
finally:
os.close(descriptor)
parent_fd = os.open(path.parent, os.O_RDONLY | getattr(os, "O_DIRECTORY", 0))
try:
os.fsync(parent_fd)
finally:
os.close(parent_fd)
def _audit_sanitize(value):
if isinstance(value, dict):
return {
key: (
"[REDACTED]"
if any(
marker in str(key).lower()
for marker in (
"authorization",
"credential",
"password",
"secret",
"token",
"private_key",
)
)
and str(key).lower() != "token_id"
else _audit_sanitize(inner)
)
for key, inner in value.items()
}
if isinstance(value, list):
return [_audit_sanitize(item) for item in value]
if isinstance(value, str) and re.search(
r"(?i)(authorization\s*:\s*bearer|(?:token|password|secret)\s*[:=])",
value,
):
return "[REDACTED]"
return value
def validate_endpoint(method: str, path: str, workspace: Path, bare_repo: Path) -> None:
if not enforcement_required():
return
if method != "POST" or path not in WRITE_ENDPOINTS:
raise PermissionError("isolated_writer_operation_forbidden")
require_service_invariants(workspace, bare_repo)
def validate_branch_request(data: dict) -> None:
if not enforcement_required():
return
branch = allowed_branch() if "branch" not in data else str(data.get("branch") or "").strip()
if branch != allowed_branch():
raise PermissionError("isolated_writer_branch_rejected")
if str(data.get("base", "main")).strip() != "main":
raise PermissionError("isolated_writer_base_rejected")
if data.get("force") or data.get("force_push"):
raise PermissionError("isolated_writer_force_rejected")
def validate_path(path: str, workspace: Path | None = None) -> str:
if not isinstance(path, str):
raise PermissionError("isolated_writer_path_rejected")
value = path
if not value or value != value.strip() or "\x00" in value or "\\" in value:
raise PermissionError("isolated_writer_path_rejected")
try:
decoded = unquote(value, errors="strict")
except (UnicodeDecodeError, ValueError):
raise PermissionError("isolated_writer_path_rejected") from None
if decoded != value:
raise PermissionError("isolated_writer_path_rejected")
windows = PureWindowsPath(value)
posix = PurePosixPath(value)
parts = value.split("/")
if (
posix.is_absolute()
or windows.is_absolute()
or bool(windows.drive)
or any(part in {"", ".", ".."} for part in parts)
):
raise PermissionError("isolated_writer_path_rejected")
first = parts[0]
service_exact_path = False
if first in DENIED_ROOTS:
try:
path_policy.require_service_path("commit_file.atomic", value)
service_exact_path = True
except path_policy.PathPolicyError as exc:
raise PermissionError(str(exc)) from None
if value != "README.md" and first not in ALLOWED_ROOTS and not service_exact_path:
raise PermissionError("isolated_writer_path_outside_allowed_roots")
if workspace is not None:
lexical_root = Path(os.path.abspath(str(workspace)))
root = lexical_root.resolve(strict=True)
if lexical_root != root:
raise PermissionError("isolated_writer_workspace_symlink_rejected")
current = root
for part in parts:
current = current / part
if current.is_symlink():
raise PermissionError("isolated_writer_path_symlink_rejected")
candidate = (root / value).resolve(strict=False)
try:
candidate.relative_to(root)
except ValueError:
raise PermissionError("isolated_writer_path_escapes_workspace") from None
return value
def validate_content_for_path(path: str, content: str) -> None:
if path.split("/", 1)[0] not in DENIED_ROOTS:
return
try:
path_policy.validate_content("commit_file.atomic", path, content)
except path_policy.PathPolicyError as exc:
raise PermissionError(str(exc)) from None
def configured_path_policy_sha256() -> str | None:
try:
return path_policy.configured_path_policy_sha256()
except path_policy.PathPolicyError as exc:
raise PermissionError(str(exc)) from None
def validate_commit_paths(paths: list[str], workspace: Path | None = None) -> None:
if enforcement_required():
for path in paths:
if str(path).strip() != ".":
validate_path(str(path), workspace)
def validate_staged_paths(workspace: Path) -> None:
if not enforcement_required():
return
for path in _git(workspace, "diff", "--cached", "--name-only").splitlines():
validate_path(path, workspace)
def validate_sync_branch(branch: str) -> None:
if enforcement_required() and branch != allowed_branch():
raise PermissionError("isolated_writer_sync_branch_rejected")
-355
View File
@@ -1,355 +0,0 @@
"""Fail-closed access to the authoritative bare Git repository."""
from __future__ import annotations
import os
import re
import subprocess
import urllib.parse
from pathlib import Path
from typing import Iterable
_ALLOWED_COMMANDS = frozenset(
{
"cat-file",
"config",
"diff",
"for-each-ref",
"ls-remote",
"merge-base",
"push",
"remote",
"rev-parse",
"show-ref",
"tag",
}
)
_OID_RE = re.compile(r"^[0-9a-f]{40}$")
_MAX_OUTPUT = 1_048_576
_TIMEOUT = 8.0
class AuthoritativeGitError(RuntimeError):
"""A fail-closed authoritative repository or ref query error."""
class AuthoritativeGitNotFound(AuthoritativeGitError):
"""A validated revision does not exist in the authoritative repository."""
class AuthoritativeGitOutputTooLarge(AuthoritativeGitError):
"""A controlled Git query exceeded the response-size boundary."""
def _redact(text: str) -> str:
value = str(text or "")
value = re.sub(r"(?i)(authorization\s*[:=]\s*)(?:bearer|token)\s+\S+", r"\1REDACTED", value)
value = re.sub(r"(?i)((?:access_)?token\s*[:=]\s*)\S+", r"\1REDACTED", value)
value = re.sub(r"(?i)(https?://[^/\s:@]+:)[^@\s/]+@", r"\1REDACTED@", value)
return value
def _configured_path(git_dir: str | Path | None = None) -> Path:
raw = str(git_dir or os.getenv("AIOS_GITOPS_AUTHORITATIVE_GIT_DIR", "")).strip()
if not raw:
raise AuthoritativeGitError("AUTHORITATIVE_REPOSITORY_UNAVAILABLE")
path = Path(raw)
if not path.is_absolute() or ".." in path.parts:
raise AuthoritativeGitError("AUTHORITATIVE_REPOSITORY_UNAVAILABLE")
try:
resolved = path.resolve(strict=True)
except OSError as exc:
raise AuthoritativeGitError("AUTHORITATIVE_REPOSITORY_UNAVAILABLE") from exc
if not resolved.is_dir():
raise AuthoritativeGitError("AUTHORITATIVE_REPOSITORY_UNAVAILABLE")
return resolved
def _run(
git_dir: Path,
args: list[str],
*,
check: bool,
timeout: float = _TIMEOUT,
) -> subprocess.CompletedProcess[str]:
if not args or args[0] not in _ALLOWED_COMMANDS:
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
try:
result = subprocess.run(
["git", "--git-dir", str(git_dir), *args],
text=True,
capture_output=True,
timeout=timeout,
env={**os.environ, "GIT_OPTIONAL_LOCKS": "0"},
)
except (OSError, subprocess.TimeoutExpired) as exc:
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED") from exc
if len(result.stdout) + len(result.stderr) > _MAX_OUTPUT:
raise AuthoritativeGitOutputTooLarge("GIT_OUTPUT_LIMIT_EXCEEDED")
if check and result.returncode != 0:
detail = _redact(result.stderr.strip() or result.stdout.strip())
raise AuthoritativeGitError(
"GIT_REF_QUERY_FAILED" + (f": {detail[:160]}" if detail else "")
)
return result
def _repository_identity_from_remote(url: str) -> str:
value = str(url or "").strip().split("?", 1)[0].rstrip("/")
if not value:
return ""
if "://" in value:
value = urllib.parse.urlparse(value).path
elif ":" in value:
value = value.rsplit(":", 1)[1]
value = value.rsplit("@", 1)[-1].strip("/")
value = re.sub(r"\.git$", "", value, flags=re.IGNORECASE)
parts = [part for part in value.split("/") if part]
if len(parts) >= 2:
value = "/".join(parts[-2:])
return value.strip("/").lower()
def verify_repository(
git_dir: str | Path | None = None,
repository_identity: str | None = None,
) -> tuple[Path, str]:
path = _configured_path(git_dir)
expected = str(
repository_identity
or os.getenv("AIOS_GITOPS_REPOSITORY_IDENTITY", "")
).strip().strip("/")
if not expected or expected.count("/") != 1:
raise AuthoritativeGitError("REPOSITORY_IDENTITY_MISMATCH")
bare = _run(path, ["rev-parse", "--is-bare-repository"], check=True)
if bare.stdout.strip().lower() != "true":
raise AuthoritativeGitError("AUTHORITATIVE_REPOSITORY_UNAVAILABLE")
remote = _run(path, ["config", "--get", "remote.gitea.url"], check=False)
actual = _repository_identity_from_remote(remote.stdout)
if actual != expected.lower():
raise AuthoritativeGitError("REPOSITORY_IDENTITY_MISMATCH")
return path, expected
def run_authoritative_git(
args: list[str],
*,
check: bool = True,
git_dir: str | Path | None = None,
timeout: float = _TIMEOUT,
) -> subprocess.CompletedProcess[str]:
path, _ = verify_repository(git_dir=git_dir)
return _run(path, list(args), check=check, timeout=timeout)
def authoritative_head_oid(
*,
git_dir: str | Path | None = None,
) -> str:
result = run_authoritative_git(["rev-parse", "--verify", "refs/heads/main"], git_dir=git_dir)
oid = result.stdout.strip().lower()
if not _OID_RE.fullmatch(oid):
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
return oid
def authoritative_tags(
*,
git_dir: str | Path | None = None,
) -> list[dict[str, str | None]]:
result = run_authoritative_git(
[
"for-each-ref",
"refs/tags",
"--format=%(refname:strip=2)\t%(objectname)\t%(objecttype)",
],
git_dir=git_dir,
)
tags: list[dict[str, str | None]] = []
for line in result.stdout.splitlines():
name, oid, object_type = (line.split("\t", 2) + ["", ""])[:3]
if not name or not _OID_RE.fullmatch(oid.lower()):
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
peeled_result = run_authoritative_git(
["rev-parse", "--verify", f"refs/tags/{name}^{{commit}}"],
git_dir=git_dir,
)
peeled = peeled_result.stdout.strip().lower()
if not _OID_RE.fullmatch(peeled):
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
tags.append(
{
"name": name,
"tag_oid": oid.lower(),
"object_type": object_type,
"peeled_target_oid": peeled,
}
)
return tags
def authoritative_branches(
*,
git_dir: str | Path | None = None,
) -> list[dict[str, str]]:
result = run_authoritative_git(
[
"for-each-ref",
"refs/heads",
"--format=%(refname:strip=2)\t%(objectname)",
],
git_dir=git_dir,
)
branches: list[dict[str, str]] = []
for line in result.stdout.splitlines():
name, oid = (line.split("\t", 1) + [""])[:2]
if not name or not _OID_RE.fullmatch(oid.lower()):
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
branches.append({"name": name, "oid": oid.lower()})
return branches
def authoritative_commit(
oid: str,
*,
git_dir: str | Path | None = None,
) -> str:
value = str(oid or "").strip().lower()
if not _OID_RE.fullmatch(value):
raise ValueError("commit OID must be a full 40-character value")
result = run_authoritative_git(
["rev-parse", "--verify", f"{value}^{{commit}}"],
git_dir=git_dir,
)
resolved = result.stdout.strip().lower()
if resolved != value:
raise AuthoritativeGitError("GIT_REF_QUERY_FAILED")
return resolved
def authoritative_lineage(
ancestor: str,
descendant: str,
*,
git_dir: str | Path | None = None,
) -> None:
authoritative_commit(ancestor, git_dir=git_dir)
authoritative_commit(descendant, git_dir=git_dir)
result = run_authoritative_git(
["merge-base", "--is-ancestor", ancestor, descendant],
check=False,
git_dir=git_dir,
)
if result.returncode != 0:
raise AuthoritativeGitError("evidence lineage mismatch")
_SAFE_REF_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._/-]{0,239}$")
def _safe_ref_name(value: str) -> bool:
return bool(
_SAFE_REF_RE.fullmatch(value)
and ".." not in value
and "//" not in value
and "@{" not in value
and not value.endswith(("/", ".", ".lock"))
and not any(part in {"", ".", ".."} for part in value.split("/"))
)
def validate_git_revision(
value: str,
field: str,
*,
git_dir: str | Path | None = None,
) -> str:
revision = str(value or "").strip()
if not revision or revision != str(value or "") or revision.startswith("-"):
raise ValueError(f"invalid {field} revision")
if _OID_RE.fullmatch(revision.lower()):
try:
return authoritative_commit(revision.lower(), git_dir=git_dir)
except AuthoritativeGitError as exc:
raise AuthoritativeGitNotFound(f"{field} revision not found") from exc
if not _safe_ref_name(revision) or any(token in revision for token in ("^", "~", ":", "\\")):
raise ValueError(f"invalid {field} revision")
if revision.startswith("refs/"):
if not revision.startswith(("refs/heads/", "refs/tags/")):
raise ValueError(f"invalid {field} revision")
candidates = [revision]
else:
candidates = [f"refs/heads/{revision}", f"refs/tags/{revision}"]
matches: list[str] = []
for candidate in candidates:
result = run_authoritative_git(
["rev-parse", "--verify", f"{candidate}^{{commit}}"],
check=False,
git_dir=git_dir,
)
oid = result.stdout.strip().lower()
if result.returncode == 0 and _OID_RE.fullmatch(oid):
matches.append(oid)
if not matches:
raise AuthoritativeGitNotFound(f"{field} revision not found")
if len(set(matches)) != 1 or len(matches) != 1:
raise ValueError(f"ambiguous {field} revision")
return matches[0]
def authoritative_diff(
from_revision: str,
to_revision: str,
*,
git_dir: str | Path | None = None,
) -> dict[str, object]:
path, _ = verify_repository(git_dir=git_dir)
from_oid = validate_git_revision(from_revision, "from", git_dir=path)
to_oid = validate_git_revision(to_revision, "to", git_dir=path)
args = ["diff", "--no-ext-diff", "--no-textconv", from_oid, to_oid]
result = run_authoritative_git(args, git_dir=path)
return {
"code": 0,
"stdout": result.stdout,
"stderr": "",
"cmd": ["git", f"--git-dir={path}", *args],
"from_oid": from_oid,
"to_oid": to_oid,
}
def repository_snapshot(
*,
git_dir: str | Path | None = None,
repository_identity: str | None = None,
) -> dict[str, object]:
configured = bool(
str(git_dir or os.getenv("AIOS_GITOPS_AUTHORITATIVE_GIT_DIR", "")).strip()
)
try:
path, identity = verify_repository(
git_dir=git_dir,
repository_identity=repository_identity,
)
head = authoritative_head_oid(git_dir=path)
return {
"repository_available": True,
"authoritative_git_dir_configured": configured,
"authoritative_git_dir_readable": True,
"authoritative_git_dir": str(path),
"authoritative_head_oid": head,
"repository_identity": identity,
"query_error": None,
}
except (AuthoritativeGitError, ValueError) as exc:
return {
"repository_available": False,
"authoritative_git_dir_configured": configured,
"authoritative_git_dir_readable": False,
"authoritative_git_dir": None,
"authoritative_head_oid": None,
"repository_identity": str(
repository_identity
or os.getenv("AIOS_GITOPS_REPOSITORY_IDENTITY", "")
).strip() or None,
"query_error": _redact(str(exc)),
}
File diff suppressed because it is too large Load Diff
-167
View File
@@ -1,167 +0,0 @@
"""Fail-closed operation authorization for an isolated Git writer."""
from __future__ import annotations
import ipaddress
import os
import re
import subprocess
from dataclasses import dataclass
from datetime import datetime, timezone
from pathlib import Path
from typing import Any
from isolated_writer_token_registry import TokenRegistry
ROUTE_OPERATIONS = {
("GET", "/api/runtime-status"): "runtime.status.read",
("GET", "/api/baseline"): "baseline.read",
("GET", "/api/patches"): "patches.read",
("POST", "/api/commit-file"): "commit_file.atomic",
}
OID_RE = re.compile(r"[0-9a-f]{40}")
class AuthorizationError(PermissionError):
def __init__(self, code: str, status: int = 403):
super().__init__(code)
self.code = code
self.status = status
@dataclass(frozen=True)
class AuthorizedGrant:
record: dict[str, Any]
operation: str
use_count: int
@property
def token_id(self) -> str:
return str(self.record["token_id"])
@property
def task_id(self) -> str:
return str(self.record["task_id"])
def operation_for(method: str, path: str) -> str:
operation = ROUTE_OPERATIONS.get((method.upper(), path))
if not operation:
raise AuthorizationError("isolated_writer_operation_forbidden")
return operation
def _parse_time(value: Any) -> datetime:
try:
parsed = datetime.fromisoformat(str(value).replace("Z", "+00:00"))
except (TypeError, ValueError):
raise AuthorizationError("writer_token_time_invalid") from None
if parsed.tzinfo is None:
raise AuthorizationError("writer_token_time_invalid")
return parsed.astimezone(timezone.utc)
def _git_is_ancestor(workspace: Path, base: str, branch: str) -> bool:
result = subprocess.run(
["git", "merge-base", "--is-ancestor", base, branch],
cwd=workspace,
stdout=subprocess.DEVNULL,
stderr=subprocess.DEVNULL,
)
return result.returncode == 0
def _validate_grant(
grant: dict[str, Any],
*,
audience: str,
source_ip: str,
operation: str,
branch: str | None,
path: str | None,
workspace: Path,
current_time: datetime,
) -> None:
if int(grant.get("schema_version", 0)) != 2:
raise AuthorizationError("writer_token_policy_v2_required", 401)
if str(grant.get("audience") or "") != audience:
raise AuthorizationError("writer_token_audience_rejected")
try:
peer = ipaddress.ip_address(source_ip)
networks = [ipaddress.ip_network(str(value), strict=True) for value in grant.get("source_cidrs", [])]
except ValueError:
raise AuthorizationError("writer_token_source_rejected") from None
if not networks or not any(peer in network for network in networks):
raise AuthorizationError("writer_token_source_rejected")
created = _parse_time(grant.get("created_at"))
not_before = _parse_time(grant.get("not_before"))
expires = _parse_time(grant.get("expires_at"))
if expires <= not_before or (expires - created).total_seconds() > 1800:
raise AuthorizationError("writer_token_lifetime_rejected")
if current_time < not_before:
raise AuthorizationError("writer_token_not_yet_valid", 401)
if current_time >= expires:
raise AuthorizationError("writer_token_expired", 401)
if bool(grant.get("revoked")):
raise AuthorizationError("writer_token_revoked", 401)
operations = grant.get("allowed_operations")
if not isinstance(operations, list) or operation not in operations:
raise AuthorizationError("writer_token_operation_rejected")
expected_branch = str(grant.get("allowed_branch") or "")
configured_branch = os.getenv("AIOS_GITOPS_ALLOWED_BRANCH", "")
if expected_branch != configured_branch:
raise AuthorizationError("writer_token_branch_rejected")
if branch is not None and branch != expected_branch:
raise AuthorizationError("writer_token_branch_rejected")
if path is not None:
paths = grant.get("allowed_paths")
if not isinstance(paths, list) or path not in paths:
raise AuthorizationError("writer_token_path_rejected")
base = str(grant.get("approved_base_oid") or "").lower()
if not OID_RE.fullmatch(base):
raise AuthorizationError("writer_token_base_rejected")
if not _git_is_ancestor(workspace, base, expected_branch):
raise AuthorizationError("writer_token_base_not_ancestor")
def authorize(
registry: TokenRegistry,
token: str,
*,
method: str,
path: str,
source_ip: str,
workspace: Path,
branch: str | None = None,
target_path: str | None = None,
current_time: datetime | None = None,
) -> AuthorizedGrant:
if os.getenv("AIOS_GITOPS_TOKEN_POLICY_VERSION_REQUIRED", "") != "2":
raise AuthorizationError("writer_token_policy_v2_not_required")
if not token:
raise AuthorizationError("unauthorized", 401)
grant = registry.find(token)
if grant is None:
raise AuthorizationError("unauthorized", 401)
operation = operation_for(method, path)
_validate_grant(
grant,
audience=os.getenv("AIOS_GITOPS_SERVICE_AUDIENCE", ""),
source_ip=source_ip,
operation=operation,
branch=branch,
path=target_path,
workspace=workspace,
current_time=current_time or datetime.now(timezone.utc),
)
limits = grant.get("max_uses")
if not isinstance(limits, dict):
raise AuthorizationError("writer_token_use_limit_missing")
try:
used = registry.consume(str(grant["token_id"]), operation, int(limits[operation]))
except (KeyError, TypeError, ValueError):
raise AuthorizationError("writer_token_use_limit_missing") from None
except PermissionError as exc:
code = str(exc)
raise AuthorizationError(code, 401 if code == "writer_token_revoked" else 403) from None
return AuthorizedGrant(grant, operation, used)
@@ -1,403 +0,0 @@
"""Atomic, idempotent commit-file Saga for the isolated writer."""
from __future__ import annotations
import fcntl
import hashlib
import json
import os
import re
import sqlite3
import subprocess
import tempfile
import threading
from pathlib import Path
from typing import Any
import aios_isolated_writer_policy as policy
OID_RE = re.compile(r"[0-9a-f]{40}")
IDEMPOTENCY_RE = re.compile(r"[A-Za-z0-9][A-Za-z0-9._:-]{7,127}")
STATES = frozenset({
"PENDING", "FILE_WRITTEN", "COMMITTED", "PUSHED", "VERIFIED", "COMPLETED",
"FAILED_RECOVERABLE", "MANUAL_REVIEW_REQUIRED",
})
_LOCKS: dict[str, threading.Lock] = {}
_LOCKS_GUARD = threading.Lock()
_LEDGER_INIT_LOCK = threading.Lock()
class CommitFileError(RuntimeError):
def __init__(self, code: str, status: int = 400):
super().__init__(code)
self.code = code
self.status = status
class SimulatedCrash(RuntimeError):
pass
def canonical_digest(data: dict[str, Any]) -> str:
payload = {
"branch": data["branch"],
"path": data["path"],
"content_sha256": hashlib.sha256(data["content"].encode("utf-8")).hexdigest(),
"message": data["message"],
"expected_head_oid": data["expected_head_oid"],
}
encoded = json.dumps(payload, sort_keys=True, separators=(",", ":")).encode("utf-8")
return hashlib.sha256(encoded).hexdigest()
class OperationLedger:
def __init__(self, path: Path | str):
self.path = Path(path)
self.path.parent.mkdir(parents=True, exist_ok=True)
if self.path.parent.is_symlink() or self.path.is_symlink():
raise PermissionError("writer_operation_ledger_symlink_rejected")
with _LEDGER_INIT_LOCK:
with self.connect() as db:
db.execute(
"""
CREATE TABLE IF NOT EXISTS writer_operations (
operation_id TEXT PRIMARY KEY,
token_id TEXT NOT NULL,
task_id TEXT NOT NULL,
idempotency_key TEXT NOT NULL UNIQUE,
operation_digest TEXT NOT NULL,
branch TEXT NOT NULL,
path TEXT NOT NULL,
content_sha256 TEXT NOT NULL,
pre_head_oid TEXT NOT NULL,
commit_oid TEXT,
push_status TEXT,
verification_status TEXT,
state TEXT NOT NULL,
created_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP,
updated_at TEXT NOT NULL DEFAULT CURRENT_TIMESTAMP,
error_code TEXT
)
"""
)
os.chmod(self.path, 0o600)
def connect(self) -> sqlite3.Connection:
db = sqlite3.connect(self.path, timeout=30)
db.row_factory = sqlite3.Row
db.execute("PRAGMA journal_mode=WAL")
db.execute("PRAGMA synchronous=FULL")
return db
def get_or_create(self, grant: Any, data: dict[str, Any], digest: str) -> dict[str, Any]:
key = data["idempotency_key"]
with self.connect() as db:
db.execute("BEGIN IMMEDIATE")
row = db.execute(
"SELECT * FROM writer_operations WHERE idempotency_key=?", (key,)
).fetchone()
if row:
db.commit()
record = dict(row)
if record["operation_digest"] != digest:
raise CommitFileError("writer_idempotency_conflict", 409)
if (
record["token_id"] != grant.token_id
or record["task_id"] != grant.task_id
):
raise CommitFileError("writer_idempotency_owner_conflict", 409)
return record
operation_id = hashlib.sha256(
f"{grant.task_id}\0{key}".encode("utf-8")
).hexdigest()
content_sha = hashlib.sha256(data["content"].encode("utf-8")).hexdigest()
db.execute(
"""
INSERT INTO writer_operations(
operation_id,token_id,task_id,idempotency_key,operation_digest,
branch,path,content_sha256,pre_head_oid,state
) VALUES(?,?,?,?,?,?,?,?,?,?)
""",
(
operation_id, grant.token_id, grant.task_id, key, digest,
data["branch"], data["path"], content_sha,
data["expected_head_oid"], "PENDING",
),
)
db.commit()
return self.get(key)
def get(self, key: str) -> dict[str, Any]:
with self.connect() as db:
row = db.execute(
"SELECT * FROM writer_operations WHERE idempotency_key=?", (key,)
).fetchone()
if row is None:
raise KeyError(key)
return dict(row)
def update(self, key: str, state: str, **fields: Any) -> dict[str, Any]:
if state not in STATES:
raise ValueError("invalid writer operation state")
allowed = {"commit_oid", "push_status", "verification_status", "error_code"}
assignments = ["state=?", "updated_at=CURRENT_TIMESTAMP"]
values: list[Any] = [state]
for name, value in fields.items():
if name not in allowed:
raise ValueError("invalid ledger field")
assignments.append(f"{name}=?")
values.append(value)
values.append(key)
with self.connect() as db:
db.execute("BEGIN IMMEDIATE")
db.execute(
f"UPDATE writer_operations SET {','.join(assignments)} WHERE idempotency_key=?",
values,
)
db.commit()
return self.get(key)
class AtomicCommitFile:
def __init__(
self,
*,
workspace: Path,
bare_repo: Path,
ledger: OperationLedger,
remote: str,
audit,
crash_after_state: str | None = None,
):
self.workspace = workspace
self.bare_repo = bare_repo
self.ledger = ledger
self.remote = remote
self.audit = audit
self.crash_after_state = crash_after_state
def _git(self, *args: str, check: bool = True) -> str:
result = subprocess.run(
["git", *args], cwd=self.workspace, text=True, capture_output=True
)
if check and result.returncode != 0:
raise CommitFileError("writer_git_operation_failed", 500)
return result.stdout.strip()
def _transition(self, key: str, state: str, **fields: Any) -> dict[str, Any]:
record = self.ledger.update(key, state, **fields)
if self.crash_after_state == state:
raise SimulatedCrash(state)
return record
def _validate(self, data: Any) -> dict[str, Any]:
if not isinstance(data, dict):
raise CommitFileError("writer_request_object_required")
required = ("branch", "path", "content", "message", "expected_head_oid", "idempotency_key")
if any(field not in data for field in required):
raise CommitFileError("writer_request_incomplete")
if data.get("force") or data.get("force_push"):
raise CommitFileError("isolated_writer_force_rejected", 403)
if set(data) != set(required):
raise CommitFileError("writer_request_fields_rejected")
item = {field: data[field] for field in required}
if item["branch"] != policy.allowed_branch():
raise CommitFileError("isolated_writer_branch_rejected", 403)
try:
item["path"] = policy.validate_path(item["path"], self.workspace)
except PermissionError as exc:
raise CommitFileError(str(exc), 403) from None
if not isinstance(item["content"], str) or "\x00" in item["content"]:
raise CommitFileError("writer_content_rejected")
try:
policy.validate_content_for_path(item["path"], item["content"])
except PermissionError as exc:
raise CommitFileError(str(exc), 403) from None
if len(item["content"].encode("utf-8")) > 1_048_576:
raise CommitFileError("writer_content_too_large", 413)
message = item["message"]
if (
not isinstance(message, str)
or not 1 <= len(message) <= 200
or "\x00" in message
or "\n" in message
or "\r" in message
):
raise CommitFileError("writer_message_rejected")
if not isinstance(item["expected_head_oid"], str) or not OID_RE.fullmatch(item["expected_head_oid"]):
raise CommitFileError("writer_expected_head_rejected")
if not isinstance(item["idempotency_key"], str) or not IDEMPOTENCY_RE.fullmatch(item["idempotency_key"]):
raise CommitFileError("writer_idempotency_key_rejected")
return item
def _atomic_write(self, relative: str, content: str) -> None:
target = self.workspace / relative
target.parent.mkdir(parents=True, exist_ok=True)
fd, temporary = tempfile.mkstemp(prefix=".writer-", dir=target.parent)
try:
with os.fdopen(fd, "wb", closefd=True) as stream:
stream.write(content.encode("utf-8"))
stream.flush()
os.fsync(stream.fileno())
os.replace(temporary, target)
parent_fd = os.open(target.parent, os.O_RDONLY | getattr(os, "O_DIRECTORY", 0))
try:
os.fsync(parent_fd)
finally:
os.close(parent_fd)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def _verify_remote(self, branch: str, oid: str) -> None:
bare = subprocess.run(
["git", "--git-dir", str(self.bare_repo), "rev-parse", f"refs/heads/{branch}"],
text=True, capture_output=True,
)
if bare.returncode != 0 or bare.stdout.strip() != oid:
raise CommitFileError("writer_authoritative_bare_oid_mismatch", 500)
remote = subprocess.run(
["git", "ls-remote", "--heads", self.remote, f"refs/heads/{branch}"],
cwd=self.workspace, text=True, capture_output=True,
)
if remote.returncode != 0:
raise CommitFileError("writer_gitea_verification_failed", 503)
fields = remote.stdout.strip().split()
if len(fields) != 2:
raise CommitFileError("writer_gitea_ref_missing", 503)
if fields[0] != oid:
raise CommitFileError("writer_gitea_oid_conflict", 409)
def execute(self, grant: Any, raw_data: Any) -> dict[str, Any]:
data = self._validate(raw_data)
digest = canonical_digest(data)
branch = data["branch"]
with _LOCKS_GUARD:
lock = _LOCKS.setdefault(branch, threading.Lock())
lock_path = self.ledger.path.parent / f"{hashlib.sha256(branch.encode()).hexdigest()}.lock"
with lock, open(lock_path, "a+b") as branch_lock:
fcntl.flock(branch_lock.fileno(), fcntl.LOCK_EX)
return self._execute_locked(grant, data, digest)
def _execute_locked(self, grant: Any, data: dict[str, Any], digest: str) -> dict[str, Any]:
key = data["idempotency_key"]
record = self.ledger.get_or_create(grant, data, digest)
if record["state"] == "COMPLETED":
return self._result(record, idempotent=True)
if self._git("rev-parse", "--abbrev-ref", "HEAD") != data["branch"]:
record = self._transition(
key, "MANUAL_REVIEW_REQUIRED", error_code="writer_current_branch_conflict"
)
self._audit(grant, data, digest, record, "manual_review_required")
raise CommitFileError("writer_current_branch_conflict", 409)
head = self._git("rev-parse", "HEAD")
commit_oid = record.get("commit_oid")
if head != record["pre_head_oid"] and head != commit_oid:
record = self._transition(
key, "MANUAL_REVIEW_REQUIRED", error_code="writer_branch_head_conflict"
)
self._audit(grant, data, digest, record, "manual_review_required")
raise CommitFileError("writer_branch_head_conflict", 409)
try:
state = record["state"]
if state == "MANUAL_REVIEW_REQUIRED":
raise CommitFileError("writer_manual_review_required", 409)
if state == "FAILED_RECOVERABLE":
if record.get("push_status") == "confirmed":
state = "PUSHED"
else:
state = "COMMITTED" if commit_oid else "FILE_WRITTEN"
if state == "PENDING":
self._atomic_write(data["path"], data["content"])
record = self._transition(key, "FILE_WRITTEN")
state = record["state"]
if state == "FILE_WRITTEN":
actual_sha = hashlib.sha256(
(self.workspace / data["path"]).read_bytes()
).hexdigest()
if actual_sha != record["content_sha256"]:
raise CommitFileError("writer_file_content_conflict", 409)
staged_before = self._git("diff", "--cached", "--name-only", check=False).splitlines()
if staged_before:
raise CommitFileError("writer_extra_staged_files_rejected", 409)
self._git("add", "--", data["path"])
staged = self._git("diff", "--cached", "--name-only").splitlines()
if staged != [data["path"]]:
raise CommitFileError("writer_extra_staged_files_rejected", 409)
self._git("commit", "-m", data["message"])
commit_oid = self._git("rev-parse", "HEAD")
record = self._transition(key, "COMMITTED", commit_oid=commit_oid)
state = record["state"]
if state == "COMMITTED":
commit_oid = str(record["commit_oid"])
pushed = subprocess.run(
["git", "push", "--porcelain", self.remote, f"HEAD:refs/heads/{data['branch']}"],
cwd=self.workspace, text=True, capture_output=True,
)
if pushed.returncode != 0:
self._transition(
key, "FAILED_RECOVERABLE",
push_status="failed", error_code="writer_gitea_push_failed",
)
raise CommitFileError("writer_gitea_push_failed", 503)
record = self._transition(key, "PUSHED", push_status="confirmed")
state = record["state"]
if state == "PUSHED":
commit_oid = str(record["commit_oid"])
self._verify_remote(data["branch"], commit_oid)
record = self._transition(
key, "VERIFIED", verification_status="bare+gitea"
)
state = record["state"]
if state == "VERIFIED":
record = self._transition(key, "COMPLETED", error_code=None)
self._audit(grant, data, digest, record, "completed")
return self._result(record, idempotent=False)
except SimulatedCrash:
raise
except CommitFileError as exc:
latest = self.ledger.get(key)
if latest["state"] not in {"FAILED_RECOVERABLE", "MANUAL_REVIEW_REQUIRED"}:
failure_state = (
"MANUAL_REVIEW_REQUIRED"
if exc.code in {
"writer_gitea_oid_conflict",
"writer_file_content_conflict",
"writer_current_branch_conflict",
}
else "FAILED_RECOVERABLE"
)
latest = self._transition(key, failure_state, error_code=exc.code)
self._audit(grant, data, digest, latest, "failed")
raise
def _audit(self, grant: Any, data: dict[str, Any], digest: str, record: dict[str, Any], result: str) -> None:
self.audit({
"token_id": grant.token_id,
"task_id": grant.task_id,
"audience": grant.record["audience"],
"source_ip": grant.record.get("_source_ip", "socket-peer"),
"operation": grant.operation,
"branch": data["branch"],
"path": data["path"],
"idempotency_key": data["idempotency_key"],
"operation_digest": digest,
"path_policy_sha256": policy.configured_path_policy_sha256(),
"pre_head_oid": record["pre_head_oid"],
"commit_oid": record.get("commit_oid"),
"result": result,
})
@staticmethod
def _result(record: dict[str, Any], *, idempotent: bool) -> dict[str, Any]:
return {
"status": "completed",
"state": record["state"],
"operation_id": record["operation_id"],
"branch": record["branch"],
"path": record["path"],
"commit_oid": record.get("commit_oid"),
"push_status": record.get("push_status"),
"verification_status": record.get("verification_status"),
"idempotent": idempotent,
}
@@ -1,422 +0,0 @@
"""Strict service-level exact-path policy for isolated Git writers."""
from __future__ import annotations
import grp
import hashlib
import json
import os
import re
import stat
from dataclasses import dataclass
from functools import lru_cache
from pathlib import Path, PurePosixPath, PureWindowsPath
from typing import Any
from urllib.parse import unquote
MAX_POLICY_BYTES = 65_536
POLICY_FIELDS = frozenset(
{
"schema_version",
"policy_id",
"audience",
"repository_identity",
"allowed_branch",
"allowed_operations",
}
)
OPERATION_FIELDS = frozenset({"exact_paths", "max_content_bytes", "content_profile"})
SUPPORTED_OPERATIONS = frozenset({"commit_file.atomic"})
SUPPORTED_CONTENT_PROFILES = frozenset({"systemd-codex-review-worker-v1"})
_SECRET_RE = re.compile(
r"(?i)(authorization\s*:|bearer\s+|(?:token|password|secret)\s*[:=])"
)
_SYSTEMD_NAME_RE = re.compile(r"[A-Za-z][A-Za-z0-9]*\Z")
_SYSTEMD_MAX_LINE_BYTES = 4096
_SYSTEMD_ALLOWED_DIRECTIVES = {
"Unit": frozenset({"Description", "After", "Wants"}),
"Service": frozenset(
{
"Type",
"User",
"Group",
"WorkingDirectory",
"Environment",
"ExecStart",
"Restart",
"RestartSec",
"NoNewPrivileges",
"PrivateTmp",
"PrivateDevices",
"ProtectSystem",
"ProtectHome",
"ProtectKernelTunables",
"ProtectKernelModules",
"ProtectControlGroups",
"LockPersonality",
"RestrictSUIDSGID",
"RestrictRealtime",
"MemoryDenyWriteExecute",
"RuntimeDirectory",
"StateDirectory",
"CacheDirectory",
"ReadOnlyPaths",
"ReadWritePaths",
"UMask",
}
),
"Install": frozenset({"WantedBy"}),
}
_SYSTEMD_MULTI_DIRECTIVES = frozenset(
{
("Service", "Environment"),
("Service", "ReadOnlyPaths"),
("Service", "ReadWritePaths"),
}
)
_SYSTEMD_EXACT_SINGLE_VALUES = {
("Unit", "Description"): "AIOS Codex Review/Test Worker",
("Unit", "After"): "network-online.target",
("Unit", "Wants"): "network-online.target",
("Service", "Type"): "simple",
("Service", "User"): "aiman",
("Service", "Group"): "gitops",
(
"Service",
"WorkingDirectory",
): "/usr/libexec/aios-codex-review-test-worker/current",
(
"Service",
"ExecStart",
): "/usr/bin/python3 -m task_plane.workers.codex_review_test",
("Service", "Restart"): "on-failure",
("Service", "RestartSec"): "2",
("Service", "NoNewPrivileges"): "true",
("Service", "PrivateTmp"): "true",
("Service", "PrivateDevices"): "true",
("Service", "ProtectSystem"): "strict",
("Service", "ProtectHome"): "true",
("Service", "ProtectKernelTunables"): "true",
("Service", "ProtectKernelModules"): "true",
("Service", "ProtectControlGroups"): "true",
("Service", "LockPersonality"): "true",
("Service", "RestrictSUIDSGID"): "true",
("Service", "RestrictRealtime"): "true",
("Service", "MemoryDenyWriteExecute"): "true",
("Service", "RuntimeDirectory"): "aios-codex-review-test-worker",
("Service", "StateDirectory"): "aios-codex-review-test-worker",
("Service", "CacheDirectory"): "aios-codex-review-test-worker",
("Service", "UMask"): "0077",
("Install", "WantedBy"): "multi-user.target",
}
_SYSTEMD_EXACT_MULTI_VALUES = {
("Service", "Environment"): (
"AIOS_CODEX_WORKER_MODE=shadow",
"AIOS_TASK_PLANE_EXECUTION_ENABLED=false",
"PYTHONDONTWRITEBYTECODE=1",
),
("Service", "ReadOnlyPaths"): (
"/data/aios-gitops/repo.git",
"/usr/libexec/aios-codex-review-test-worker",
),
("Service", "ReadWritePaths"): (
"/run/aios-codex-review-test-worker",
"/var/lib/aios-codex-review-test-worker",
"/var/cache/aios-codex-review-test-worker",
),
}
class PathPolicyError(PermissionError):
pass
@dataclass(frozen=True)
class OperationPolicy:
exact_paths: tuple[str, ...]
max_content_bytes: int
content_profile: str
@dataclass(frozen=True)
class ServicePathPolicy:
schema_version: int
policy_id: str
audience: str
repository_identity: str
allowed_branch: str
allowed_operations: dict[str, OperationPolicy]
sha256: str
def operation(self, name: str) -> OperationPolicy:
try:
return self.allowed_operations[name]
except KeyError:
raise PathPolicyError("isolated_writer_service_operation_rejected") from None
def require_exact_path(self, operation: str, path: str) -> OperationPolicy:
selected = self.operation(operation)
if path not in selected.exact_paths:
raise PathPolicyError("isolated_writer_service_path_rejected")
return selected
def _canonical_json(value: Any) -> bytes:
return (
json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
+ "\n"
).encode("utf-8")
def validate_relative_path_syntax(path: str) -> str:
if not isinstance(path, str):
raise PathPolicyError("isolated_writer_path_rejected")
if not path or path != path.strip() or "\x00" in path or "\\" in path:
raise PathPolicyError("isolated_writer_path_rejected")
try:
decoded = unquote(path, errors="strict")
except (UnicodeDecodeError, ValueError):
raise PathPolicyError("isolated_writer_path_rejected") from None
if decoded != path:
raise PathPolicyError("isolated_writer_path_rejected")
posix = PurePosixPath(path)
windows = PureWindowsPath(path)
parts = path.split("/")
if (
posix.is_absolute()
or windows.is_absolute()
or bool(windows.drive)
or any(part in {"", ".", ".."} for part in parts)
):
raise PathPolicyError("isolated_writer_path_rejected")
return path
def _require_fields(value: dict[str, Any], allowed: frozenset[str], code: str) -> None:
if set(value) != allowed:
raise PathPolicyError(code)
def _parse_policy(document: Any, canonical: bytes) -> ServicePathPolicy:
if not isinstance(document, dict):
raise PathPolicyError("isolated_writer_path_policy_object_required")
_require_fields(document, POLICY_FIELDS, "isolated_writer_path_policy_fields_rejected")
if document["schema_version"] != 1:
raise PathPolicyError("isolated_writer_path_policy_schema_rejected")
text_fields = (
"policy_id",
"audience",
"repository_identity",
"allowed_branch",
)
if any(
not isinstance(document[field], str) or not document[field].strip()
for field in text_fields
):
raise PathPolicyError("isolated_writer_path_policy_value_rejected")
raw_operations = document["allowed_operations"]
if not isinstance(raw_operations, dict) or not raw_operations:
raise PathPolicyError("isolated_writer_path_policy_operations_rejected")
if not set(raw_operations).issubset(SUPPORTED_OPERATIONS):
raise PathPolicyError("isolated_writer_path_policy_operations_rejected")
operations: dict[str, OperationPolicy] = {}
for name, raw in raw_operations.items():
if not isinstance(raw, dict):
raise PathPolicyError("isolated_writer_path_policy_operation_rejected")
_require_fields(
raw,
OPERATION_FIELDS,
"isolated_writer_path_policy_operation_fields_rejected",
)
paths = raw["exact_paths"]
if (
not isinstance(paths, list)
or not paths
or any(not isinstance(item, str) for item in paths)
):
raise PathPolicyError("isolated_writer_path_policy_paths_rejected")
validated = tuple(validate_relative_path_syntax(item) for item in paths)
if len(set(validated)) != len(validated):
raise PathPolicyError("isolated_writer_path_policy_paths_rejected")
maximum = raw["max_content_bytes"]
if not isinstance(maximum, int) or isinstance(maximum, bool) or not 1 <= maximum <= 1_048_576:
raise PathPolicyError("isolated_writer_path_policy_size_rejected")
profile = raw["content_profile"]
if profile not in SUPPORTED_CONTENT_PROFILES:
raise PathPolicyError("isolated_writer_path_policy_profile_rejected")
operations[name] = OperationPolicy(validated, maximum, profile)
return ServicePathPolicy(
schema_version=1,
policy_id=document["policy_id"],
audience=document["audience"],
repository_identity=document["repository_identity"],
allowed_branch=document["allowed_branch"],
allowed_operations=operations,
sha256=hashlib.sha256(canonical).hexdigest(),
)
def load_path_policy(
path: Path,
*,
expected_owner_uid: int = 0,
expected_group_gid: int | None = None,
expected_parent_mode: int = 0o750,
expected_file_mode: int = 0o640,
) -> ServicePathPolicy:
lexical = Path(os.path.abspath(str(path)))
if lexical.is_symlink():
raise PathPolicyError("isolated_writer_path_policy_symlink_rejected")
try:
parent_info = lexical.parent.stat(follow_symlinks=False)
file_info = lexical.stat(follow_symlinks=False)
except FileNotFoundError:
raise PathPolicyError("isolated_writer_path_policy_missing") from None
if stat.S_ISLNK(parent_info.st_mode) or not stat.S_ISDIR(parent_info.st_mode):
raise PathPolicyError("isolated_writer_path_policy_parent_rejected")
if stat.S_ISLNK(file_info.st_mode) or not stat.S_ISREG(file_info.st_mode):
raise PathPolicyError("isolated_writer_path_policy_file_rejected")
if lexical.resolve(strict=True) != lexical:
raise PathPolicyError("isolated_writer_path_policy_symlink_rejected")
if (
parent_info.st_uid != expected_owner_uid
or stat.S_IMODE(parent_info.st_mode) != expected_parent_mode
):
raise PathPolicyError("isolated_writer_path_policy_parent_permissions_rejected")
if (
file_info.st_uid != expected_owner_uid
or stat.S_IMODE(file_info.st_mode) != expected_file_mode
):
raise PathPolicyError("isolated_writer_path_policy_permissions_rejected")
if expected_group_gid is not None and (
parent_info.st_gid != expected_group_gid or file_info.st_gid != expected_group_gid
):
raise PathPolicyError("isolated_writer_path_policy_group_rejected")
if file_info.st_size <= 0 or file_info.st_size > MAX_POLICY_BYTES:
raise PathPolicyError("isolated_writer_path_policy_file_size_rejected")
raw = lexical.read_bytes()
try:
document = json.loads(raw.decode("utf-8"))
except (UnicodeDecodeError, json.JSONDecodeError):
raise PathPolicyError("isolated_writer_path_policy_json_rejected") from None
canonical = _canonical_json(document)
if raw != canonical:
raise PathPolicyError("isolated_writer_path_policy_not_canonical")
return _parse_policy(document, canonical)
@lru_cache(maxsize=1)
def configured_path_policy() -> ServicePathPolicy | None:
configured = os.getenv("AIOS_GITOPS_PATH_POLICY_FILE", "").strip()
if not configured:
return None
if not Path(configured).is_absolute():
raise PathPolicyError("isolated_writer_path_policy_path_rejected")
try:
gitops_gid = grp.getgrnam("gitops").gr_gid
except KeyError:
raise PathPolicyError("isolated_writer_path_policy_group_rejected") from None
loaded = load_path_policy(Path(configured), expected_group_gid=gitops_gid)
if loaded.audience != os.getenv("AIOS_GITOPS_SERVICE_AUDIENCE", ""):
raise PathPolicyError("isolated_writer_path_policy_audience_rejected")
if loaded.repository_identity != os.getenv("AIOS_GITOPS_REPOSITORY_IDENTITY", ""):
raise PathPolicyError("isolated_writer_path_policy_repository_rejected")
if loaded.allowed_branch != os.getenv("AIOS_GITOPS_ALLOWED_BRANCH", ""):
raise PathPolicyError("isolated_writer_path_policy_branch_rejected")
return loaded
def clear_path_policy_cache() -> None:
configured_path_policy.cache_clear()
def require_service_path(operation: str, path: str) -> OperationPolicy:
loaded = configured_path_policy()
if loaded is None:
raise PathPolicyError("isolated_writer_service_path_policy_required")
return loaded.require_exact_path(operation, path)
def configured_path_policy_sha256() -> str | None:
loaded = configured_path_policy()
return None if loaded is None else loaded.sha256
def validate_content(operation: str, path: str, content: str) -> None:
selected = require_service_path(operation, path)
try:
encoded = content.encode("utf-8", errors="strict")
except (AttributeError, UnicodeEncodeError):
raise PathPolicyError("isolated_writer_systemd_profile_rejected") from None
if len(encoded) > selected.max_content_bytes:
raise PathPolicyError("isolated_writer_service_content_too_large")
if selected.content_profile == "systemd-codex-review-worker-v1":
_validate_systemd_codex_review_worker(content)
return
raise PathPolicyError("isolated_writer_path_policy_profile_rejected")
def _validate_systemd_codex_review_worker(content: str) -> None:
if not isinstance(content, str):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
try:
content.encode("utf-8", errors="strict")
except UnicodeEncodeError:
raise PathPolicyError("isolated_writer_systemd_profile_rejected") from None
if (
"\x00" in content
or "\r" in content
or _SECRET_RE.search(content)
or not content.endswith("\n")
):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
sections: set[str] = set()
values: dict[tuple[str, str], list[str]] = {}
section: str | None = None
for line in content.splitlines():
if len(line.encode("utf-8")) > _SYSTEMD_MAX_LINE_BYTES:
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
if not line:
continue
if line[0].isspace() or line.startswith(("#", ";")):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
if line.startswith("[") and line.endswith("]"):
candidate = line[1:-1]
if (
not _SYSTEMD_NAME_RE.fullmatch(candidate)
or candidate not in _SYSTEMD_ALLOWED_DIRECTIVES
or candidate in sections
):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
sections.add(candidate)
section = candidate
continue
if section is None or "=" not in line:
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
directive, value = line.split("=", 1)
if (
not _SYSTEMD_NAME_RE.fullmatch(directive)
or directive not in _SYSTEMD_ALLOWED_DIRECTIVES[section]
or not value
or value != value.strip()
):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
key = (section, directive)
if key in values and key not in _SYSTEMD_MULTI_DIRECTIVES:
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
values.setdefault(key, []).append(value)
if sections != set(_SYSTEMD_ALLOWED_DIRECTIVES):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
if set(values) != set(_SYSTEMD_EXACT_SINGLE_VALUES) | set(
_SYSTEMD_EXACT_MULTI_VALUES
):
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
for key, expected in _SYSTEMD_EXACT_SINGLE_VALUES.items():
if values.get(key) != [expected]:
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
for key, expected in _SYSTEMD_EXACT_MULTI_VALUES.items():
observed = values.get(key, [])
if len(observed) != len(set(observed)) or tuple(observed) != expected:
raise PathPolicyError("isolated_writer_systemd_profile_rejected")
@@ -1,241 +0,0 @@
"""Digest-only, task-scoped token registry for isolated writers."""
from __future__ import annotations
import argparse
import hashlib
import hmac
import json
import os
import grp
import pwd
import secrets
import sqlite3
import stat
import threading
from datetime import datetime, timezone
from pathlib import Path
from typing import Any
_REGISTRY_INIT_LOCK = threading.Lock()
def utc_now() -> str:
return datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
def token_digest(token: str) -> str:
return hashlib.sha256(token.encode("utf-8")).hexdigest()
class TokenRegistry:
def __init__(self, path: Path | str):
self.path = Path(path)
self._prepare()
def _prepare(self) -> None:
self.path.parent.mkdir(parents=True, exist_ok=True)
if self.path.parent.is_symlink() or self.path.is_symlink():
raise PermissionError("writer_token_registry_symlink_rejected")
with _REGISTRY_INIT_LOCK:
with self._connect() as db:
db.executescript(
"""
CREATE TABLE IF NOT EXISTS token_grants (
token_id TEXT PRIMARY KEY,
token_digest TEXT NOT NULL UNIQUE,
grant_json TEXT NOT NULL,
revoked INTEGER NOT NULL DEFAULT 0,
created_at TEXT NOT NULL
);
CREATE TABLE IF NOT EXISTS token_uses (
token_id TEXT NOT NULL,
operation TEXT NOT NULL,
use_count INTEGER NOT NULL DEFAULT 0,
PRIMARY KEY(token_id, operation),
FOREIGN KEY(token_id) REFERENCES token_grants(token_id)
);
"""
)
os.chmod(self.path, 0o600)
def _connect(self) -> sqlite3.Connection:
db = sqlite3.connect(self.path, timeout=30, isolation_level=None)
db.row_factory = sqlite3.Row
db.execute("PRAGMA journal_mode=WAL")
db.execute("PRAGMA synchronous=FULL")
db.execute("PRAGMA foreign_keys=ON")
return db
def issue(self, grant: dict[str, Any], token: str | None = None) -> tuple[str, dict[str, Any]]:
value = token or secrets.token_urlsafe(32)
if len(value.encode("utf-8")) < 32:
raise ValueError("writer_token_entropy_insufficient")
record = dict(grant)
record["schema_version"] = 2
record["token_digest"] = token_digest(value)
record["revoked"] = False
record.setdefault("created_at", utc_now())
required = (
"token_id",
"audience",
"task_id",
"source_cidrs",
"allowed_branch",
"approved_base_oid",
"allowed_operations",
"allowed_paths",
"not_before",
"expires_at",
"max_uses",
"issued_by",
)
if any(not record.get(field) for field in required):
raise ValueError("writer_token_grant_incomplete")
encoded = json.dumps(record, sort_keys=True, separators=(",", ":"))
with self._connect() as db:
db.execute("BEGIN IMMEDIATE")
try:
db.execute(
"""
INSERT INTO token_grants(token_id, token_digest, grant_json, revoked, created_at)
VALUES(?,?,?,?,?)
""",
(
str(record["token_id"]),
str(record["token_digest"]),
encoded,
0,
str(record["created_at"]),
),
)
db.execute("COMMIT")
except Exception:
db.execute("ROLLBACK")
raise
return value, record
def find(self, token: str) -> dict[str, Any] | None:
candidate = token_digest(token)
with self._connect() as db:
rows = db.execute(
"SELECT token_digest, grant_json, revoked FROM token_grants"
).fetchall()
match = None
for row in rows:
if hmac.compare_digest(str(row["token_digest"]), candidate):
match = dict(json.loads(str(row["grant_json"])))
match["revoked"] = bool(row["revoked"])
return match
def consume(self, token_id: str, operation: str, maximum: int) -> int:
if maximum < 1:
raise PermissionError("writer_token_operation_exhausted")
with self._connect() as db:
db.execute("BEGIN IMMEDIATE")
try:
row = db.execute(
"SELECT revoked FROM token_grants WHERE token_id=?",
(token_id,),
).fetchone()
if row is None or bool(row["revoked"]):
raise PermissionError("writer_token_revoked")
used_row = db.execute(
"SELECT use_count FROM token_uses WHERE token_id=? AND operation=?",
(token_id, operation),
).fetchone()
used = int(used_row["use_count"]) if used_row else 0
if used >= maximum:
raise PermissionError("writer_token_operation_exhausted")
current = used + 1
db.execute(
"""
INSERT INTO token_uses(token_id, operation, use_count)
VALUES(?,?,?)
ON CONFLICT(token_id, operation)
DO UPDATE SET use_count=excluded.use_count
""",
(token_id, operation, current),
)
db.execute("COMMIT")
return current
except Exception:
db.execute("ROLLBACK")
raise
def revoke(self, token_id: str) -> bool:
with self._connect() as db:
db.execute("BEGIN IMMEDIATE")
changed = db.execute(
"UPDATE token_grants SET revoked=1 WHERE token_id=? AND revoked=0",
(token_id,),
).rowcount
db.execute("COMMIT")
return bool(changed)
def usage(self, token_id: str, operation: str) -> int:
with self._connect() as db:
row = db.execute(
"SELECT use_count FROM token_uses WHERE token_id=? AND operation=?",
(token_id, operation),
).fetchone()
return int(row["use_count"]) if row else 0
def _write_secret_file(
path: Path,
token: str,
*,
owner: str | None = None,
group: str | None = None,
) -> None:
path.parent.mkdir(parents=True, exist_ok=True)
if path.parent.is_symlink() or path.is_symlink():
raise PermissionError("writer_token_file_symlink_rejected")
flags = os.O_WRONLY | os.O_CREAT | os.O_EXCL
if hasattr(os, "O_NOFOLLOW"):
flags |= os.O_NOFOLLOW
fd = os.open(path, flags, 0o600)
try:
if not stat.S_ISREG(os.fstat(fd).st_mode):
raise PermissionError("writer_token_file_not_regular")
if owner is not None or group is not None:
uid = pwd.getpwnam(owner).pw_uid if owner is not None else -1
gid = grp.getgrnam(group).gr_gid if group is not None else -1
os.fchown(fd, uid, gid)
os.write(fd, token.encode("utf-8"))
os.fsync(fd)
finally:
os.close(fd)
def main(argv: list[str] | None = None) -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--registry", required=True)
sub = parser.add_subparsers(dest="command", required=True)
issue = sub.add_parser("issue")
issue.add_argument("--grant-file", required=True)
issue.add_argument("--token-file", required=True)
issue.add_argument("--owner")
issue.add_argument("--group")
revoke = sub.add_parser("revoke")
revoke.add_argument("--token-id", required=True)
args = parser.parse_args(argv)
registry = TokenRegistry(args.registry)
if args.command == "issue":
grant = json.loads(Path(args.grant_file).read_text(encoding="utf-8"))
token, record = registry.issue(grant)
_write_secret_file(
Path(args.token_file),
token,
owner=args.owner,
group=args.group,
)
print(json.dumps({"token_id": record["token_id"], "created": True}, sort_keys=True))
return 0
changed = registry.revoke(args.token_id)
print(json.dumps({"token_id": args.token_id, "revoked": changed}, sort_keys=True))
return 0
if __name__ == "__main__":
raise SystemExit(main())
File diff suppressed because it is too large Load Diff
@@ -1,564 +0,0 @@
#!/usr/bin/env python3
"""Execution-readiness adapters and the fail-closed Shadow controller service."""
from __future__ import annotations
import argparse
import base64
import dataclasses
import hashlib
import json
import os
import subprocess
import threading
import time
from http import HTTPStatus
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from pathlib import Path
from typing import Any, Callable, Mapping, Protocol
from urllib.parse import parse_qs, urlparse
from cryptography.hazmat.primitives.asymmetric.ed25519 import (
Ed25519PrivateKey,
)
from silent_release_controller import _sanitize, utc_now
from silent_release_hardening import (
BuilderAttestation,
CapabilityAuthorizationError,
ReleaseRecoveryWorker,
SignedApprovalVerifier,
SignedReleasePlanApproval,
SilentReleaseActionService,
TrustedCandidateBundle,
_canonical_json,
_durable_json,
_safe_child,
)
class BuilderSigner(Protocol):
@property
def key_id(self) -> str:
...
def sign(self, payload: bytes) -> bytes:
...
class LocalTestBuilderSigner:
"""Ephemeral in-memory signer. It is forbidden for production execution."""
test_only = True
def __init__(self, key_id: str, private_key: Ed25519PrivateKey):
self._key_id = str(key_id)
self._private_key = private_key
@property
def key_id(self) -> str:
return self._key_id
@classmethod
def generate(
cls, key_id: str = "shadow-builder-test-key"
) -> "LocalTestBuilderSigner":
return cls(key_id, Ed25519PrivateKey.generate())
@property
def public_key(self):
return self._private_key.public_key()
def sign(self, payload: bytes) -> bytes:
return self._private_key.sign(payload)
class ExternalKmsBuilderSigner:
"""Signer facade whose private operation remains in KMS/HSM."""
def __init__(
self,
key_id: str,
credential_reference: str,
sign_callback: Callable[[str, bytes], bytes],
):
if not str(credential_reference).startswith(
("kms://", "hsm://", "secret-manager://")
):
raise ValueError("production builder requires a KMS/HSM reference")
self._key_id = str(key_id)
self.credential_reference = str(credential_reference)
self._sign_callback = sign_callback
@property
def key_id(self) -> str:
return self._key_id
def sign(self, payload: bytes) -> bytes:
return bytes(self._sign_callback(self.credential_reference, payload))
class SignedTrustedBuilderAdapter:
"""Signs evidence produced by a trusted collector, never caller evidence."""
def __init__(
self,
collector: Callable[[Mapping[str, Any]], TrustedCandidateBundle],
signer: BuilderSigner,
):
self.collector = collector
self.signer = signer
def build(self, intent: Mapping[str, Any]) -> TrustedCandidateBundle:
forbidden = {
"artifact_sha256",
"canary",
"checks",
"manifest_sha256",
"tests",
"tree_oid",
}
if forbidden.intersection(intent):
raise CapabilityAuthorizationError(
"trusted builder rejects caller-supplied evidence"
)
unsigned = self.collector(dict(intent))
if unsigned.attestation is not None:
raise CapabilityAuthorizationError(
"collector cannot self-attest its evidence"
)
candidate = unsigned.candidate
test_sha = hashlib.sha256(
_canonical_json(_sanitize(candidate.tests))
).hexdigest()
canary_sha = hashlib.sha256(
_canonical_json(_sanitize(candidate.canary))
).hexdigest()
attestation_id = "builder-attestation-" + hashlib.sha256(
(
candidate.candidate_id
+ "\0"
+ candidate.source_commit
+ "\0"
+ candidate.artifact_sha256
).encode("utf-8")
).hexdigest()[:24]
attestation = BuilderAttestation(
builder_attestation_id=attestation_id,
builder_key_id=self.signer.key_id,
source_commit=candidate.source_commit,
tree_oid=candidate.tree_oid,
artifact_sha256=candidate.artifact_sha256,
manifest_sha256=candidate.manifest_sha256,
package_closure_sha256=(
unsigned.release_plan.deployment.package_closure_sha256
),
test_evidence_sha256=test_sha,
canary_evidence_sha256=canary_sha,
issued_at=utc_now(),
signature="",
)
attestation = dataclasses.replace(
attestation,
signature=base64.b64encode(
self.signer.sign(attestation.canonical_bytes())
).decode("ascii"),
)
return dataclasses.replace(unsigned, attestation=attestation)
class ApprovalBackendAdapter(Protocol):
@property
def available(self) -> bool:
...
def verify(
self,
approval: SignedReleasePlanApproval,
*,
candidate,
release_plan,
policy_sha256: str,
) -> SignedReleasePlanApproval:
...
class LocalTestApprovalSigner:
"""Test-only approval signer used solely for Shadow acceptance."""
test_only = True
def __init__(self, key_id: str, private_key: Ed25519PrivateKey):
self.key_id = str(key_id)
self.private_key = private_key
@classmethod
def generate(
cls, key_id: str = "shadow-owner-test-key"
) -> "LocalTestApprovalSigner":
return cls(key_id, Ed25519PrivateKey.generate())
def sign(
self, approval: SignedReleasePlanApproval
) -> SignedReleasePlanApproval:
unsigned = dataclasses.replace(
approval, key_id=self.key_id, signature=""
)
return dataclasses.replace(
unsigned,
signature=base64.b64encode(
self.private_key.sign(unsigned.canonical_bytes())
).decode("ascii"),
)
class ProductionApprovalBackendAdapter:
"""Verification-only production adapter configured by a secret/KMS ref."""
def __init__(
self,
verifier: SignedApprovalVerifier,
credential_reference: str,
):
if not str(credential_reference).startswith(
("kms://", "hsm://", "secret-manager://")
):
raise ValueError(
"production approval backend requires a KMS/secret reference"
)
self.verifier = verifier
self.credential_reference = str(credential_reference)
@property
def available(self) -> bool:
return True
def verify(
self,
approval: SignedReleasePlanApproval,
*,
candidate,
release_plan,
policy_sha256: str,
) -> SignedReleasePlanApproval:
return self.verifier.verify_release_plan(
approval, candidate, release_plan, policy_sha256
)
class UnavailableProductionApprovalBackend:
credential_reference = None
@property
def available(self) -> bool:
return False
def verify(self, *args, **kwargs):
raise CapabilityAuthorizationError(
"production approval backend is not configured"
)
class LeaderLock:
def __init__(self, path: Path):
self.path = Path(path)
self.handle = None
def __enter__(self):
self.path.parent.mkdir(parents=True, exist_ok=True)
os.chmod(self.path.parent, 0o700)
self.handle = self.path.open("a+", encoding="utf-8")
os.chmod(self.path, 0o600)
try:
import fcntl
fcntl.flock(
self.handle.fileno(), fcntl.LOCK_EX | fcntl.LOCK_NB
)
except BlockingIOError as exc:
self.handle.close()
self.handle = None
raise RuntimeError("silent release leader is already active") from exc
return self
def __exit__(self, exc_type, exc, tb):
if self.handle is not None:
try:
import fcntl
fcntl.flock(self.handle.fileno(), fcntl.LOCK_UN)
finally:
self.handle.close()
self.handle = None
class PersistentSilentReleaseController:
"""Idempotent main-event controller with durable startup/recovery reports."""
def __init__(
self,
root: Path,
service: SilentReleaseActionService,
*,
main_probe: Callable[[], str],
recovery_worker: ReleaseRecoveryWorker | None = None,
):
self.root = Path(root)
self.service = service
self.main_probe = main_probe
self.recovery_worker = recovery_worker
self.leader_path = self.root / "controller.leader.lock"
self.status_path = self.root / "controller-status.json"
def startup_verify(self) -> dict[str, Any]:
checked = 0
failed = 0
for journal in sorted(self.service.sagas.root.glob("*.events.jsonl")):
candidate_id = journal.name[: -len(".events.jsonl")]
saga = self.service.sagas.load(candidate_id)
checked += 1
if not saga.journal_integrity:
failed += 1
status = {
"checked_sagas": checked,
"execution_enabled": self.service.execution_enabled,
"failed_closed": failed,
"shadow_mode": not self.service.execution_enabled,
"side_effects_executed": False,
"status": "READY" if failed == 0 else "FAIL_CLOSED",
}
_durable_json(self.status_path, status)
return status
def run_once(self) -> dict[str, Any]:
with LeaderLock(self.leader_path):
startup = self.startup_verify()
if startup["status"] != "READY":
result = {
"execution_enabled": self.service.execution_enabled,
"intent": None,
"notification": None,
"recovery": None,
"shadow_mode": not self.service.execution_enabled,
"side_effects_executed": False,
"startup": startup,
"status": "FAIL_CLOSED",
}
_durable_json(self.status_path, result)
return _sanitize(result)
source_commit = str(self.main_probe())
event_id = f"main-{source_commit}"
intent = self.service.create_release_intent(
{"event_id": event_id, "source_commit": source_commit}
)
notification = None
if intent.get("candidate_id"):
notification = self.service.ensure_owner_notification(
str(intent["candidate_id"])
)
recovery = (
self.recovery_worker.run_once()
if self.recovery_worker is not None
else self.service.get_recovery_status()
)
result = {
"execution_enabled": self.service.execution_enabled,
"intent": intent,
"notification": notification,
"recovery": recovery,
"shadow_mode": not self.service.execution_enabled,
"side_effects_executed": False,
"startup": startup,
}
_durable_json(self.status_path, result)
return _sanitize(result)
class ShadowActionHandler(BaseHTTPRequestHandler):
service: SilentReleaseActionService
controller: PersistentSilentReleaseController
def log_message(self, fmt, *args):
return
def _json(self, payload: Mapping[str, Any], status=HTTPStatus.OK):
body = json.dumps(_sanitize(dict(payload)), sort_keys=True).encode()
self.send_response(int(status))
self.send_header("Content-Type", "application/json")
self.send_header("Content-Length", str(len(body)))
self.end_headers()
self.wfile.write(body)
def _body(self) -> dict[str, Any]:
size = int(self.headers.get("Content-Length", "0"))
if size > 1024 * 1024:
raise ValueError("request body is too large")
return json.loads(self.rfile.read(size) or b"{}")
def do_GET(self):
parsed = urlparse(self.path)
query = parse_qs(parsed.query)
candidate_id = str((query.get("candidate_id") or [""])[0])
try:
if parsed.path == "/healthz":
return self._json(
{
"execution_enabled": False,
"shadow_mode": True,
"side_effects_executed": False,
"status": "ok",
}
)
routes = {
"/getReleaseCandidate": self.service.get_release_candidate,
"/getReleasePlan": self.service.get_release_plan,
"/getReleaseAcceptance": self.service.get_acceptance,
"/getReleaseSaga": self.service.get_release_saga,
"/getCapabilityStatus": self.service.get_capability_status,
}
if parsed.path in routes:
return self._json(routes[parsed.path](candidate_id))
if parsed.path == "/getRecoveryStatus":
return self._json(self.service.get_recovery_status())
self._json({"error": "not_found"}, HTTPStatus.NOT_FOUND)
except (ValueError, FileNotFoundError) as exc:
self._json(
{"error": type(exc).__name__}, HTTPStatus.BAD_REQUEST
)
def do_POST(self):
parsed = urlparse(self.path)
try:
body = self._body()
if parsed.path == "/createReleaseIntent":
return self._json(self.service.create_release_intent(body))
if parsed.path in {
"/submitSignedApproval",
"/submitOwnerApproval",
}:
return self._json(self.service.submit_owner_approval(body))
if parsed.path == "/startDeployment":
return self._json(
self.service.start_deployment(
str(body.get("candidate_id") or "")
)
)
self._json({"error": "not_found"}, HTTPStatus.NOT_FOUND)
except (ValueError, FileNotFoundError, CapabilityAuthorizationError) as exc:
self._json(
{"error": type(exc).__name__}, HTTPStatus.BAD_REQUEST
)
def git_main_probe(repository: Path) -> Callable[[], str]:
repository = Path(repository)
def probe() -> str:
completed = subprocess.run(
[
"git",
f"--git-dir={repository}",
"rev-parse",
"refs/heads/main",
],
check=True,
capture_output=True,
text=True,
timeout=15,
env={"PATH": os.environ.get("PATH", "")},
)
return completed.stdout.strip()
return probe
def serve_shadow(
*,
state_root: Path,
repository: Path,
host: str,
port: int,
poll_seconds: float,
) -> None:
service = SilentReleaseActionService.shadow(state_root)
controller = PersistentSilentReleaseController(
state_root,
service,
main_probe=git_main_probe(repository),
)
controller.run_once()
stop = threading.Event()
def controller_loop() -> None:
while not stop.wait(poll_seconds):
try:
controller.run_once()
except Exception as exc: # fail closed and keep the API available
_durable_json(
controller.status_path,
{
"error": type(exc).__name__,
"execution_enabled": False,
"shadow_mode": True,
"side_effects_executed": False,
"status": "FAIL_CLOSED",
},
)
handler = type(
"ConfiguredShadowActionHandler",
(ShadowActionHandler,),
{"service": service, "controller": controller},
)
server = ThreadingHTTPServer((host, port), handler)
api_thread = threading.Thread(
target=server.serve_forever,
name="silent-release-shadow-api",
daemon=False,
)
controller_thread = threading.Thread(
target=controller_loop,
name="silent-release-shadow-controller",
daemon=True,
)
api_thread.start()
controller_thread.start()
try:
while api_thread.is_alive():
api_thread.join(timeout=1)
finally:
stop.set()
server.shutdown()
server.server_close()
controller_thread.join(timeout=max(1.0, poll_seconds + 1.0))
def main() -> None:
parser = argparse.ArgumentParser()
parser.add_argument("--state-root", required=True, type=Path)
parser.add_argument("--repository", required=True, type=Path)
parser.add_argument("--host", default="127.0.0.1")
parser.add_argument("--port", default=18891, type=int)
parser.add_argument("--poll-seconds", default=30.0, type=float)
arguments = parser.parse_args()
if arguments.poll_seconds <= 0:
raise SystemExit("poll seconds must be positive")
if os.environ.get("AIOS_GITOPS_WRITES_ENABLED", "false").lower() != "false":
raise SystemExit("GitOps writes must remain disabled")
if (
os.environ.get(
"AIOS_SILENT_RELEASE_EXECUTION_ENABLED", "false"
).lower()
!= "false"
):
raise SystemExit("Silent Release execution must remain disabled")
serve_shadow(
state_root=arguments.state_root,
repository=arguments.repository,
host=arguments.host,
port=arguments.port,
poll_seconds=arguments.poll_seconds,
)
if __name__ == "__main__":
main()
File diff suppressed because it is too large Load Diff
-21
View File
@@ -1,21 +0,0 @@
"""Task Plane runtime package."""
from .models import (
EvidenceEnvelope,
ProbeEvidence,
ProbeOutcome,
TaskPlan,
TaskRecord,
TaskStatus,
TaskStateTransitionError,
)
__all__ = [
"EvidenceEnvelope",
"ProbeEvidence",
"ProbeOutcome",
"TaskPlan",
"TaskRecord",
"TaskStatus",
"TaskStateTransitionError",
]
@@ -1,17 +0,0 @@
"""Adapters for executing tasks via worker backends."""
from .base import WorkerAdapter
from .openclaw import OpenClawAdapter
from .codex import CodexAdapter
from .hermes import HermesAdapter
from .a2a import A2AAdapter
from .systemd_worker import SystemdWorkerAdapter
__all__ = [
"WorkerAdapter",
"OpenClawAdapter",
"CodexAdapter",
"HermesAdapter",
"A2AAdapter",
"SystemdWorkerAdapter",
]
@@ -1,31 +0,0 @@
from __future__ import annotations
from .base import WorkerAdapter
class A2AAdapter(WorkerAdapter):
worker_name = "a2a"
def capabilities(self) -> list[str]:
return ["task.exchange", "agent.handoff"]
def claim(self, task):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "claimed"}
def execute(self, task, claim):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "SHADOW_VERIFIED", "claim_id": claim.claim_id, "execution_mode": "shadow", "production_side_effects": False}
def heartbeat(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "alive"}
def report_progress(self, claim, progress):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "progress": progress}
def probe(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": claim.status}
def cancel(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "cancelled"}
def collect_result(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "result": "not_collected"}
@@ -1,41 +0,0 @@
"""Base worker adapter contract."""
from __future__ import annotations
import abc
class WorkerAdapter(abc.ABC):
worker_name = "base"
@abc.abstractmethod
def capabilities(self) -> list[str]:
raise NotImplementedError
@abc.abstractmethod
def claim(self, task):
raise NotImplementedError
@abc.abstractmethod
def execute(self, task, claim):
raise NotImplementedError
@abc.abstractmethod
def heartbeat(self, claim):
raise NotImplementedError
@abc.abstractmethod
def report_progress(self, claim, progress):
raise NotImplementedError
@abc.abstractmethod
def probe(self, claim):
raise NotImplementedError
@abc.abstractmethod
def cancel(self, claim):
raise NotImplementedError
@abc.abstractmethod
def collect_result(self, claim):
raise NotImplementedError
@@ -1,31 +0,0 @@
from __future__ import annotations
from .base import WorkerAdapter
class CodexAdapter(WorkerAdapter):
worker_name = "codex"
def capabilities(self) -> list[str]:
return ["code.read", "code.write.patch", "test.python", "review.security"]
def claim(self, task):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "claimed"}
def execute(self, task, claim):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "SHADOW_VERIFIED", "claim_id": claim.claim_id, "execution_mode": "shadow", "production_side_effects": False}
def heartbeat(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "alive"}
def report_progress(self, claim, progress):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "progress": progress}
def probe(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": claim.status}
def cancel(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "cancelled"}
def collect_result(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "result": "not_collected"}
@@ -1,31 +0,0 @@
from __future__ import annotations
from .base import WorkerAdapter
class HermesAdapter(WorkerAdapter):
worker_name = "hermes"
def capabilities(self) -> list[str]:
return ["architecture.review", "policy.review", "threat.model"]
def claim(self, task):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "claimed"}
def execute(self, task, claim):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "SHADOW_VERIFIED", "claim_id": claim.claim_id, "execution_mode": "shadow", "production_side_effects": False}
def heartbeat(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "alive"}
def report_progress(self, claim, progress):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "progress": progress}
def probe(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": claim.status}
def cancel(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "cancelled"}
def collect_result(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "result": "not_collected"}
@@ -1,31 +0,0 @@
from __future__ import annotations
from .base import WorkerAdapter
class OpenClawAdapter(WorkerAdapter):
worker_name = "openclaw"
def capabilities(self) -> list[str]:
return ["code.read", "code.write.patch", "test.execute", "runtime.shadow"]
def claim(self, task):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "claimed"}
def execute(self, task, claim):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "SHADOW_VERIFIED", "claim_id": claim.claim_id, "execution_mode": "shadow", "production_side_effects": False}
def heartbeat(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "alive"}
def report_progress(self, claim, progress):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "progress": progress}
def probe(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": claim.status}
def cancel(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "cancelled"}
def collect_result(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "result": "not_collected"}
@@ -1,31 +0,0 @@
from __future__ import annotations
from .base import WorkerAdapter
class SystemdWorkerAdapter(WorkerAdapter):
worker_name = "systemd_worker"
def capabilities(self) -> list[str]:
return ["systemd.exec", "runtime.shell"]
def claim(self, task):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "claimed"}
def execute(self, task, claim):
return {"worker": self.worker_name, "task_id": task.task_id, "status": "SHADOW_VERIFIED", "claim_id": claim.claim_id, "execution_mode": "shadow", "production_side_effects": False}
def heartbeat(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "alive"}
def report_progress(self, claim, progress):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "progress": progress}
def probe(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": claim.status}
def cancel(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "status": "cancelled"}
def collect_result(self, claim):
return {"worker": self.worker_name, "claim_id": claim.claim_id, "result": "not_collected"}
@@ -1,27 +0,0 @@
"""Agent registry abstraction."""
from __future__ import annotations
from .errors import NotFoundError
class AgentRegistry:
def __init__(self):
self._agents: dict[str, dict] = {}
def register(self, agent_id: str, metadata: dict) -> None:
self._agents[agent_id] = {"agent_id": agent_id, **metadata}
def unregister(self, agent_id: str) -> None:
if agent_id not in self._agents:
raise NotFoundError("agent")
del self._agents[agent_id]
def list(self) -> list[dict]:
return list(self._agents.values())
def get(self, agent_id: str) -> dict:
try:
return self._agents[agent_id]
except KeyError:
raise NotFoundError("agent")
-557
View File
@@ -1,557 +0,0 @@
"""Task Plane HTTP API implementation (stdlib-only)."""
from __future__ import annotations
import json
import os
import time
import uuid
import urllib.parse
from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
from typing import Any, Callable
from .delegation import DelegationEngine
from .errors import (
ConflictError,
ForbiddenError,
NotFoundError,
PayloadTooLargeError,
TaskPlaneHTTPError,
UnauthorizedError,
ValidationError,
)
from .evidence import EvidenceService
from .journal import TaskJournal
from .lease import LeaseManager
from .models import (
EvidenceEnvelope,
ResultEnvelope,
TaskPlan,
TaskRecord,
iso8601_now,
contains_secret_text,
validate_identifier,
)
from .policy import PolicyEngine
from .store import TaskPlaneStore
from .worker_registry import WorkerRegistry
def _json_payload(data: dict[str, Any], status: int = 200) -> tuple[int, bytes]:
body = json.dumps(data, ensure_ascii=False).encode("utf-8")
return status, body
def _bounded_int(value: Any, name: str, minimum: int, maximum: int) -> int:
try:
number = int(value)
except (TypeError, ValueError) as exc:
raise ValidationError(f"invalid {name}") from exc
if not minimum <= number <= maximum:
raise ValidationError(f"invalid {name}")
return number
class TaskPlaneService:
def __init__(self, store: TaskPlaneStore, policy: PolicyEngine | None = None):
self.store = store
self.policy = policy or PolicyEngine(
execution_enabled=False,
identity_capabilities={
"anon": {"task.creator", "task.read", "task.audit"}
},
)
self.journal = TaskJournal(store)
self.lease = LeaseManager(store)
self.workers = WorkerRegistry(store)
self.evidence = EvidenceService(store)
self.delegation = DelegationEngine(store)
self.store.probe_attestation_verifier = self.policy.probe_attestation_verifier
def create_task(self, body: dict[str, Any], idempotency_key: str | None = None) -> dict[str, Any]:
if body.get("parent_task_id"):
raise ForbiddenError("child tasks must use the delegate endpoint")
issuer = body.get("issuer", {})
actor = issuer.get("id", "anon") if isinstance(issuer, dict) else "anon"
self.policy.assert_allowed("create", actor)
try:
plan = TaskPlan.from_dict(body)
except (TypeError, ValueError) as exc:
raise ValidationError("invalid task plan") from exc
task = self.store.create_task_with_event(plan, idempotency_key=idempotency_key)
return task.to_dict()
def list_tasks(self, actor: str | None = None) -> list[dict[str, Any]]:
if actor is not None:
self.policy.assert_allowed("read", actor)
return [task.to_dict(include_task=False) for task in self.store.list_tasks()]
def get_task(self, task_id: str, actor: str | None = None) -> dict[str, Any]:
if actor is not None:
self.policy.assert_allowed("read", actor, task=self.store.get_task(task_id))
return self.store.get_task(task_id).to_dict()
def get_worker(self, worker_id: str, actor: str | None = None) -> dict[str, Any]:
if actor is not None:
self.policy.assert_allowed("read", actor)
return self.workers.get(worker_id).to_dict()
def list_workers(self, actor: str | None = None) -> list[dict[str, Any]]:
if actor is not None:
self.policy.assert_allowed("read", actor)
return [w.to_dict() for w in self.workers.list()]
def register_worker(self, body: dict[str, Any]) -> dict[str, Any]:
actor = str(body.get("actor_identity") or body.get("identity") or "")
is_admin = actor in self.policy.admin_identities
try:
existing = self.store.get_worker(str(body.get("worker_id") or ""))
except NotFoundError:
existing = None
if existing is not None and str(body.get("identity") or "") != existing.identity:
raise ConflictError("worker identity takeover rejected")
self.policy.assert_readonly_worker_registration(actor, body)
self.policy.assert_allowed(
"worker.admin" if is_admin else "worker.register", actor
)
try:
worker = self.workers.register(
body,
actor_identity=actor,
is_admin=is_admin,
expected_revision=body.get("expected_revision"),
)
except (TypeError, ValueError, KeyError) as exc:
raise ValidationError("invalid worker registration") from exc
return worker.to_dict()
def worker_heartbeat(self, worker_id: str, payload: dict[str, Any]) -> dict[str, Any]:
interval = payload.get("heartbeat_interval_seconds")
actor = str(payload.get("actor_identity") or "")
self.policy.assert_readonly_worker_binding(actor, worker_id)
self.policy.assert_allowed("worker.register", actor)
return self.workers.heartbeat(
worker_id,
heartbeat_interval_seconds=interval,
actor_identity=actor,
expected_revision=payload.get("expected_revision"),
).to_dict()
def disable_worker(self, worker_id: str, payload: dict[str, Any]) -> dict[str, Any]:
actor = str(payload.get("actor_identity") or "")
self.policy.assert_allowed("worker.admin", actor)
return self.store.disable_worker(
worker_id,
actor_identity=actor,
expected_revision=int(payload.get("expected_revision")),
reason=str(payload.get("reason") or "disabled by administrator"),
).to_dict()
def restore_worker(self, worker_id: str, payload: dict[str, Any]) -> dict[str, Any]:
actor = str(payload.get("actor_identity") or "")
self.policy.assert_allowed("worker.admin", actor)
return self.store.restore_worker(
worker_id,
actor_identity=actor,
expected_revision=int(payload.get("expected_revision")),
).to_dict()
def claim(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
actor = body.get("actor_identity") or body.get("worker_identity") or "worker"
self.policy.assert_allowed("claim", actor)
worker_id = body.get("worker_id")
if not worker_id:
raise ValidationError("worker_id required")
task = self.store.get_task(task_id)
self.policy.assert_readonly_worker_binding(actor, str(worker_id))
self.policy.assert_allowed("claim", actor, task=task)
lease_seconds = _bounded_int(
body.get("lease_seconds", 30), "lease_seconds", 1, 3600
)
claim = self.lease.claim(
task_id=task_id,
worker_id=worker_id,
worker_identity=actor,
lease_seconds=lease_seconds,
expected_revision=task.revision,
operation_digest=task.task_plan.operation_digest(),
)
return {"claim_id": claim.claim_id, **claim.to_dict()}
def heartbeat(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
claim_id = body.get("claim_id")
if not claim_id:
raise ValidationError("claim_id required")
actor = body.get("actor_identity") or body.get("worker_identity") or "worker"
task = self.store.get_task(task_id)
claim = self.store.get_claim(claim_id)
self.policy.assert_readonly_worker_binding(actor, claim.worker_id)
self.policy.assert_allowed("heartbeat", actor, task=task)
if claim.worker_identity != actor:
raise ForbiddenError("claim identity mismatch")
lease_seconds = _bounded_int(
body.get("lease_seconds", 30), "lease_seconds", 1, 3600
)
claim = self.lease.heartbeat(task_id, claim_id, lease_seconds)
return {"claim_id": claim_id, "lease_until": claim.lease_until}
def progress(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
claim_id = body.get("claim_id")
if not claim_id:
raise ValidationError("claim_id required")
actor = body.get("actor_identity") or body.get("worker_identity") or "worker"
progress = body.get("progress")
stage = body.get("stage", "RUNNING")
message = body.get("message")
if progress is None:
raise ValidationError("progress required")
task = self.store.get_task(task_id)
claim = self.store.get_claim(claim_id)
self.policy.assert_readonly_worker_binding(actor, claim.worker_id)
self.policy.assert_allowed("progress", actor, task=task)
if task.task_plan.data_classification == "credential-secret":
if message not in {None, ""} or contains_secret_text(stage):
raise ValidationError(
"credential-secret progress must contain metadata only"
)
lease_seconds = _bounded_int(
body.get("lease_seconds", 30), "lease_seconds", 1, 3600
)
claim = self.store.record_progress(
task_id,
claim_id,
actor,
progress=progress,
stage=str(stage),
message=None if message is None else str(message),
lease_seconds=lease_seconds,
)
return {"claim_id": claim.claim_id, "status": claim.status, "progress": progress}
def complete(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
claim_id = body.get("claim_id")
if not claim_id:
raise ValidationError("claim_id required")
actor = body.get("actor_identity") or body.get("worker_identity") or "worker"
task = self.store.get_task(task_id)
status = body.get("status", "COMPLETED")
outputs = body.get("outputs", [])
evidences = body.get("evidence", [])
if not isinstance(outputs, list):
raise ValidationError("outputs must be list")
if not isinstance(evidences, list):
raise ValidationError("evidence must be list")
if any(not isinstance(ev, dict) for ev in evidences):
raise ValidationError("evidence items must be objects")
claim = self.store.get_claim(claim_id)
self.policy.assert_readonly_worker_binding(actor, claim.worker_id)
self.policy.assert_allowed("result", actor, task=task)
if claim.worker_identity != actor:
raise ForbiddenError("claim identity mismatch")
if task.task_plan.data_classification == "credential-secret":
allowed_result_keys = {
"kind",
"hash",
"sha256",
"count",
"status",
"error_code",
"signature",
"attestation_ref",
}
for entry in [*outputs, *evidences]:
if (
not isinstance(entry, dict)
or set(entry) - allowed_result_keys
or contains_secret_text(entry)
):
raise ValidationError(
"credential-secret result must contain hash/count/status only"
)
result = ResultEnvelope(
task_id=task_id,
claim_id=claim_id,
worker_identity=actor,
status=status,
outputs=outputs,
evidence=evidences,
started_at=str(body.get("started_at") or iso8601_now()),
completed_at=str(body.get("completed_at") or iso8601_now()),
result_digest=body.get("result_digest"),
signature=body.get("signature"),
attestation_ref=body.get("attestation_ref"),
operation_digest=claim.operation_digest,
)
evidence_decision = self.policy.evidence_verifier.verify(task, evidences)
if not evidence_decision.allowed:
raise ForbiddenError(evidence_decision.reason)
result_decision = self.policy.verify_result(result, task)
if not result_decision.allowed:
raise ForbiddenError(result_decision.reason)
result.result_digest = result.computed_digest()
previous = self.store._last_evidence_hash(task_id)
evidence_entries = []
for evidence in evidences:
envelope = EvidenceEnvelope.create(
task_id=task_id,
claim_id=claim_id,
kind=str(evidence.get("kind") or evidence.get("type") or "generic"),
worker_identity=actor,
payload=evidence,
previous_evidence_hash=previous,
signature=evidence.get("signature"),
attestation_ref=evidence.get("attestation_ref"),
)
evidence_entries.append(envelope)
previous = envelope.evidence_digest
self.store.complete_claim_with_evidence_and_event(
task_id,
claim_id,
status
if status
in {"COMPLETED", "FAILED_FINAL", "FAILED_RECOVERABLE", "CANCELLED"}
else "COMPLETED",
result.to_dict(),
evidence_entries,
)
return result.to_dict()
def fail(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
claim_id = body.get("claim_id")
if not claim_id:
raise ValidationError("claim_id required")
reason = body.get("reason", "failed")
actor = body.get("actor_identity") or body.get("worker_identity") or "worker"
self.policy.assert_allowed("fail", actor)
task = self.store.get_task(task_id)
self.policy.assert_allowed("fail", actor, task=task)
claim = self.store.get_claim(claim_id)
if claim.worker_identity != actor:
raise ForbiddenError("claim identity mismatch")
status = "FAILED_RECOVERABLE" if body.get("recoverable", False) else "FAILED_FINAL"
self.store.fail_claim_with_event(task_id, claim_id, status, str(reason))
return {"task_id": task_id, "status": status, "reason": reason}
def cancel(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
actor = body.get("actor_identity") or body.get("worker_identity") or "owner"
self.policy.assert_allowed("cancel", actor)
reason = body.get("reason", "cancelled")
task = self.store.get_task(task_id)
self.policy.assert_allowed("cancel", actor, task=task)
updated = self.store.cancel_task_with_event(
task_id, task.revision, reason=reason
)
return updated.to_dict()
def delegate(self, task_id: str, body: dict[str, Any]) -> dict[str, Any]:
actor = body.get("actor_identity") or "owner"
self.policy.assert_allowed("delegate", actor)
parent = self.store.get_task(task_id)
self.policy.assert_allowed("delegate", actor, task=parent)
if "task" in body:
try:
child_plan = TaskPlan.from_dict(body["task"])
except (TypeError, ValueError) as exc:
raise ValidationError("invalid child task plan") from exc
else:
plan_body = dict(body)
plan_body.pop("actor_identity", None)
try:
child_plan = TaskPlan.from_dict(plan_body)
except (TypeError, ValueError) as exc:
raise ValidationError("invalid child task plan") from exc
child_plan.parent_task_id = task_id
child = self.delegation.delegate(task_id, child_plan, caller=actor)
return child.to_dict()
def get_events(self, task_id: str, actor: str | None = None) -> list[dict[str, Any]]:
if actor is not None:
self.policy.assert_allowed("audit", actor, task=self.store.get_task(task_id))
self.store.get_task(task_id)
return self.store.list_events(task_id)
def get_evidence(self, task_id: str, actor: str | None = None) -> list[dict[str, Any]]:
if actor is not None:
self.policy.assert_allowed("audit", actor, task=self.store.get_task(task_id))
self.store.get_task(task_id)
return self.evidence.list(task_id)
def get_children(self, task_id: str, actor: str | None = None) -> list[str]:
if actor is not None:
self.policy.assert_allowed("read", actor, task=self.store.get_task(task_id))
self.store.get_task(task_id)
return self.store.list_children(task_id)
class TaskPlaneRequestHandler(BaseHTTPRequestHandler):
service: TaskPlaneService | None = None
authenticator: Callable[[str], str | None] | None = None
max_body_bytes = 1_000_000
def log_message(self, format: str, *args: Any) -> None:
# BaseHTTPRequestHandler logs the raw request target; suppress it so
# attacker-controlled paths or query strings cannot leak credentials.
return
def _json(self) -> tuple[str | None, dict[str, Any]]:
try:
length = int(self.headers.get("Content-Length", "0"))
except ValueError as exc:
raise ValidationError("invalid content length") from exc
if length < 0:
raise ValidationError("invalid content length")
if length > self.max_body_bytes:
raise PayloadTooLargeError("payload exceeds limit")
raw = self.rfile.read(length) if length > 0 else b""
if not raw:
return None, {}
try:
decoded = raw.decode("utf-8")
payload = json.loads(decoded)
if not isinstance(payload, dict):
raise ValidationError("JSON payload must be an object")
return decoded, payload
except (UnicodeDecodeError, json.JSONDecodeError) as exc:
raise ValidationError("invalid JSON payload") from exc
def _response(self, status: int, payload: Any) -> None:
encoded = json.dumps(payload, ensure_ascii=False).encode("utf-8")
self.send_response(status)
self.send_header("Content-Type", "application/json; charset=utf-8")
self.send_header("Content-Length", str(len(encoded)))
self.end_headers()
self.wfile.write(encoded)
def _authenticated_identity(self) -> str:
authenticator = type(self).authenticator
if authenticator is None:
raise UnauthorizedError("authentication unavailable")
header = self.headers.get("Authorization", "")
if not header.startswith("Bearer "):
raise UnauthorizedError("bearer authentication required")
token = header[7:]
if not token or any(ord(ch) < 33 for ch in token):
raise UnauthorizedError("invalid bearer token")
identity = authenticator(token)
if not identity:
raise UnauthorizedError("invalid credentials")
return str(identity)
def _handle(self, handler):
try:
if self.service is None:
raise TaskPlaneHTTPError("service not configured")
result = handler()
if isinstance(result, tuple):
status, payload = result
else:
status, payload = 200, result
self._response(status, payload)
except TaskPlaneHTTPError as exc:
self._response(exc.status_code, exc.to_payload())
except Exception as exc: # defensive
if isinstance(exc, TaskPlaneHTTPError):
raise
self._response(
500,
{"error": "internal_error", "message": "internal server error"},
)
def do_GET(self): # noqa: N802
def route():
path = urllib.parse.urlparse(self.path).path
parts = [p for p in path.split("/") if p]
for part in parts[1:]:
if part not in {"events", "evidence", "children", "heartbeat"}:
try:
validate_identifier(part, "resource identifier")
except ValueError as exc:
raise ValidationError("invalid resource identifier") from exc
identity = self._authenticated_identity()
if len(parts) == 1 and parts[0] == "tasks":
return 200, self.service.list_tasks(identity)
if len(parts) == 2 and parts[0] == "tasks":
task_id = parts[1]
return 200, self.service.get_task(task_id, identity)
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "events":
return 200, self.service.get_events(parts[1], identity)
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "evidence":
return 200, self.service.get_evidence(parts[1], identity)
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "children":
return 200, self.service.get_children(parts[1], identity)
if len(parts) == 1 and parts[0] == "workers":
return 200, self.service.list_workers(identity)
if len(parts) == 2 and parts[0] == "workers":
return 200, self.service.get_worker(parts[1], identity)
raise NotFoundError("unknown endpoint")
self._handle(route)
def do_POST(self): # noqa: N802
def route():
body_text, body = self._json()
identity = self._authenticated_identity()
path = urllib.parse.urlparse(self.path).path
parts = [p for p in path.split("/") if p]
headers = {k.lower(): v for k, v in self.headers.items()}
if len(parts) == 1 and parts[0] == "tasks":
issuer = body.get("issuer")
if issuer and (
not isinstance(issuer, dict) or issuer.get("id") != identity
):
raise ForbiddenError("issuer identity mismatch")
body["issuer"] = {"type": "authenticated", "id": identity}
return 201, self.service.create_task(body or {}, headers.get("idempotency-key"))
if body.get("actor_identity") not in {None, identity}:
raise ForbiddenError("actor identity mismatch")
body["actor_identity"] = identity
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "claim":
return 200, self.service.claim(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "heartbeat":
return 200, self.service.heartbeat(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "progress":
return 200, self.service.progress(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "result":
return 200, self.service.complete(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "fail":
return 200, self.service.fail(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "cancel":
return 200, self.service.cancel(parts[1], body or {})
if len(parts) == 3 and parts[0] == "tasks" and parts[2] == "delegate":
return 200, self.service.delegate(parts[1], body or {})
if len(parts) == 2 and parts[0] == "workers" and parts[1] == "register":
if (
body.get("identity") != identity
and identity not in self.service.policy.admin_identities
):
raise ForbiddenError("worker identity mismatch")
return 201, self.service.register_worker(body or {})
if len(parts) == 3 and parts[0] == "workers" and parts[2] == "heartbeat":
return 200, self.service.worker_heartbeat(parts[1], body or {})
if len(parts) == 3 and parts[0] == "workers" and parts[2] == "disable":
return 200, self.service.disable_worker(parts[1], body or {})
if len(parts) == 3 and parts[0] == "workers" and parts[2] == "restore":
return 200, self.service.restore_worker(parts[1], body or {})
raise NotFoundError("unknown endpoint")
self._handle(route)
def run_server(
host: str,
port: int,
service: TaskPlaneService,
*,
authenticator: Callable[[str], str | None] | None = None,
):
handler = TaskPlaneRequestHandler
handler.service = service
handler.authenticator = authenticator
server = ThreadingHTTPServer((host, port), handler)
server.serve_forever()
@@ -1,36 +0,0 @@
"""Capability matcher helpers."""
from __future__ import annotations
from fnmatch import fnmatch
def _normalize_capability(capability: str) -> str:
return capability.strip().lower()
def capability_matches(granted: str, required: str) -> bool:
granted = _normalize_capability(granted)
required = _normalize_capability(required)
if granted == required:
return True
# A granted wildcard such as "test.*" should match required "test.unit"
if "*" in granted:
return fnmatch(required, granted)
# A required wildcard is treated as a suffix/prefix requirement for callers.
if "*" in required:
return fnmatch(granted, required)
return False
def has_capabilities(granted_caps: list[str], required_caps: list[str]) -> bool:
granted = [_normalize_capability(c) for c in granted_caps]
for required in required_caps:
required = _normalize_capability(required)
if not any(capability_matches(g, required) for g in granted):
return False
return True
def normalize_capabilities(caps: list[str]) -> list[str]:
return sorted({_normalize_capability(cap) for cap in caps})
@@ -1,87 +0,0 @@
"""Delegation guardrails."""
from __future__ import annotations
from .errors import ConflictError, ForbiddenError, NotFoundError
from .models import Budget, TaskPlan, TaskRecord
from .store import TaskPlaneStore
class DelegationEngine:
def __init__(self, store: TaskPlaneStore):
self.store = store
def can_delegate(self, parent_id: str, child_plan: TaskPlan) -> bool:
parent = self.store.get_task(parent_id)
if parent.status == "COMPLETED":
raise ConflictError("cannot delegate completed task")
if parent.task_plan.max_delegation_depth <= 0:
raise ConflictError("max delegation depth not configured")
depth = self._depth(parent_id)
if depth >= parent.task_plan.max_delegation_depth:
raise ConflictError("max delegation depth reached")
if child_plan.parent_task_id not in {None, parent_id}:
raise ConflictError("delegation parent mismatch")
current = parent_id
while current:
if child_plan.task_id and current == child_plan.task_id:
raise ConflictError("delegation cycle")
current = self.store.parent_of(current)
return True
def delegate(self, parent_id: str, child_plan: TaskPlan, *, caller: str) -> TaskRecord:
self.can_delegate(parent_id, child_plan)
parent = self.store.get_task(parent_id)
if child_plan.required_capabilities and not set(child_plan.required_capabilities).issubset(
set(parent.task_plan.required_capabilities or [])
):
raise ForbiddenError("child capabilities exceed parent constraints")
risk_rank = {"low": 1, "medium": 2, "high": 3}
if risk_rank.get(child_plan.risk, 99) > risk_rank.get(parent.task_plan.risk, -1):
raise ForbiddenError("child risk exceeds parent constraints")
child_plan.parent_task_id = parent_id
child_plan.correlation_id = parent.task_plan.correlation_id or parent.task_id
child_plan.preferred_workers = parent.task_plan.preferred_workers
child_plan.max_delegation_depth = min(
child_plan.max_delegation_depth,
parent.task_plan.max_delegation_depth,
)
child_plan.approval_policy = parent.task_plan.approval_policy
parent_budget = self.store.get_task_budget(parent_id)
required = Budget()
for name in ("time_seconds", "attempts", "tokens", "cost"):
available = getattr(parent_budget, name)
requested = getattr(child_plan.budgets, name)
if name == "attempts" and requested is None:
requested = 1
if available is not None:
requested = 0 if requested is None else requested
if requested > available:
raise ConflictError(f"budget exceeded for {name}")
setattr(child_plan.budgets, name, requested)
setattr(required, name, requested)
# A retried identical delegation is a read, not a second budget charge.
if child_plan.task_id:
try:
existing = self.store.get_task(child_plan.task_id)
except NotFoundError:
existing = None
if existing is not None:
if (
existing.task_plan.parent_task_id == parent_id
and existing.task_plan.operation_digest()
== child_plan.operation_digest()
):
return existing
raise ConflictError("delegation task_id conflict")
return self.store.delegate_task_with_budget_and_events(
parent_id, child_plan, caller=caller
)
def _depth(self, task_id: str) -> int:
return self.store._compute_depth(task_id)
-71
View File
@@ -1,71 +0,0 @@
"""Error types for task plane operations."""
from __future__ import annotations
class TaskPlaneError(Exception):
"""Base exception for API-compatible failures."""
status_code = 500
error_code = "task_plane_error"
def __init__(self, message: str, *, code: str | None = None, details: dict | None = None):
super().__init__(message)
self.message = message
self.details = details or {}
if code:
self.error_code = code
def to_payload(self) -> dict:
return {"error": self.error_code, "message": self.message, "details": self.details}
class TaskPlaneHTTPError(TaskPlaneError):
"""Base class with explicit HTTP status."""
status_code = 500
class ValidationError(TaskPlaneHTTPError):
status_code = 400
error_code = "validation_error"
class UnauthorizedError(TaskPlaneHTTPError):
status_code = 401
error_code = "unauthorized"
class ForbiddenError(TaskPlaneHTTPError):
status_code = 403
error_code = "forbidden"
class NotFoundError(TaskPlaneHTTPError):
status_code = 404
error_code = "not_found"
class ConflictError(TaskPlaneHTTPError):
status_code = 409
error_code = "conflict"
class PayloadTooLargeError(TaskPlaneHTTPError):
status_code = 413
error_code = "payload_too_large"
class FailedDependencyError(TaskPlaneHTTPError):
status_code = 424
error_code = "failed_dependency"
class LockedError(TaskPlaneHTTPError):
status_code = 423
error_code = "locked"
class ServiceUnavailableError(TaskPlaneHTTPError):
status_code = 503
error_code = "service_unavailable"
-108
View File
@@ -1,108 +0,0 @@
"""CloudEvents-compatible event envelope and durable outbox helpers."""
from __future__ import annotations
from dataclasses import dataclass
from datetime import datetime, timezone
from typing import Callable, Protocol
from .store import TaskPlaneStore
@dataclass
class CloudEvent:
specversion: str = "1.0"
id: str = ""
type: str = ""
source: str = "spiffe://aios/task-plane"
subject: str = ""
datacontenttype: str = "application/json"
data: dict | None = None
@property
def time(self) -> str:
return datetime.now(timezone.utc).replace(microsecond=0).isoformat().replace("+00:00", "Z")
def asdict(self) -> dict:
return {
"specversion": self.specversion,
"id": self.id,
"type": self.type,
"source": self.source,
"subject": self.subject,
"time": self.time,
"datacontenttype": self.datacontenttype,
"data": self.data or {},
}
NOT_DELIVERED = "NOT_DELIVERED"
DELIVERED = "DELIVERED"
FAILED_DELIVERY = "FAILED"
class NotificationAdapter(Protocol):
name: str
def deliver(self, envelope: dict) -> str | None:
"""Return a provider receipt, or None when delivery is unconfirmed."""
class CallbackNotificationAdapter:
"""Adapter seam for webhook, NATS, GPT, dashboard, and WeChat transports."""
def __init__(self, name: str, callback: Callable[[dict], str | None]):
self.name = name
self._callback = callback
def deliver(self, envelope: dict) -> str | None:
return self._callback(envelope)
class WebhookAdapter(CallbackNotificationAdapter):
def __init__(self, callback: Callable[[dict], str | None]):
super().__init__("webhook", callback)
class NatsJetStreamAdapter(CallbackNotificationAdapter):
def __init__(self, callback: Callable[[dict], str | None]):
super().__init__("nats-jetstream", callback)
class GPTResultCallbackAdapter(CallbackNotificationAdapter):
def __init__(self, callback: Callable[[dict], str | None]):
super().__init__("gpt-result-callback", callback)
class DashboardAdapter(CallbackNotificationAdapter):
def __init__(self, callback: Callable[[dict], str | None]):
super().__init__("dashboard", callback)
class WeChatGatewayAdapter(CallbackNotificationAdapter):
def __init__(self, callback: Callable[[dict], str | None]):
super().__init__("wechat-gateway", callback)
class DurableOutboxDispatcher:
def __init__(self, store: TaskPlaneStore):
self.store = store
def dispatch(self, message_id: str, adapter: NotificationAdapter) -> dict:
row = self.store.get_outbox_message(message_id)
import json
envelope = json.loads(row["payload"])
try:
receipt = adapter.deliver(envelope)
except Exception:
return self.store.record_outbox_delivery(
message_id,
receipt=None,
error=f"{adapter.name} delivery failed",
)
return self.store.record_outbox_delivery(
message_id,
receipt=str(receipt) if receipt else None,
error=None if receipt else f"{adapter.name} receipt unavailable",
)
-31
View File
@@ -1,31 +0,0 @@
"""Evidence helpers."""
from __future__ import annotations
import hashlib
import json
from .store import TaskPlaneStore
def evidence_digest(payload: dict) -> str:
body = json.dumps(payload, sort_keys=True, separators=(",", ":"))
return hashlib.sha256(body.encode("utf-8")).hexdigest()
class EvidenceService:
def __init__(self, store: TaskPlaneStore):
self.store = store
def add(self, task_id: str, kind: str, payload: dict, worker_identity: str) -> str:
return self.store.add_evidence(task_id, kind, payload, worker_identity)
def list(self, task_id: str) -> list[dict]:
return self.store.list_evidence(task_id)
def verify(self, task_id: str, required_kinds: list[str]) -> bool:
evidences = self.list(task_id)
observed = {e["kind"] for e in evidences}
return set(required_kinds).issubset(
observed
) and self.store.verify_evidence_integrity(task_id)
-22
View File
@@ -1,22 +0,0 @@
"""Event journaling helpers."""
from __future__ import annotations
from .errors import ValidationError
from .store import TaskPlaneStore
class TaskJournal:
def __init__(self, store: TaskPlaneStore):
self.store = store
def append(self, task_id: str, event_type: str, data: dict, *, subject: str | None = None) -> None:
if not event_type:
raise ValidationError("event_type required")
self.store.append_event(task_id, event_type, data, subject=subject)
def list(self, task_id: str) -> list[dict]:
return self.store.list_events(task_id)
def check_chain(self, task_id: str) -> bool:
return self.store.verify_event_chain(task_id)
-32
View File
@@ -1,32 +0,0 @@
"""Claim and lease management helpers."""
from __future__ import annotations
from .errors import LockedError
from .models import ProbeEvidence
from .store import TaskPlaneStore
class LeaseManager:
def __init__(self, store: TaskPlaneStore):
self.store = store
def claim(self, task_id: str, worker_id: str, worker_identity: str, lease_seconds: int, expected_revision: int, operation_digest: str):
return self.store.claim_task(task_id, worker_id, worker_identity, lease_seconds, expected_revision, operation_digest)
def heartbeat(self, task_id: str, claim_id: str, lease_seconds: int):
claim = self.store.heartbeat_claim(task_id, claim_id, lease_seconds)
return claim
def probe(self, claim_id: str):
claim = self.store.get_claim(claim_id)
if claim.status != "ACTIVE":
raise LockedError("claim not active")
return claim
def recover(self, task_id: str, claim_id: str, probe: ProbeEvidence):
return self.store.recover_expired_claim(
task_id,
claim_id,
probe,
)
-829
View File
@@ -1,829 +0,0 @@
"""Core data models for the task plane."""
from __future__ import annotations
import hashlib
import json
import re
import time
import uuid
import urllib.parse
from enum import StrEnum
from dataclasses import dataclass, field
from datetime import datetime, timezone
from typing import Any
SENSITIVE_KEYS = {
"secret",
"password",
"token",
"api_key",
"access_token",
"refresh_token",
"authorization",
"private_key",
"client_secret",
"credential",
}
IDENTIFIER_RE = re.compile(r"[A-Za-z0-9][A-Za-z0-9._:-]{0,127}")
TYPE_RE = re.compile(r"[A-Za-z0-9][A-Za-z0-9._-]{0,127}")
RISK_LEVELS = {"low", "medium", "high"}
DATA_CLASSIFICATIONS = {
"public",
"internal",
"confidential",
"credential-secret",
}
SECRET_TEXT_PATTERNS = (
re.compile(r"(?i)\bAuthorization\s*:\s*Bearer\s+\S+"),
re.compile(r"(?i)\b(?:password|passwd|api[_-]?key|access[_-]?token|secret)\s*[:=]\s*\S+"),
re.compile(r"\bsk-[A-Za-z0-9_-]{12,}\b"),
re.compile(r"\bAKIA[0-9A-Z]{16}\b"),
re.compile(r"-----BEGIN (?:RSA |EC |OPENSSH )?PRIVATE KEY-----"),
)
TASK_PLAN_KEYS = {
"schema_version",
"task_id",
"task_type",
"issuer",
"intent",
"required_capabilities",
"preferred_workers",
"risk",
"approval_policy",
"dependencies",
"max_delegation_depth",
"budgets",
"evidence_requirements",
"parent_task_id",
"correlation_id",
"metadata",
"data_classification",
"execution_zone",
"network_policy",
"logging_policy",
"allowed_input_paths",
"allowed_output_paths",
"output_policy",
}
def _now_ts() -> float:
return time.time()
def canonical_json(value: Any) -> str:
return json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":"))
def compute_sha256_hex(value: Any) -> str:
return hashlib.sha256(canonical_json(value).encode("utf-8")).hexdigest()
def contains_secret_text(value: Any) -> bool:
if isinstance(value, str):
return any(pattern.search(value) for pattern in SECRET_TEXT_PATTERNS)
if isinstance(value, dict):
return any(contains_secret_text(item) for item in value.values())
if isinstance(value, (list, tuple)):
return any(contains_secret_text(item) for item in value)
return False
def validate_relative_paths(values: Any, name: str) -> list[str]:
paths = _normalize_list(values)
for value in paths:
decoded = urllib.parse.unquote(value)
parts = value.replace("\\", "/").split("/")
if (
decoded != value
or "\\" in value
or value.startswith("/")
or re.match(r"^[A-Za-z]:", value)
or any(part in {"", ".", ".."} for part in parts)
):
raise ValueError(f"invalid {name}")
return paths
def validate_identifier(value: Any, name: str) -> str:
text = str(value or "")
if (
not IDENTIFIER_RE.fullmatch(text)
or urllib.parse.unquote(text) != text
or any(ord(ch) < 32 or ord(ch) == 127 for ch in text)
):
raise ValueError(f"invalid {name}")
return text
def validate_type(value: Any, name: str) -> str:
text = str(value or "")
if not TYPE_RE.fullmatch(text):
raise ValueError(f"invalid {name}")
return text
def redact_value(value: Any) -> Any:
if isinstance(value, dict):
redacted = {}
for key, inner in value.items():
lowered = str(key).lower()
if lowered in SENSITIVE_KEYS or lowered.endswith(
("_password", "_secret", "_access_token", "_refresh_token", "_private_key")
):
redacted[key] = "[REDACTED]"
else:
redacted[key] = redact_value(inner)
return redacted
if isinstance(value, list):
return [redact_value(v) for v in value]
if isinstance(value, tuple):
return tuple(redact_value(v) for v in value)
return value
class TaskStateTransitionError(ValueError):
pass
class TaskStatus(StrEnum):
CREATED = "CREATED"
QUEUED = "QUEUED"
CLAIMED = "CLAIMED"
RUNNING = "RUNNING"
WAITING_CHILD = "WAITING_CHILD"
VERIFYING = "VERIFYING"
COMPLETED = "COMPLETED"
FAILED_RECOVERABLE = "FAILED_RECOVERABLE"
FAILED_FINAL = "FAILED_FINAL"
CANCELLED = "CANCELLED"
MANUAL_REVIEW_REQUIRED = "MANUAL_REVIEW_REQUIRED"
TASK_STATES = {
status.value for status in TaskStatus
}
TASK_NEXT_STATES = {
"CREATED": {"QUEUED", "CLAIMED", "CANCELLED", "MANUAL_REVIEW_REQUIRED"},
"QUEUED": {"CLAIMED", "CANCELLED", "FAILED_RECOVERABLE", "FAILED_FINAL"},
"CLAIMED": {"RUNNING", "FAILED_RECOVERABLE", "FAILED_FINAL", "CANCELLED"},
"RUNNING": {"VERIFYING", "WAITING_CHILD", "COMPLETED", "FAILED_RECOVERABLE", "FAILED_FINAL", "CANCELLED"},
"WAITING_CHILD": {"RUNNING", "VERIFYING", "FAILED_RECOVERABLE", "FAILED_FINAL", "CANCELLED"},
"VERIFYING": {"COMPLETED", "FAILED_RECOVERABLE", "FAILED_FINAL", "CANCELLED"},
"COMPLETED": set(),
"FAILED_RECOVERABLE": {"CLAIMED", "RUNNING", "CANCELLED"},
"FAILED_FINAL": set(),
"CANCELLED": set(),
"MANUAL_REVIEW_REQUIRED": {"RUNNING", "CANCELLED"},
}
@dataclass
class Budget:
time_seconds: int | None = None
attempts: int | None = None
tokens: int | None = None
cost: float | None = None
@classmethod
def from_dict(cls, data: dict[str, Any] | None) -> "Budget":
if not data:
return cls()
if not isinstance(data, dict) or set(data) - {
"time_seconds",
"attempts",
"tokens",
"cost",
}:
raise ValueError("invalid budget")
for name in ("time_seconds", "attempts", "tokens", "cost"):
value = data.get(name)
if value is not None:
try:
if float(value) < 0:
raise ValueError("budget values must be non-negative")
except (TypeError, ValueError) as exc:
raise ValueError("invalid budget value") from exc
return cls(
time_seconds=_to_positive_int(data.get("time_seconds")),
attempts=_to_positive_int(data.get("attempts")),
tokens=_to_positive_int(data.get("tokens")),
cost=_to_float(data.get("cost")),
)
def to_dict(self) -> dict[str, Any]:
return {
"time_seconds": self.time_seconds,
"attempts": self.attempts,
"tokens": self.tokens,
"cost": self.cost,
}
def merge(self, other: "Budget") -> "Budget":
return Budget(
time_seconds=_first_non_none(self.time_seconds, other.time_seconds),
attempts=_first_non_none(self.attempts, other.attempts),
tokens=_first_non_none(self.tokens, other.tokens),
cost=_first_non_none(self.cost, other.cost),
)
@dataclass
class TaskPlan:
schema_version: int = 1
task_id: str | None = None
task_type: str = "code.change"
issuer: dict[str, Any] = field(default_factory=dict)
intent: str | None = None
required_capabilities: list[str] = field(default_factory=list)
preferred_workers: list[str] = field(default_factory=list)
risk: str = "medium"
approval_policy: str = "patch-only"
dependencies: list[str] = field(default_factory=list)
max_delegation_depth: int = 5
budgets: Budget = field(default_factory=Budget)
evidence_requirements: list[str] = field(default_factory=list)
parent_task_id: str | None = None
correlation_id: str | None = None
metadata: dict[str, Any] = field(default_factory=dict)
data_classification: str = "internal"
execution_zone: str = "isolated-development"
network_policy: str = "controlled"
logging_policy: str = "redacted"
allowed_input_paths: list[str] = field(default_factory=list)
allowed_output_paths: list[str] = field(default_factory=list)
output_policy: str = "standard"
@classmethod
def from_dict(cls, data: dict[str, Any] | None) -> "TaskPlan":
if not isinstance(data, dict):
raise ValueError("task plan payload must be an object")
unknown = set(data) - TASK_PLAN_KEYS
if unknown:
raise ValueError("unknown task plan fields")
if "task_id" not in data or data["task_id"] is None:
data = dict(data)
data["task_id"] = f"task-{uuid.uuid4()}"
risk = str(data.get("risk", "medium"))
if risk not in RISK_LEVELS:
raise ValueError("invalid risk")
issuer = data.get("issuer", {})
if not isinstance(issuer, dict):
raise ValueError("issuer must be an object")
schema_version = int(data.get("schema_version", 1))
if schema_version != 1:
raise ValueError("unsupported schema_version")
max_depth = _to_positive_int(data.get("max_delegation_depth"), default=5)
if max_depth is None or max_depth > 32:
raise ValueError("invalid max_delegation_depth")
classification = str(data.get("data_classification", "internal"))
if classification not in DATA_CLASSIFICATIONS:
raise ValueError("invalid data_classification")
execution_zone = validate_type(
data.get("execution_zone", "isolated-development"), "execution_zone"
)
network_policy = validate_type(
data.get("network_policy", "controlled"), "network_policy"
)
logging_policy = validate_type(
data.get("logging_policy", "redacted"), "logging_policy"
)
output_policy = validate_type(
data.get("output_policy", "standard"), "output_policy"
)
if contains_secret_text(
{
"intent": data.get("intent"),
"metadata": data.get("metadata", {}),
}
):
raise ValueError("task content contains secret material")
if classification == "credential-secret":
if execution_zone != "secure-file-transformer":
raise ValueError("credential-secret requires secure-file-transformer")
if network_policy != "denied":
raise ValueError("credential-secret requires network denied")
if logging_policy != "metadata-only":
raise ValueError("credential-secret requires metadata-only logging")
if output_policy != "hash-count-status-only":
raise ValueError("credential-secret requires restricted output")
return cls(
schema_version=schema_version,
task_id=validate_identifier(data["task_id"], "task_id"),
task_type=validate_type(data.get("task_type", "code.change"), "task_type"),
issuer=issuer,
intent=data.get("intent"),
required_capabilities=_normalize_list(data.get("required_capabilities", [])),
preferred_workers=_normalize_list(data.get("preferred_workers", [])),
risk=risk,
approval_policy=validate_type(
data.get("approval_policy", "patch-only"), "approval_policy"
),
dependencies=_normalize_list(data.get("dependencies", [])),
max_delegation_depth=max_depth,
budgets=Budget.from_dict(data.get("budgets")),
evidence_requirements=_normalize_list(data.get("evidence_requirements", [])),
parent_task_id=data.get("parent_task_id"),
correlation_id=data.get("correlation_id"),
metadata=data.get("metadata", {}) if isinstance(data.get("metadata", {}), dict) else {},
data_classification=classification,
execution_zone=execution_zone,
network_policy=network_policy,
logging_policy=logging_policy,
allowed_input_paths=validate_relative_paths(
data.get("allowed_input_paths", []), "allowed_input_paths"
),
allowed_output_paths=validate_relative_paths(
data.get("allowed_output_paths", []), "allowed_output_paths"
),
output_policy=output_policy,
)
def to_dict(self, *, redact: bool = False) -> dict[str, Any]:
payload = {
"schema_version": self.schema_version,
"task_id": self.task_id,
"task_type": self.task_type,
"issuer": self.issuer,
"intent": self.intent,
"required_capabilities": list(self.required_capabilities),
"preferred_workers": list(self.preferred_workers),
"risk": self.risk,
"approval_policy": self.approval_policy,
"dependencies": list(self.dependencies),
"max_delegation_depth": self.max_delegation_depth,
"budgets": self.budgets.to_dict(),
"evidence_requirements": list(self.evidence_requirements),
"parent_task_id": self.parent_task_id,
"correlation_id": self.correlation_id,
"metadata": dict(self.metadata),
"data_classification": self.data_classification,
"execution_zone": self.execution_zone,
"network_policy": self.network_policy,
"logging_policy": self.logging_policy,
"allowed_input_paths": list(self.allowed_input_paths),
"allowed_output_paths": list(self.allowed_output_paths),
"output_policy": self.output_policy,
}
if redact:
payload = redact_value(payload)
return payload
def operation_digest(self) -> str:
return compute_sha256_hex(self.to_dict())
@dataclass
class TaskRecord:
task_id: str
status: str
revision: int
task_plan: TaskPlan
created_at: float
updated_at: float
root_task_id: str
idempotency_key: str | None = None
failure_reason: str | None = None
def with_status(self, status: str) -> "TaskRecord":
allowed = TASK_NEXT_STATES.get(self.status, set())
if status != self.status and (not allowed and self.status not in {"COMPLETED", "FAILED_FINAL", "CANCELLED"}):
raise TaskStateTransitionError(f"Cannot transition {self.status} -> {status}")
if status not in TASK_STATES:
raise TaskStateTransitionError(f"Invalid status {status}")
if status != self.status and self.status in TASK_NEXT_STATES and status not in TASK_NEXT_STATES[self.status]:
raise TaskStateTransitionError(f"Cannot transition {self.status} -> {status}")
return TaskRecord(
task_id=self.task_id,
status=status,
revision=self.revision + 1,
task_plan=self.task_plan,
created_at=self.created_at,
updated_at=_now_ts(),
root_task_id=self.root_task_id,
idempotency_key=self.idempotency_key,
failure_reason=self.failure_reason,
)
def to_dict(self, *, include_task=True, redact_payload: bool = True) -> dict[str, Any]:
payload = {
"task_id": self.task_id,
"status": self.status,
"revision": self.revision,
"root_task_id": self.root_task_id,
"parent_task_id": self.task_plan.parent_task_id,
"created_at": _fmt_ts(self.created_at),
"updated_at": _fmt_ts(self.updated_at),
"failure_reason": self.failure_reason,
}
if include_task:
payload["plan"] = self.task_plan.to_dict(redact=redact_payload)
if self.idempotency_key:
payload["idempotency_key"] = self.idempotency_key
return payload
@dataclass
class WorkerRegistration:
worker_id: str
worker_type: str
identity: str
capabilities: list[str]
trust_zone: str
max_concurrency: int
heartbeat_interval_seconds: int
supported_protocols: list[str]
registered_at: float
last_heartbeat: float
status: str = "active"
disabled: bool = False
owner_identity: str = ""
registration_revision: int = 1
registration_policy: str = "self"
created_at: float = 0.0
updated_at: float = 0.0
disabled_reason: str | None = None
@classmethod
def from_dict(cls, data: dict[str, Any]) -> "WorkerRegistration":
if not isinstance(data, dict):
raise ValueError("worker registration payload must be an object")
now = _now_ts()
worker_id = validate_identifier(data.get("worker_id"), "worker_id")
worker_type = validate_type(data.get("worker_type"), "worker_type")
identity = str(data.get("identity") or "").strip()
if not identity or any(ord(ch) < 32 for ch in identity):
raise ValueError("invalid identity")
max_concurrency = _to_positive_int(data.get("max_concurrency"), default=1)
heartbeat_interval = _to_positive_int(
data.get("heartbeat_interval_seconds"), default=15
)
if not max_concurrency:
raise ValueError("max_concurrency must be positive")
if not heartbeat_interval:
raise ValueError("heartbeat_interval_seconds must be positive")
return cls(
worker_id=worker_id,
worker_type=worker_type,
identity=identity,
capabilities=_normalize_list(data.get("capabilities", [])),
trust_zone=str(data.get("trust_zone") or "default"),
max_concurrency=max_concurrency,
heartbeat_interval_seconds=heartbeat_interval,
supported_protocols=_normalize_list(
data.get("supported_protocols", ["aios-task-v1"])
),
registered_at=float(data.get("registered_at") or now),
last_heartbeat=float(data.get("last_heartbeat") or now),
status=str(data.get("status") or "active"),
disabled=bool(data.get("disabled", False)),
owner_identity=str(data.get("owner_identity") or identity),
registration_revision=int(data.get("registration_revision") or 1),
registration_policy=str(data.get("registration_policy") or "self"),
created_at=float(data.get("created_at") or data.get("registered_at") or now),
updated_at=float(data.get("updated_at") or now),
disabled_reason=data.get("disabled_reason"),
)
def to_dict(self, redact_payload: bool = True) -> dict[str, Any]:
payload = {
"worker_id": self.worker_id,
"worker_type": self.worker_type,
"identity": self.identity,
"capabilities": list(self.capabilities),
"trust_zone": self.trust_zone,
"max_concurrency": self.max_concurrency,
"heartbeat_interval_seconds": self.heartbeat_interval_seconds,
"supported_protocols": list(self.supported_protocols),
"registered_at": _fmt_ts(self.registered_at),
"last_heartbeat": _fmt_ts(self.last_heartbeat),
"status": self.status,
"disabled": self.disabled,
"owner_identity": self.owner_identity,
"registration_revision": self.registration_revision,
"registration_policy": self.registration_policy,
"created_at": _fmt_ts(self.created_at or self.registered_at),
"updated_at": _fmt_ts(self.updated_at or self.last_heartbeat),
"disabled_reason": self.disabled_reason,
}
if redact_payload:
payload = redact_value(payload)
return payload
@dataclass(frozen=True)
class EvidenceEnvelope:
schema_version: int
evidence_id: str
task_id: str
claim_id: str
kind: str
worker_identity: str
payload: dict[str, Any]
created_at: float
previous_evidence_hash: str
evidence_digest: str
signature: str | None = None
attestation_ref: str | None = None
@classmethod
def create(
cls,
*,
task_id: str,
claim_id: str,
kind: str,
worker_identity: str,
payload: dict[str, Any],
previous_evidence_hash: str,
signature: str | None = None,
attestation_ref: str | None = None,
evidence_id: str | None = None,
created_at: float | None = None,
) -> "EvidenceEnvelope":
envelope = cls(
schema_version=1,
evidence_id=evidence_id or f"evd-{uuid.uuid4()}",
task_id=validate_identifier(task_id, "task_id"),
claim_id=validate_identifier(claim_id, "claim_id"),
kind=validate_type(kind, "evidence kind"),
worker_identity=str(worker_identity),
payload=redact_value(payload),
created_at=float(created_at or _now_ts()),
previous_evidence_hash=str(previous_evidence_hash or ""),
evidence_digest="",
signature=signature,
attestation_ref=attestation_ref,
)
return dataclass_replace(
envelope, evidence_digest=compute_sha256_hex(envelope.digest_payload())
)
def digest_payload(self) -> dict[str, Any]:
return {
"schema_version": self.schema_version,
"evidence_id": self.evidence_id,
"task_id": self.task_id,
"claim_id": self.claim_id,
"kind": self.kind,
"worker_identity": self.worker_identity,
"payload": self.payload,
"created_at": self.created_at,
"previous_evidence_hash": self.previous_evidence_hash,
"attestation_ref": self.attestation_ref,
}
def verify(
self,
*,
expected_task_id: str,
expected_claim_id: str,
expected_previous_hash: str,
attestation_verifier: Any = None,
) -> bool:
if (
self.schema_version != 1
or self.task_id != expected_task_id
or self.claim_id != expected_claim_id
or self.previous_evidence_hash != expected_previous_hash
or self.evidence_digest != compute_sha256_hex(self.digest_payload())
):
return False
if self.signature or self.attestation_ref:
return True if attestation_verifier is None else bool(attestation_verifier(self))
return True
def to_dict(self) -> dict[str, Any]:
return {
**self.digest_payload(),
"evidence_digest": self.evidence_digest,
"signature": self.signature,
}
class ProbeOutcome(StrEnum):
NOT_APPLIED = "NOT_APPLIED"
APPLIED_VERIFIED = "APPLIED_VERIFIED"
INDETERMINATE = "INDETERMINATE"
CONFLICT = "CONFLICT"
@dataclass(frozen=True)
class ProbeEvidence:
schema_version: int
probe_id: str
task_id: str
claim_id: str
worker_id: str
worker_identity: str
operation_digest: str
outcome: ProbeOutcome
observed_at: float
expires_at: float
probe_executor_identity: str
evidence_digest: str
signature: str | None = None
attestation_ref: str | None = None
@classmethod
def create(cls, **data: Any) -> "ProbeEvidence":
outcome = data["outcome"]
if not isinstance(outcome, ProbeOutcome):
outcome = ProbeOutcome(str(outcome))
probe = cls(
schema_version=1,
probe_id=data.get("probe_id") or f"probe-{uuid.uuid4()}",
task_id=validate_identifier(data["task_id"], "task_id"),
claim_id=validate_identifier(data["claim_id"], "claim_id"),
worker_id=validate_identifier(data["worker_id"], "worker_id"),
worker_identity=str(data["worker_identity"]),
operation_digest=str(data["operation_digest"]),
outcome=outcome,
observed_at=float(data["observed_at"]),
expires_at=float(data["expires_at"]),
probe_executor_identity=str(data["probe_executor_identity"]),
evidence_digest="",
signature=data.get("signature"),
attestation_ref=data.get("attestation_ref"),
)
return dataclass_replace(
probe, evidence_digest=compute_sha256_hex(probe.digest_payload())
)
def digest_payload(self) -> dict[str, Any]:
return {
"schema_version": self.schema_version,
"probe_id": self.probe_id,
"task_id": self.task_id,
"claim_id": self.claim_id,
"worker_id": self.worker_id,
"worker_identity": self.worker_identity,
"operation_digest": self.operation_digest,
"outcome": self.outcome.value,
"observed_at": self.observed_at,
"expires_at": self.expires_at,
"probe_executor_identity": self.probe_executor_identity,
"attestation_ref": self.attestation_ref,
}
def verify(self, *, now: float, attestation_verifier: Any) -> bool:
if (
self.schema_version != 1
or self.expires_at <= now
or self.observed_at > now
or self.evidence_digest != compute_sha256_hex(self.digest_payload())
or not (self.signature or self.attestation_ref)
):
return False
return bool(attestation_verifier and attestation_verifier(self))
def to_dict(self) -> dict[str, Any]:
return {
**self.digest_payload(),
"evidence_digest": self.evidence_digest,
"signature": self.signature,
}
def dataclass_replace(value: Any, **changes: Any) -> Any:
from dataclasses import replace
return replace(value, **changes)
@dataclass
class WorkerClaim:
claim_id: str
task_id: str
worker_id: str
worker_identity: str
claimed_at: float
lease_until: float
attempt: int
task_revision: int
operation_digest: str
status: str = "ACTIVE"
def to_dict(self, redact_payload: bool = True) -> dict[str, Any]:
payload = {
"claim_id": self.claim_id,
"task_id": self.task_id,
"worker_id": self.worker_id,
"worker_identity": self.worker_identity,
"claimed_at": _fmt_ts(self.claimed_at),
"lease_until": _fmt_ts(self.lease_until),
"attempt": self.attempt,
"task_revision": self.task_revision,
"operation_digest": self.operation_digest,
"status": self.status,
}
if redact_payload:
payload = redact_value(payload)
return payload
@dataclass
class ResultEnvelope:
task_id: str
claim_id: str
worker_identity: str
status: str
outputs: list[dict[str, Any]]
evidence: list[dict[str, Any]]
started_at: str
completed_at: str
result_digest: str | None = None
signature: str | None = None
attestation_ref: str | None = None
operation_digest: str | None = None
trust_status: str = "untrusted_shadow_result"
def digest_payload(self) -> dict[str, Any]:
return {
"task_id": self.task_id,
"claim_id": self.claim_id,
"worker_identity": self.worker_identity,
"status": self.status,
"outputs": self.outputs,
"evidence": self.evidence,
"started_at": self.started_at,
"completed_at": self.completed_at,
"operation_digest": self.operation_digest,
"attestation_ref": self.attestation_ref,
}
def computed_digest(self) -> str:
return compute_sha256_hex(self.digest_payload())
def to_dict(self, redact_payload: bool = True) -> dict[str, Any]:
payload = {
"task_id": self.task_id,
"claim_id": self.claim_id,
"worker_identity": self.worker_identity,
"status": self.status,
"outputs": self.outputs,
"evidence": self.evidence,
"started_at": self.started_at,
"completed_at": self.completed_at,
"result_digest": self.result_digest,
"signature": self.signature,
"attestation_ref": self.attestation_ref,
"operation_digest": self.operation_digest,
"trust_status": self.trust_status,
}
if redact_payload:
payload = redact_value(payload)
return payload
def _normalize_list(value: Any) -> list[str]:
if not isinstance(value, list):
raise ValueError("expected list")
if len(value) > 128 or any(
not isinstance(v, str)
or not v
or len(v) > 256
or any(ord(ch) < 32 for ch in v)
for v in value
):
raise ValueError("invalid list item")
return list(value)
def _to_positive_int(value: Any, default: int | None = None) -> int | None:
if value is None:
return default
try:
number = int(value)
except (TypeError, ValueError):
return default
return number if number >= 0 else default
def _to_float(value: Any) -> float | None:
if value is None:
return None
try:
return float(value)
except (TypeError, ValueError):
return None
def _first_non_none(primary: Any, fallback: Any) -> Any:
if primary is not None:
return primary
return fallback
def iso8601_now() -> str:
return datetime.now(timezone.utc).replace(microsecond=0).isoformat().replace("+00:00", "Z")
def _fmt_ts(value: float | None) -> str | None:
if value is None:
return None
return datetime.fromtimestamp(value, tz=timezone.utc).replace(microsecond=0).isoformat().replace("+00:00", "Z")
-226
View File
@@ -1,226 +0,0 @@
"""Policy and verification logic (fail-closed)."""
from __future__ import annotations
from dataclasses import dataclass
from typing import Callable
from .errors import ForbiddenError, LockedError
from .models import ProbeEvidence, ResultEnvelope, TaskRecord
RISK_RANK = {"low": 1, "medium": 2, "high": 3}
READONLY_WORKER_ACTIONS = frozenset(
{"claim", "result", "fail", "progress", "heartbeat"}
)
ALL_ACTION_CAPABILITIES = frozenset(
{
"task.creator",
"worker.self_register",
"worker.admin",
"task.claim",
"task.cancel",
"task.delegate",
"task.read",
"task.audit",
}
)
@dataclass
class PolicyDecision:
allowed: bool
reason: str = ""
@dataclass(frozen=True)
class ReadonlyWorkerProfile:
"""A narrowly-scoped identity allowed to process safe Shadow tasks only."""
worker_id: str
identity: str
capabilities: frozenset[str]
task_types: frozenset[str] = frozenset({"code.review", "test.execute"})
def permits_task(self, task: TaskRecord) -> bool:
plan = task.task_plan
if plan.task_type not in self.task_types or plan.risk != "low":
return False
if plan.data_classification not in {"public", "internal"}:
return False
if plan.metadata.get("production_side_effects") is not False:
return False
return set(plan.required_capabilities).issubset(self.capabilities)
class IdentityVerifier:
def __init__(self, trusted_identities: set[str] | None = None):
self.trusted_identities = trusted_identities
def verify(self, identity: str) -> bool:
if self.trusted_identities is None:
return True
return identity in self.trusted_identities
class EvidenceVerifier:
def __init__(self, mandatory: list[str] | None = None):
self.mandatory = set(mandatory or [])
def verify(self, task: TaskRecord, evidences: list[dict]) -> PolicyDecision:
required = set(task.task_plan.evidence_requirements or [])
if not required and not self.mandatory:
return PolicyDecision(True)
observed = {e.get("kind") for e in evidences}
missing = required.union(self.mandatory) - observed
if missing:
return PolicyDecision(False, f"missing evidence: {', '.join(sorted(missing))}")
return PolicyDecision(True)
class PolicyEngine:
"""
Fail-closed policy engine with minimal but explicit decisions.
"""
def __init__(
self,
*,
identity_verifier: IdentityVerifier | None = None,
evidence_verifier: EvidenceVerifier | None = None,
execution_enabled: bool = False,
max_risk: str = "medium",
result_signature_verifier: Callable[[ResultEnvelope], bool] | None = None,
probe_attestation_verifier: Callable[[ProbeEvidence], bool] | None = None,
identity_capabilities: dict[str, set[str]] | None = None,
admin_identities: set[str] | None = None,
readonly_worker_execution_enabled: bool = False,
readonly_worker_profiles: list[ReadonlyWorkerProfile] | None = None,
) -> None:
if max_risk not in RISK_RANK:
raise ValueError("invalid max_risk")
self.identity_verifier = identity_verifier or IdentityVerifier()
self.evidence_verifier = evidence_verifier or EvidenceVerifier()
self.execution_enabled = execution_enabled
self.max_risk = max_risk
self.result_signature_verifier = result_signature_verifier
self.probe_attestation_verifier = probe_attestation_verifier
self.identity_capabilities = {
identity: set(capabilities)
for identity, capabilities in (identity_capabilities or {}).items()
}
self.admin_identities = set(admin_identities or ())
self.readonly_worker_execution_enabled = readonly_worker_execution_enabled
self.readonly_worker_profiles = {
profile.identity: profile for profile in (readonly_worker_profiles or [])
}
def _readonly_worker_allows(
self, action: str, actor: str, task: TaskRecord | None
) -> bool:
profile = self.readonly_worker_profiles.get(actor)
return bool(
self.readonly_worker_execution_enabled
and action in READONLY_WORKER_ACTIONS
and profile is not None
and (task is None or profile.permits_task(task))
)
def assert_readonly_worker_binding(self, actor: str, worker_id: str) -> None:
profile = self.readonly_worker_profiles.get(actor)
if profile is not None and profile.worker_id != worker_id:
raise ForbiddenError("worker identity binding mismatch")
def assert_readonly_worker_registration(self, actor: str, payload: dict) -> None:
profile = self.readonly_worker_profiles.get(actor)
if profile is None:
return
if (
str(payload.get("worker_id") or "") != profile.worker_id
or str(payload.get("identity") or "") != profile.identity
or set(payload.get("capabilities") or []) != set(profile.capabilities)
or str(payload.get("trust_zone") or "") != "isolated-review"
or int(payload.get("max_concurrency") or 0) != 1
or int(payload.get("heartbeat_interval_seconds") or 0) != 15
):
raise ForbiddenError("readonly worker registration mismatch")
def verify_result(
self, result: ResultEnvelope, task: TaskRecord | None = None
) -> PolicyDecision:
computed = result.computed_digest()
if result.result_digest and result.result_digest != computed:
return PolicyDecision(False, "result digest mismatch")
sensitive = bool(
task
and (
task.task_plan.data_classification
in {"confidential", "credential-secret"}
or any(
capability.startswith(
("code.write", "deploy", "secret", "production")
)
for capability in task.task_plan.required_capabilities
)
)
)
if sensitive and not (result.signature or result.attestation_ref):
return PolicyDecision(False, "trusted result proof required")
if result.signature is not None or result.attestation_ref is not None:
if self.result_signature_verifier is None:
return PolicyDecision(False, "result signature verifier unavailable")
if not self.result_signature_verifier(result):
return PolicyDecision(False, "result signature invalid")
result.trust_status = "trusted_result"
else:
result.trust_status = "untrusted_shadow_result"
return PolicyDecision(True)
def authorize(self, action: str, actor: str, task: TaskRecord | None = None, context: dict | None = None) -> PolicyDecision:
context = context or {}
if not self.identity_verifier.verify(actor):
return PolicyDecision(False, "identity untrusted")
action_capabilities = {
"create": "task.creator",
"read": "task.read",
"audit": "task.audit",
"claim": "task.claim",
"result": "task.claim",
"fail": "task.claim",
"progress": "task.claim",
"heartbeat": "task.claim",
"cancel": "task.cancel",
"delegate": "task.delegate",
"worker.register": "worker.self_register",
"worker.admin": "worker.admin",
}
required_capability = action_capabilities.get(action)
if required_capability:
granted = set(
self.identity_capabilities.get(
actor, self.identity_capabilities.get("*", set())
)
)
if required_capability not in granted:
return PolicyDecision(False, f"action {action} not permitted")
if action in {"claim", "result", "fail", "cancel", "progress", "delegate", "heartbeat"} and not self.execution_enabled:
if not self._readonly_worker_allows(action, actor, task):
return PolicyDecision(False, "execution disabled")
if action == "delegate" and task and task.status not in {"COMPLETED", "FAILED_FINAL", "CANCELLED"}:
risk = str(context.get("risk", task.task_plan.risk))
if RISK_RANK.get(risk, 2) > RISK_RANK.get(self.max_risk, 1):
return PolicyDecision(False, f"risk {risk} exceeds policy")
return PolicyDecision(True)
def assert_allowed(self, action: str, actor: str, task: TaskRecord | None = None, context: dict | None = None):
decision = self.authorize(action, actor, task, context=context)
if not decision.allowed:
if decision.reason == "identity untrusted":
raise ForbiddenError(decision.reason)
if decision.reason == "execution disabled":
raise LockedError("execution disabled")
raise ForbiddenError(decision.reason)
@@ -1,76 +0,0 @@
"""Task scheduler primitives."""
from __future__ import annotations
import time
from .errors import NotFoundError
from .capabilities import has_capabilities
from .models import WorkerClaim
from .store import TaskPlaneStore
class TaskScheduler:
def __init__(self, store: TaskPlaneStore):
self.store = store
def next_queued_task(self, worker_id: str):
worker = self.store.get_worker(worker_id)
tasks = sorted(
self.store.list_tasks(),
key=lambda task: (
-int(task.task_plan.metadata.get("priority", 0)),
task.created_at,
task.task_id,
),
)
for task in tasks:
if task.status not in {"CREATED", "QUEUED", "FAILED_RECOVERABLE"}:
continue
if task.task_plan.task_type == "worker.registry.audit":
continue
if task.status == "FAILED_RECOVERABLE" and not self.store.task_has_retryable_probe(
task.task_id
):
continue
if not has_capabilities(
worker.capabilities, task.task_plan.required_capabilities
):
continue
if any(
self.store.get_task(dependency).status != "COMPLETED"
for dependency in task.task_plan.dependencies
):
continue
if (
task.task_plan.data_classification == "credential-secret"
and worker.worker_type != "secure_file_transformer"
and "secure_file_transformer" not in worker.capabilities
):
continue
return task
return None
def schedule_for_worker(self, worker_id: str, lease_seconds: int = 15) -> WorkerClaim:
worker = self.store.get_worker(worker_id)
if worker.disabled:
from .errors import LockedError
raise LockedError("worker disabled")
if time.time() > worker.last_heartbeat + (
worker.heartbeat_interval_seconds * 3
):
from .errors import LockedError
raise LockedError("worker offline")
task = self.next_queued_task(worker_id)
if task is None:
raise NotFoundError("no queued tasks available")
claim = self.store.claim_task(
task.task_id,
worker_id,
worker.identity,
lease_seconds=lease_seconds,
expected_revision=task.revision,
operation_digest=task.task_plan.operation_digest(),
)
return claim
File diff suppressed because it is too large Load Diff
@@ -1,68 +0,0 @@
"""Worker registry abstraction and read helpers."""
from __future__ import annotations
import time
from .capabilities import has_capabilities
from .errors import NotFoundError
from .models import WorkerRegistration
from .store import TaskPlaneStore
class WorkerRegistry:
def __init__(self, store: TaskPlaneStore):
self.store = store
def register(
self,
payload: dict,
*,
actor_identity: str,
is_admin: bool,
expected_revision: int | None,
) -> WorkerRegistration:
reg = WorkerRegistration.from_dict(payload)
return self.store.register_worker(
reg,
actor_identity=actor_identity,
is_admin=is_admin,
expected_revision=expected_revision,
)
def heartbeat(
self,
worker_id: str,
heartbeat_interval_seconds: int | None = None,
*,
actor_identity: str,
expected_revision: int | None,
) -> WorkerRegistration:
return self.store.heartbeat_worker(
worker_id,
heartbeat_interval_seconds=heartbeat_interval_seconds,
actor_identity=actor_identity,
expected_revision=expected_revision,
)
def list(self):
return self.store.list_workers()
def get(self, worker_id: str) -> WorkerRegistration:
try:
return self.store.get_worker(worker_id)
except NotFoundError:
raise
def can_execute(self, worker_id: str, required_capabilities: list[str]) -> bool:
worker = self.store.get_worker(worker_id)
return has_capabilities(worker.capabilities, required_capabilities)
def is_online(self, worker_id: str, *, now: float | None = None) -> bool:
worker = self.store.get_worker(worker_id)
deadline = worker.last_heartbeat + (worker.heartbeat_interval_seconds * 3)
return not worker.disabled and (time.time() if now is None else now) <= deadline
def stale_workers(self, *, now: float | None = None) -> list[WorkerRegistration]:
current = time.time() if now is None else now
return [worker for worker in self.list() if not self.is_online(worker.worker_id, now=current)]
+141
View File
@@ -0,0 +1,141 @@
#!/usr/bin/env python3
"""Prove two real central-decision to Gateway-dispatch to Codex lifecycle cycles."""
from __future__ import annotations
import argparse
import json
from pathlib import Path
import sqlite3
import time
import uuid
from health_probe_common import (
ProbeError, gateway_base, load_config, read_token, request_json, state_version,
)
REQUIRED_MARKERS = {"CODEX_AUTO_CLAIM", "TASK_STARTED", "TASK_EXECUTED", "TASK_ENDED", "EVENT_POSTED"}
def marker_counts(path: Path, command_id: str) -> dict[str, int]:
counts = {marker: 0 for marker in REQUIRED_MARKERS}
with path.open("r", encoding="utf-8") as handle:
for line in handle:
try:
record = json.loads(line)
except json.JSONDecodeError:
continue
marker = record.get("marker")
if record.get("command_id") == command_id and marker in counts:
counts[marker] += 1
return counts
def watcher_status(database: Path, command_id: str) -> str | None:
with sqlite3.connect(f"file:{database}?mode=ro", uri=True, timeout=2) as db:
row = db.execute("SELECT status FROM commands WHERE command_id=?", (command_id,)).fetchone()
return None if row is None else str(row[0])
def wait_for_cycle(config: dict[str, object], token: str, cycle: int) -> tuple[str, str]:
endpoint = str(config["gateway_event_url"])
project = str(config["project_key"])
event_id = f"a0-cycle-{cycle}-" + uuid.uuid4().hex
work_unit_id = f"A0-CODEX-CYCLE-{cycle}-" + uuid.uuid4().hex[:16]
event = {
"schema_version": 1,
"project_key": project,
"contract_commit": config["contract_commit"],
"event_id": event_id,
"source_agent": "system",
"event_type": "a0_unattended_acceptance_requested",
"work_unit_id": work_unit_id,
"status": "READY",
"state_version": state_version(endpoint, project, token),
"implementation_commit": None,
"result_commit": None,
"blocker": None,
"evidence_refs": [],
"payload": {
"acceptance_contract": {
"verdict": "DISPATCH",
"work_unit_id": work_unit_id,
"track_id": "A0_AUTOMATION",
"gate_id": "A0_AUTOMATION_CLOSED",
"primary_agent": "codex",
"fallback_agent": "openclaw",
"clean_base_commit": config["probe_clean_base_commit"],
"allowed_paths": [],
"required_checks": ["acceptance_probe"],
"forbidden_operations": ["production_mutation", "git_mutation"],
"instruction": "Run the fixed side-effect-free acceptance probe only.",
}
},
}
request_json(endpoint, token, method="POST", payload=event)
deadline = time.monotonic() + int(config["cycle_timeout_seconds"])
channel = None
while time.monotonic() < deadline:
_, status = request_json(endpoint + "/" + event_id, token)
if status.get("status") == "TRIGGER_FAILED":
raise ProbeError("central_trigger_failed")
if status.get("status") == "DECIDED" and status.get("decision_applied") is True:
channel = status.get("trigger_channel")
break
time.sleep(float(config["poll_interval_seconds"]))
if channel not in {"workspace_agent", "responses_api"}:
raise ProbeError("central_decision_timeout")
database = Path(str(config["codex_database_path"]))
audit = Path(str(config["codex_audit_log"]))
while time.monotonic() < deadline:
if database.exists() and audit.exists() and watcher_status(database, work_unit_id) == "COMPLETED":
counts = marker_counts(audit, work_unit_id)
if all(counts[marker] == 1 for marker in REQUIRED_MARKERS):
for suffix in ("started", "ended"):
_, lifecycle = request_json(endpoint + "/" + work_unit_id + ":" + suffix, token)
if lifecycle.get("event_id") != work_unit_id + ":" + suffix:
raise ProbeError("gateway_lifecycle_missing")
return event_id, str(channel)
time.sleep(float(config["poll_interval_seconds"]))
raise ProbeError("gateway_dispatch_or_codex_lifecycle_timeout")
def project_summaries_since(database: Path, project: str, since: int) -> int:
with sqlite3.connect(f"file:{database}?mode=ro", uri=True, timeout=2) as db:
row = db.execute(
"SELECT COUNT(*) FROM lifecycle WHERE project_key=? AND kind='PROJECT_SUMMARY' AND created_at>=?",
(project, since),
).fetchone()
return 0 if row is None else int(row[0])
def run(config_path: Path) -> int:
config = load_config(config_path)
endpoint = str(config["gateway_event_url"])
gateway_base(endpoint)
token = read_token(Path(str(config["gateway_token_file"])))
started = int(time.time())
results = [wait_for_cycle(config, token, cycle) for cycle in (1, 2)]
if len({item[0] for item in results}) != 2:
raise ProbeError("duplicate_cycle_event")
if project_summaries_since(
Path(str(config["gateway_database_path"])), str(config["project_key"]), started
) < 2:
raise ProbeError("project_summary_not_persisted")
return 0
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--config", required=True, type=Path)
args = parser.parse_args()
try:
return run(args.config)
except (KeyError, OSError, ProbeError, sqlite3.Error, ValueError):
return 1
if __name__ == "__main__":
raise SystemExit(main())
+172
View File
@@ -0,0 +1,172 @@
#!/usr/bin/env python3
"""Prove a real central-model decision through the authenticated Gateway."""
from __future__ import annotations
import argparse
import json
import os
import stat
import sys
import time
from pathlib import Path
import uuid
from health_probe_common import (
ProbeError, atomic_receipt, gateway_base, load_config, read_token,
request_json, state_version,
)
PENDING_PATH = Path("/run/aios-health/central-gpt.pending.json")
def _load_pending(
path: Path, contract: str, *, owners: set[int] | None = None
) -> tuple[str, int] | None:
allowed_owners = {0} if owners is None else owners
if not path.exists():
return None
info = path.lstat()
if (
not path.is_absolute()
or stat.S_ISLNK(info.st_mode)
or not stat.S_ISREG(info.st_mode)
or stat.S_IMODE(info.st_mode) & ~0o640
or (os.name == "posix" and info.st_uid not in allowed_owners)
or path.resolve(strict=True) != Path(os.path.abspath(path))
):
raise ProbeError("unsafe_pending_marker")
value = json.loads(path.read_text(encoding="utf-8"))
if not isinstance(value, dict) or set(value) != {"contract_commit", "event_id", "state_version"}:
raise ProbeError("invalid_pending_marker")
event_id = value["event_id"]
version = value["state_version"]
if (
value["contract_commit"] != contract
or not isinstance(event_id, str)
or not event_id.startswith("health-central-")
or len(event_id) != 47
or not isinstance(version, int)
or version < 0
):
raise ProbeError("invalid_pending_marker")
return event_id, version
def _clear_pending(path: Path) -> None:
if not path.exists():
return
info = path.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
raise ProbeError("unsafe_pending_marker")
path.unlink()
def _safe_reason(exc: BaseException) -> str:
reason = str(exc)
if reason and len(reason) <= 64 and all(char.islower() or char.isdigit() or char == "_" for char in reason):
return reason
return "probe_error"
def run(config_path: Path, pending_path: Path = PENDING_PATH) -> int:
config = load_config(config_path)
endpoint = config["gateway_event_url"]
gateway_base(endpoint)
token = read_token(Path(config["gateway_token_file"]))
project = config["project_key"]
contract = config["contract_commit"]
pending = _load_pending(pending_path, contract)
is_new = pending is None
if is_new:
event_id = "health-central-" + uuid.uuid4().hex
version = state_version(endpoint, project, token)
atomic_receipt(
pending_path,
{"contract_commit": contract, "event_id": event_id, "state_version": version},
)
else:
event_id, version = pending
event = {
"schema_version": 1,
"project_key": project,
"contract_commit": contract,
"event_id": event_id,
"source_agent": "system",
"event_type": "central_gpt_health_probe",
"work_unit_id": "CENTRAL-GPT-HEALTH",
"status": "READY",
"state_version": version,
"implementation_commit": None,
"result_commit": None,
"blocker": None,
"evidence_refs": [],
"payload": {"probe": "side_effect_free", "nonce": event_id},
}
should_post = is_new
if not is_new:
try:
_, observed = request_json(endpoint + "/" + event_id, token)
except ProbeError as exc:
if str(exc) != "gateway_http_404":
raise
should_post = True
else:
observed = None
if should_post:
try:
status, _ = request_json(endpoint, token, method="POST", payload=event)
except ProbeError as exc:
if str(exc) == "gateway_fail_closed_409":
_clear_pending(pending_path)
raise
if status not in {200, 202}:
raise ProbeError("event_not_accepted")
observed = None
deadline = time.monotonic() + int(config["central_timeout_seconds"])
while time.monotonic() < deadline:
_, observed = request_json(endpoint + "/" + event_id, token)
if observed.get("status") in {"DECIDED", "TRIGGER_FAILED", "DEAD_LETTER"}:
break
time.sleep(float(config["poll_interval_seconds"]))
if isinstance(observed, dict) and observed.get("status") == "DEAD_LETTER":
_clear_pending(pending_path)
raise ProbeError("central_event_dead_letter")
if not isinstance(observed, dict) or observed.get("status") != "DECIDED" or observed.get("decision_applied") is not True:
raise ProbeError("central_decision_not_proven")
channel = observed.get("trigger_channel")
if channel not in {"workspace_agent", "responses_api", "codex_local"}:
raise ProbeError("central_channel_not_proven")
_clear_pending(pending_path)
now = int(time.time())
atomic_receipt(
Path(config["central_receipt_path"]),
{
"subject": "central_gpt",
"health": "HEALTHY",
"live_probe_succeeded": True,
"observed_at_epoch": now,
"event_id": event_id,
"decision_applied": True,
"state_version_after": observed["state_version_after"],
"workspace_agent": {"validated_live": channel == "workspace_agent"},
"responses_api": {"validated_live": channel == "responses_api"},
"codex_local": {"validated_live": channel == "codex_local"},
},
)
return 0
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--config", required=True, type=Path)
args = parser.parse_args()
try:
return run(args.config)
except (KeyError, OSError, ProbeError, ValueError) as exc:
print(f"CENTRAL_GPT_PROBE=BLOCKED reason={_safe_reason(exc)}", file=sys.stderr)
return 1
if __name__ == "__main__":
raise SystemExit(main())
+136
View File
@@ -0,0 +1,136 @@
#!/usr/bin/env python3
from __future__ import annotations
import argparse
import json
import logging
import os
from pathlib import Path
import time
import urllib.error
import urllib.request
from services.aios_orchestrator_gateway.codex_local import CodexLocalClient
from services.aios_orchestrator_gateway.config import Config
from services.aios_orchestrator_gateway.models import Event
from services.aios_orchestrator_gateway.securefs import read_secure_json
LOG = logging.getLogger("aios-codex-central-decision")
class CodexCentralWorker:
def __init__(self, config: Config, client: CodexLocalClient | None = None):
if config.central_channel != "codex_local" or config.codex_queue_dir is None:
raise RuntimeError("codex_local_not_selected")
self.config = config
self.client = client or CodexLocalClient()
self.root = config.codex_queue_dir
self.pending = self.root / "pending"
self.processing = self.root / "processing"
self.processed = self.root / "processed"
self.dead = self.root / "dead-letter"
def recover_interrupted(self) -> None:
for path in sorted(self.processing.glob("*.json")):
target = self.pending / path.name
if target.exists():
path.unlink()
else:
path.replace(target)
def run_one(self) -> bool:
candidates = sorted(self.pending.glob("*.json"))
if not candidates:
return False
source = candidates[0]
claimed = self.processing / source.name
try:
source.replace(claimed)
except FileNotFoundError:
return False
wrapper = read_secure_json(claimed, expected_uid=None)
attempts = wrapper.get("attempts")
next_attempt_at = wrapper.get("next_attempt_at")
event = Event.from_dict(wrapper.get("event"))
if (
not isinstance(attempts, int) or attempts < 0
or not isinstance(next_attempt_at, (int, float)) or next_attempt_at < 0
):
self._dead_letter(claimed)
return True
if next_attempt_at > time.time():
claimed.replace(self.pending / claimed.name)
return False
if event.contract_commit != self.config.contract_commit:
self._dead_letter(claimed)
return True
try:
decision = self.client.decide(event)
self._submit(decision.to_dict(), event.event_id)
except Exception as exc:
LOG.warning("codex_decision_failed category=%s", type(exc).__name__)
wrapper["attempts"] = attempts + 1
wrapper["next_attempt_at"] = int(time.time()) + self.config.retry_base_seconds
self._rewrite(claimed, wrapper)
if wrapper["attempts"] > self.config.max_retries:
self._dead_letter(claimed)
else:
claimed.replace(self.pending / claimed.name)
return True
claimed.replace(self.processed / claimed.name)
return True
def _submit(self, decision: dict, event_id: str) -> None:
body = json.dumps(decision, separators=(",", ":"), sort_keys=True).encode()
url = f"http://{self.config.host}:{self.config.port}/v1/orchestrator/decisions"
request = urllib.request.Request(
url, data=body, method="POST",
headers={
"Authorization": f"Bearer {self.config.callback_token()}",
"Content-Type": "application/json",
"Idempotency-Key": event_id,
},
)
try:
with urllib.request.urlopen(request, timeout=30) as response:
if response.status not in {200, 202}:
raise RuntimeError("gateway_decision_rejected")
except (urllib.error.URLError, TimeoutError, OSError) as exc:
raise RuntimeError("gateway_decision_unavailable") from exc
@staticmethod
def _rewrite(path: Path, value: dict) -> None:
temp = path.with_name(f".{path.name}.tmp-{os.getpid()}")
fd = os.open(temp, os.O_WRONLY | os.O_CREAT | os.O_EXCL | getattr(os, "O_NOFOLLOW", 0), 0o640)
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(value, handle, separators=(",", ":"), sort_keys=True)
handle.flush(); os.fsync(handle.fileno())
os.replace(temp, path)
def _dead_letter(self, path: Path) -> None:
target = self.dead / path.name
if target.exists():
path.unlink()
else:
path.replace(target)
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--gateway-config", required=True, type=Path)
parser.add_argument("--once", action="store_true")
args = parser.parse_args()
logging.basicConfig(level=logging.INFO, format="%(levelname)s %(name)s %(message)s")
worker = CodexCentralWorker(Config.from_file(args.gateway_config))
worker.recover_interrupted()
if args.once:
worker.run_one()
return 0
while True:
if not worker.run_one():
time.sleep(2)
if __name__ == "__main__":
raise SystemExit(main())
+72
View File
@@ -0,0 +1,72 @@
#!/usr/bin/env python3
from __future__ import annotations
import argparse
import json
import os
from pathlib import Path
import sys
import tempfile
import time
import uuid
from services.aios_orchestrator_gateway.codex_local import CodexLocalClient
from services.aios_orchestrator_gateway.config import Config
from services.aios_orchestrator_gateway.models import Event
def _atomic_receipt(path: Path, value: dict) -> None:
path.parent.mkdir(mode=0o700, parents=True, exist_ok=True)
fd, temp = tempfile.mkstemp(prefix=".codex-health-", dir=path.parent)
try:
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(value, handle, separators=(",", ":"), sort_keys=True)
handle.flush(); os.fsync(handle.fileno())
os.chmod(temp, 0o600)
os.replace(temp, path)
finally:
if os.path.exists(temp): os.unlink(temp)
def run(config_path: Path, receipt: Path) -> int:
config = Config.from_file(config_path)
if config.central_channel != "codex_local":
raise RuntimeError("codex_local_not_selected")
event_id = "codex-health-" + uuid.uuid4().hex
event = Event.from_dict({
"schema_version": 1, "project_key": config.project_key,
"contract_commit": config.contract_commit, "event_id": event_id,
"source_agent": "system", "event_type": "codex_local_health_probe",
"work_unit_id": "CODEX-LOCAL-HEALTH", "status": "READY",
"state_version": 0, "implementation_commit": None,
"result_commit": None, "blocker": "health_probe_has_no_release_evidence",
"evidence_refs": [], "payload": {"allowed_work_unit_ids": []},
})
decision = CodexLocalClient().decide(event)
if decision.verdict == "DISPATCH" or decision.next_work_units:
raise RuntimeError("unsafe_health_decision")
_atomic_receipt(receipt, {
"subject": "codex_local", "health": "HEALTHY",
"codex_login_status": 0, "codex_exec": 0,
"decision_schema_valid": True, "observed_at_epoch": int(time.time()),
})
print("CODEX_CENTRAL_HEALTH=PASS")
return 0
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--gateway-config", required=True, type=Path)
parser.add_argument("--receipt", required=True, type=Path)
args = parser.parse_args()
try:
return run(args.gateway_config, args.receipt)
except Exception as exc:
reason = str(exc)
safe = reason if reason.replace("_", "").isalnum() and len(reason) <= 64 else type(exc).__name__
print(f"CODEX_CENTRAL_HEALTH=BLOCKED reason={safe}", file=sys.stderr)
return 1
if __name__ == "__main__":
raise SystemExit(main())
+113
View File
@@ -0,0 +1,113 @@
#!/usr/bin/env python3
"""Queue one fixed no-side-effect command and prove the watcher lifecycle."""
from __future__ import annotations
import argparse
import json
import os
from pathlib import Path
import sqlite3
import stat
import time
import uuid
from health_probe_common import (
ProbeError, atomic_receipt, gateway_base, load_config, read_token,
request_json, state_version,
)
MARKERS = {"CODEX_AUTO_CLAIM", "TASK_STARTED", "TASK_EXECUTED", "TASK_ENDED", "EVENT_POSTED"}
def _write_command(path: Path, value: dict[str, object]) -> None:
directory = path.parent
info = directory.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode) or stat.S_IMODE(info.st_mode) & 0o002:
raise ProbeError("unsafe_command_queue")
data = (json.dumps(value, sort_keys=True, separators=(",", ":")) + "\n").encode()
fd = os.open(path, os.O_WRONLY | os.O_CREAT | os.O_EXCL | getattr(os, "O_NOFOLLOW", 0), 0o640)
try:
os.write(fd, data)
os.fsync(fd)
finally:
os.close(fd)
def _status(database: Path, command_id: str) -> str | None:
with sqlite3.connect(f"file:{database}?mode=ro", uri=True, timeout=2) as db:
row = db.execute("SELECT status FROM commands WHERE command_id=?", (command_id,)).fetchone()
return None if row is None else str(row[0])
def _markers(path: Path, command_id: str) -> dict[str, int]:
result = {name: 0 for name in MARKERS}
with path.open("r", encoding="utf-8") as handle:
for line in handle:
try:
item = json.loads(line)
except json.JSONDecodeError:
continue
if item.get("command_id") == command_id and item.get("marker") in result:
result[item["marker"]] += 1
return result
def run(config_path: Path) -> int:
config = load_config(config_path)
endpoint = config["gateway_event_url"]
gateway_base(endpoint)
token = read_token(Path(config["gateway_token_file"]))
command_id = "health-codex-" + uuid.uuid4().hex
version = state_version(endpoint, config["project_key"], token)
command = {
"schema_version": 1,
"command_id": command_id,
"project_key": config["project_key"],
"work_unit_id": command_id,
"contract_commit": config["contract_commit"],
"state_version": version,
"execution": {"kind": "acceptance_probe", "input": "fixed-side-effect-free-health-probe-v1"},
}
_write_command(Path(config["codex_command_queue"]) / f"{command_id}.json", command)
deadline = time.monotonic() + int(config["codex_timeout_seconds"])
database = Path(config["codex_database_path"])
audit = Path(config["codex_audit_log"])
while time.monotonic() < deadline:
if database.exists() and audit.exists() and _status(database, command_id) == "COMPLETED":
counts = _markers(audit, command_id)
if all(counts[name] == 1 for name in MARKERS):
for suffix in ("started", "ended"):
_, event = request_json(endpoint + "/" + command_id + ":" + suffix, token)
if event.get("event_id") != command_id + ":" + suffix:
raise ProbeError("gateway_lifecycle_missing")
atomic_receipt(
Path(config["codex_receipt_path"]),
{
"subject": "codex",
"health": "HEALTHY",
"live_probe_succeeded": True,
"observed_at_epoch": int(time.time()),
"command_id": command_id,
"lifecycle_complete": True,
"duplicate_side_effects": False,
},
)
return 0
time.sleep(float(config["poll_interval_seconds"]))
raise ProbeError("codex_lifecycle_not_proven")
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--config", required=True, type=Path)
args = parser.parse_args()
try:
return run(args.config)
except (KeyError, OSError, ProbeError, sqlite3.Error, ValueError):
return 1
if __name__ == "__main__":
raise SystemExit(main())
+940
View File
@@ -0,0 +1,940 @@
#!/usr/bin/env python3
"""Durable GPT + one-executor quorum and failover controller.
The controller persists decisions before dispatch and never stores credentials.
Health is accepted only from metadata-only, live-probe receipts.
"""
from __future__ import annotations
import argparse
import hashlib
import json
import os
import sqlite3
import stat
import subprocess
import tempfile
import time
import urllib.error
import urllib.request
import urllib.parse
from dataclasses import dataclass
from enum import Enum
from pathlib import Path
from typing import Any, Callable, Iterable, Mapping
ALLOWED_LIFECYCLE_EVENTS = frozenset(
{
"TASK_DISPATCHED",
"TASK_STARTED",
"TASK_ENDED",
"DECISION_APPLIED",
"PROJECT_SUMMARY",
"PROJECT_COMPLETED",
}
)
EXECUTORS = ("openclaw", "codex")
SECRET_KEYS = frozenset(
{
"secret",
"token",
"password",
"api_key",
"authorization",
"credential",
"private_key",
"environment",
"env",
"argv",
"stdout",
"stderr",
}
)
class Health(str, Enum):
HEALTHY = "HEALTHY"
UNHEALTHY = "UNHEALTHY"
UNKNOWN = "UNKNOWN"
class ControllerError(RuntimeError):
"""Fail-closed controller error."""
class EventConflict(ControllerError):
"""An event id was reused with a different payload."""
class StaleStateVersion(ControllerError):
"""A dispatch attempted to overwrite a newer task state."""
@dataclass(frozen=True)
class QuorumDecision:
allowed: bool
priority: str
status: str
active_executor: str | None
standby_executor: str | None
repair_required: bool
def canonical_json(value: Any) -> str:
return json.dumps(value, sort_keys=True, separators=(",", ":"), ensure_ascii=True)
def digest_json(value: Any) -> str:
return hashlib.sha256(canonical_json(value).encode("utf-8")).hexdigest()
def _contains_forbidden_key(value: Any) -> bool:
if isinstance(value, Mapping):
for key, child in value.items():
normalized = str(key).strip().lower().replace("-", "_")
if normalized in SECRET_KEYS or _contains_forbidden_key(child):
return True
elif isinstance(value, list):
return any(_contains_forbidden_key(item) for item in value)
return False
def validate_metadata_only(value: Mapping[str, Any]) -> None:
if _contains_forbidden_key(value):
raise ControllerError("secret-bearing or output-bearing fields are forbidden")
def atomic_json_write(path: Path, value: Mapping[str, Any], mode: int = 0o600) -> None:
path.parent.mkdir(parents=True, exist_ok=True, mode=0o700)
validate_secure_directory(path.parent)
fd, temporary = tempfile.mkstemp(prefix=f".{path.name}.", dir=str(path.parent))
try:
os.fchmod(fd, mode)
with os.fdopen(fd, "w", encoding="utf-8") as handle:
handle.write(canonical_json(value))
handle.write("\n")
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary, path)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def validate_secure_directory(path: Path, *, expected_uid: int | None = None) -> None:
info = path.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode):
raise ControllerError(f"unsafe directory type rejected: {path}")
if os.name == "posix" and path.resolve(strict=True) != Path(os.path.abspath(path)):
raise ControllerError(f"symlinked directory component rejected: {path}")
if os.name == "posix" and info.st_mode & 0o022:
raise ControllerError(f"group/world-writable directory rejected: {path}")
if expected_uid is not None and info.st_uid != expected_uid:
raise ControllerError(f"unexpected owner for directory {path}")
def read_secure_json(path: Path, *, expected_uid: int | None = None) -> dict[str, Any]:
info = path.lstat()
if stat.S_ISLNK(info.st_mode):
raise ControllerError(f"symlink rejected: {path}")
if not stat.S_ISREG(info.st_mode):
raise ControllerError(f"not a regular file: {path}")
if os.name == "posix" and info.st_mode & 0o022:
raise ControllerError(f"group/world-writable file rejected: {path}")
validate_secure_directory(path.parent, expected_uid=expected_uid)
if os.name == "posix" and path.resolve(strict=True) != Path(os.path.abspath(path)):
raise ControllerError(f"symlinked file component rejected: {path}")
if expected_uid is not None and info.st_uid != expected_uid:
raise ControllerError(f"unexpected owner for {path}")
with path.open("r", encoding="utf-8") as handle:
value = json.load(handle)
if not isinstance(value, dict):
raise ControllerError(f"JSON object required: {path}")
validate_metadata_only(value)
return value
def read_gateway_token(path: Path) -> str:
if not path.is_absolute():
raise ControllerError("gateway token path must be absolute")
try:
info = path.lstat()
except FileNotFoundError as exc:
raise ControllerError("gateway token unavailable") from exc
if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
raise ControllerError("gateway token must be a regular non-symlink file")
if stat.S_IMODE(info.st_mode) & ~0o640:
raise ControllerError("gateway token mode exceeds 0640")
if os.name == "posix" and info.st_uid not in {0, os.geteuid()}:
raise ControllerError("gateway token owner is not approved")
if os.name == "posix" and path.resolve(strict=True) != Path(os.path.abspath(path)):
raise ControllerError("gateway token has a symlinked path component")
flags = os.O_RDONLY | getattr(os, "O_NOFOLLOW", 0)
fd = os.open(path, flags)
try:
opened = os.fstat(fd)
if not stat.S_ISREG(opened.st_mode) or opened.st_ino != info.st_ino or opened.st_dev != info.st_dev:
raise ControllerError("gateway token changed during validation")
data = os.read(fd, 16_385)
if os.read(fd, 1):
raise ControllerError("gateway token too large")
finally:
os.close(fd)
if not data or len(data) > 16_384 or b"\n" in data or b"\r" in data:
raise ControllerError("gateway token format rejected")
try:
token = data.decode("utf-8")
except UnicodeDecodeError as exc:
raise ControllerError("gateway token encoding rejected") from exc
if not token:
raise ControllerError("gateway token unavailable")
return token
def validate_health_receipt(
receipt: Mapping[str, Any],
subject: str,
*,
now: float | None = None,
max_age_seconds: int = 60,
) -> Health:
validate_metadata_only(receipt)
if receipt.get("subject") != subject:
return Health.UNKNOWN
try:
health = Health(str(receipt.get("health", "UNKNOWN")))
except ValueError:
return Health.UNKNOWN
if health is not Health.HEALTHY:
return health
if receipt.get("live_probe_succeeded") is not True:
return Health.UNKNOWN
observed_at = receipt.get("observed_at_epoch")
if not isinstance(observed_at, (int, float)):
return Health.UNKNOWN
now = time.time() if now is None else now
if observed_at > now + 5 or now - observed_at > max_age_seconds:
return Health.UNKNOWN
if subject == "central_gpt":
workspace = receipt.get("workspace_agent")
responses = receipt.get("responses_api")
workspace_ok = isinstance(workspace, Mapping) and workspace.get("validated_live") is True
responses_ok = isinstance(responses, Mapping) and responses.get("validated_live") is True
if not (workspace_ok or responses_ok):
return Health.UNKNOWN
return Health.HEALTHY
def choose_executor(workload: str) -> str:
if workload in {"complex_code", "security_boundary", "architecture", "complex_code_security_architecture"}:
return "codex"
return "openclaw"
def evaluate_quorum(
central_gpt: Health,
openclaw: Health,
codex: Health,
workload: str,
) -> QuorumDecision:
if central_gpt is not Health.HEALTHY:
status = "BLOCKED_CENTRAL_GPT_UNKNOWN" if central_gpt is Health.UNKNOWN else "BLOCKED_CENTRAL_GPT_UNHEALTHY"
return QuorumDecision(False, "P0", status, None, None, True)
healthy = [name for name, health in (("openclaw", openclaw), ("codex", codex)) if health is Health.HEALTHY]
if not healthy:
has_unknown = Health.UNKNOWN in {openclaw, codex}
status = "BLOCKED_EXECUTOR_HEALTH_UNKNOWN" if has_unknown else "BLOCKED_NO_HEALTHY_EXECUTOR"
return QuorumDecision(False, "P1", status, None, None, True)
if len(healthy) == 1:
active = healthy[0]
standby = "codex" if active == "openclaw" else "openclaw"
return QuorumDecision(True, "P3", "EXECUTOR_QUORUM_PASS", active, standby, False)
active = choose_executor(workload)
standby = "codex" if active == "openclaw" else "openclaw"
return QuorumDecision(True, "NONE", "EXECUTOR_QUORUM_PASS", active, standby, False)
class QuorumStore:
def __init__(self, path: Path):
self.path = path
path.parent.mkdir(parents=True, exist_ok=True, mode=0o700)
expected_uid = os.geteuid() if os.name == "posix" else None
validate_secure_directory(path.parent, expected_uid=expected_uid)
if path.exists() or path.is_symlink():
info = path.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode) or info.st_nlink != 1:
raise ControllerError(f"unsafe database path rejected: {path}")
if expected_uid is not None and info.st_uid != expected_uid:
raise ControllerError(f"unexpected database owner: {path}")
self.db = sqlite3.connect(str(path), isolation_level=None, timeout=30)
if os.name == "posix":
path.chmod(0o600)
self.db.row_factory = sqlite3.Row
self.db.execute("PRAGMA journal_mode=WAL")
self.db.execute("PRAGMA synchronous=FULL")
self._initialize()
def close(self) -> None:
self.db.close()
def _initialize(self) -> None:
self.db.executescript(
"""
CREATE TABLE IF NOT EXISTS events (
event_id TEXT PRIMARY KEY,
payload_digest TEXT NOT NULL,
payload_json TEXT NOT NULL,
source TEXT NOT NULL,
status TEXT NOT NULL,
created_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS tasks (
work_unit TEXT PRIMARY KEY,
contract_commit TEXT NOT NULL,
event_id TEXT NOT NULL,
state_version INTEGER NOT NULL,
workload TEXT NOT NULL,
status TEXT NOT NULL,
active_executor TEXT,
standby_executor TEXT,
checkpoint TEXT,
lease_owner TEXT,
lease_until REAL,
interrupted_executor TEXT,
recovery_priority TEXT NOT NULL,
updated_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS lifecycle (
sequence INTEGER PRIMARY KEY AUTOINCREMENT,
work_unit TEXT NOT NULL,
event_type TEXT NOT NULL,
executor TEXT,
state_version INTEGER NOT NULL,
metadata_json TEXT NOT NULL,
created_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS effects (
effect_id TEXT PRIMARY KEY,
work_unit TEXT NOT NULL,
payload_digest TEXT NOT NULL,
status TEXT NOT NULL,
created_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS outbox (
outbox_id TEXT PRIMARY KEY,
destination TEXT NOT NULL,
payload_digest TEXT NOT NULL,
payload_json TEXT NOT NULL,
status TEXT NOT NULL,
attempts INTEGER NOT NULL DEFAULT 0,
next_attempt REAL NOT NULL,
created_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS blockers (
blocker_key TEXT PRIMARY KEY,
occurrences INTEGER NOT NULL,
status TEXT NOT NULL,
updated_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS observed_refs (
ref_name TEXT PRIMARY KEY,
oid TEXT NOT NULL,
updated_at REAL NOT NULL
);
CREATE TABLE IF NOT EXISTS summaries (
work_unit TEXT PRIMARY KEY,
summary_digest TEXT NOT NULL,
summary_json TEXT NOT NULL,
created_at REAL NOT NULL
);
CREATE UNIQUE INDEX IF NOT EXISTS lifecycle_once
ON lifecycle(work_unit, event_type, COALESCE(executor, ''), state_version);
"""
)
def ingest_event(self, event_id: str, payload: Mapping[str, Any], source: str) -> bool:
validate_metadata_only(payload)
encoded = canonical_json(payload)
digest = hashlib.sha256(encoded.encode("utf-8")).hexdigest()
existing = self.db.execute("SELECT payload_digest FROM events WHERE event_id=?", (event_id,)).fetchone()
if existing:
if existing["payload_digest"] != digest:
self.db.execute("UPDATE events SET status='DEAD_LETTER_CONFLICT' WHERE event_id=?", (event_id,))
raise EventConflict(f"event id payload conflict: {event_id}")
return False
self.db.execute(
"INSERT INTO events VALUES(?,?,?,?,?,?)",
(event_id, digest, encoded, source, "ACCEPTED", time.time()),
)
return True
def upsert_task(
self,
*,
work_unit: str,
contract_commit: str,
event_id: str,
state_version: int,
workload: str,
decision: QuorumDecision,
checkpoint: str | None = None,
) -> str:
now = time.time()
self.db.execute("BEGIN IMMEDIATE")
try:
current = self.db.execute("SELECT * FROM tasks WHERE work_unit=?", (work_unit,)).fetchone()
if current and state_version < current["state_version"]:
raise StaleStateVersion(f"stale state version for {work_unit}")
interrupted: str | None = None
recovery_priority = decision.priority
status = "BLOCKED"
if decision.allowed:
status = "READY"
if current and current["active_executor"] and current["active_executor"] != decision.active_executor:
interrupted = current["active_executor"]
recovery_priority = "P2"
status = "INTERRUPTED_AGENT_FAILOVER"
if checkpoint is None:
checkpoint = current["checkpoint"]
if current:
if contract_commit != current["contract_commit"] or event_id != current["event_id"]:
raise ControllerError("failover must preserve contract_commit and event_id")
if (
current["state_version"] == state_version
and current["active_executor"] == decision.active_executor
and current["status"] in {"RUNNING", "COMPLETED"}
):
self.db.execute("COMMIT")
return current["status"]
self.db.execute(
"""UPDATE tasks SET state_version=?, workload=?, status=?, active_executor=?,
standby_executor=?, checkpoint=?, interrupted_executor=?, recovery_priority=?, updated_at=? WHERE work_unit=?""",
(
state_version,
workload,
status,
decision.active_executor,
decision.standby_executor,
checkpoint,
interrupted,
recovery_priority,
now,
work_unit,
),
)
else:
self.db.execute(
"INSERT INTO tasks VALUES(?,?,?,?,?,?,?,?,?,?,?,?,?,?)",
(
work_unit,
contract_commit,
event_id,
state_version,
workload,
status,
decision.active_executor,
decision.standby_executor,
checkpoint,
None,
None,
interrupted,
recovery_priority,
now,
),
)
self.db.execute("COMMIT")
return status
except Exception:
self.db.execute("ROLLBACK")
raise
def claim_lease(self, work_unit: str, worker: str, ttl_seconds: int, now: float | None = None) -> bool:
now = time.time() if now is None else now
self.db.execute("BEGIN IMMEDIATE")
try:
task = self.db.execute(
"SELECT active_executor, status, lease_owner, lease_until FROM tasks WHERE work_unit=?",
(work_unit,),
).fetchone()
if not task:
raise ControllerError(f"unknown work unit: {work_unit}")
if worker != task["active_executor"]:
self.db.execute("ROLLBACK")
return False
if task["status"] in {"BLOCKED", "COMPLETED"}:
self.db.execute("ROLLBACK")
return False
lease_live = task["lease_owner"] and task["lease_until"] and task["lease_until"] > now
if lease_live and task["lease_owner"] != worker:
self.db.execute("ROLLBACK")
return False
self.db.execute(
"UPDATE tasks SET lease_owner=?, lease_until=?, status='RUNNING', updated_at=? WHERE work_unit=?",
(worker, now + ttl_seconds, now, work_unit),
)
self.db.execute("COMMIT")
return True
except Exception:
if self.db.in_transaction:
self.db.execute("ROLLBACK")
raise
def dispatch_task(
self,
*,
work_unit: str,
contract_commit: str,
event_id: str,
state_version: int,
workload: str,
decision: QuorumDecision,
checkpoint: str | None = None,
) -> str:
status = self.upsert_task(
work_unit=work_unit,
contract_commit=contract_commit,
event_id=event_id,
state_version=state_version,
workload=workload,
decision=decision,
checkpoint=checkpoint,
)
if decision.allowed and status != "COMPLETED":
metadata = {
"contract_commit": contract_commit,
"event_id": event_id,
"status": status,
}
self.append_lifecycle(work_unit, "TASK_DISPATCHED", decision.active_executor, state_version, metadata)
self.enqueue(
f"dispatch:{work_unit}:{state_version}:{decision.active_executor}",
"gateway",
{"event_type": "TASK_DISPATCHED", "work_unit": work_unit, **metadata},
)
return status
def start_task(self, work_unit: str, executor: str, state_version: int, lease_ttl_seconds: int) -> bool:
if not self.claim_lease(work_unit, executor, lease_ttl_seconds):
return False
self.append_lifecycle(work_unit, "TASK_STARTED", executor, state_version, {"status": "RUNNING"})
self.enqueue(
f"started:{work_unit}:{state_version}:{executor}",
"gateway",
{"event_type": "TASK_STARTED", "work_unit": work_unit, "executor": executor, "state_version": state_version},
)
return True
def end_task(
self,
work_unit: str,
executor: str,
state_version: int,
result: str,
summary: Mapping[str, Any],
*,
project_completed: bool = False,
) -> None:
validate_metadata_only(summary)
self.db.execute("BEGIN IMMEDIATE")
try:
task = self.db.execute("SELECT active_executor,state_version FROM tasks WHERE work_unit=?", (work_unit,)).fetchone()
if not task or task["active_executor"] != executor:
raise ControllerError("only the active executor may end a task")
if task["state_version"] != state_version:
raise StaleStateVersion(f"stale state version for {work_unit}")
self.db.execute(
"UPDATE tasks SET status='COMPLETED', lease_owner=NULL, lease_until=NULL, updated_at=? WHERE work_unit=?",
(time.time(), work_unit),
)
self.append_lifecycle(work_unit, "TASK_ENDED", executor, state_version, {"result": result})
self.append_lifecycle(work_unit, "DECISION_APPLIED", executor, state_version, {"result": result})
self.append_lifecycle(work_unit, "PROJECT_SUMMARY", executor, state_version, summary)
if project_completed:
self.append_lifecycle(work_unit, "PROJECT_COMPLETED", executor, state_version, {"result": result})
self.persist_summary(work_unit, summary)
outbound = (
("TASK_ENDED", {"result": result}),
("DECISION_APPLIED", {"result": result}),
("PROJECT_SUMMARY", dict(summary)),
)
for event_type, metadata in outbound:
self.enqueue(
f"{event_type.lower()}:{work_unit}:{state_version}:{executor}",
"gateway",
{
"event_type": event_type,
"work_unit": work_unit,
"executor": executor,
"state_version": state_version,
**metadata,
},
)
if project_completed:
self.enqueue(
f"project_completed:{work_unit}:{state_version}:{executor}",
"gateway",
{
"event_type": "PROJECT_COMPLETED",
"work_unit": work_unit,
"executor": executor,
"state_version": state_version,
"result": result,
},
)
self.db.execute("COMMIT")
except Exception:
if self.db.in_transaction:
self.db.execute("ROLLBACK")
raise
def record_effect(self, effect_id: str, work_unit: str, payload: Mapping[str, Any]) -> bool:
validate_metadata_only(payload)
digest = digest_json(payload)
existing = self.db.execute("SELECT payload_digest FROM effects WHERE effect_id=?", (effect_id,)).fetchone()
if existing:
if existing["payload_digest"] != digest:
raise EventConflict(f"effect id payload conflict: {effect_id}")
return False
self.db.execute(
"INSERT INTO effects VALUES(?,?,?,?,?)",
(effect_id, work_unit, digest, "COMPLETED", time.time()),
)
return True
def append_lifecycle(
self,
work_unit: str,
event_type: str,
executor: str | None,
state_version: int,
metadata: Mapping[str, Any],
) -> None:
if event_type not in ALLOWED_LIFECYCLE_EVENTS:
raise ControllerError(f"unsupported lifecycle event: {event_type}")
validate_metadata_only(metadata)
self.db.execute(
"INSERT OR IGNORE INTO lifecycle(work_unit,event_type,executor,state_version,metadata_json,created_at) VALUES(?,?,?,?,?,?)",
(work_unit, event_type, executor, state_version, canonical_json(metadata), time.time()),
)
def persist_summary(self, work_unit: str, summary: Mapping[str, Any]) -> None:
validate_metadata_only(summary)
encoded = canonical_json(summary)
digest = hashlib.sha256(encoded.encode("utf-8")).hexdigest()
self.db.execute(
"INSERT OR REPLACE INTO summaries VALUES(?,?,?,?)",
(work_unit, digest, encoded, time.time()),
)
def enqueue(self, outbox_id: str, destination: str, payload: Mapping[str, Any]) -> bool:
validate_metadata_only(payload)
encoded = canonical_json(payload)
digest = hashlib.sha256(encoded.encode("utf-8")).hexdigest()
existing = self.db.execute("SELECT payload_digest FROM outbox WHERE outbox_id=?", (outbox_id,)).fetchone()
if existing:
if existing["payload_digest"] != digest:
raise EventConflict(f"outbox id payload conflict: {outbox_id}")
return False
self.db.execute(
"INSERT INTO outbox VALUES(?,?,?,?,?,?,?,?)",
(outbox_id, destination, digest, encoded, "PENDING", 0, time.time(), time.time()),
)
return True
def record_blocker(self, blocker_key: str, threshold: int) -> str:
now = time.time()
row = self.db.execute("SELECT occurrences FROM blockers WHERE blocker_key=?", (blocker_key,)).fetchone()
occurrences = (row["occurrences"] if row else 0) + 1
status = "DEAD_LETTER" if occurrences > threshold else "RETRY"
self.db.execute(
"INSERT INTO blockers VALUES(?,?,?,?) ON CONFLICT(blocker_key) DO UPDATE SET occurrences=excluded.occurrences,status=excluded.status,updated_at=excluded.updated_at",
(blocker_key, occurrences, status, now),
)
return status
def deliver_outbox(self, sender: Callable[[Mapping[str, Any]], None], max_attempts: int, now: float | None = None) -> tuple[int, int]:
now = time.time() if now is None else now
sent = 0
failed = 0
rows = self.db.execute(
"SELECT * FROM outbox WHERE status='PENDING' AND next_attempt<=? ORDER BY created_at,outbox_id LIMIT 100",
(now,),
).fetchall()
for row in rows:
payload = json.loads(row["payload_json"])
validate_metadata_only(payload)
try:
sender(payload)
except Exception:
attempts = row["attempts"] + 1
status = "DEAD_LETTER" if attempts >= max_attempts else "PENDING"
delay = min(300, 2 ** min(attempts, 8))
self.db.execute(
"UPDATE outbox SET attempts=?,status=?,next_attempt=? WHERE outbox_id=?",
(attempts, status, now + delay, row["outbox_id"]),
)
failed += 1
else:
self.db.execute(
"UPDATE outbox SET attempts=attempts+1,status='SENT',next_attempt=? WHERE outbox_id=?",
(now, row["outbox_id"]),
)
sent += 1
return sent, failed
def observe_ref(self, ref_name: str, oid: str) -> bool:
row = self.db.execute("SELECT oid FROM observed_refs WHERE ref_name=?", (ref_name,)).fetchone()
changed = row is not None and row["oid"] != oid
self.db.execute(
"INSERT INTO observed_refs VALUES(?,?,?) ON CONFLICT(ref_name) DO UPDATE SET oid=excluded.oid,updated_at=excluded.updated_at",
(ref_name, oid, time.time()),
)
return changed
def spool_after_http_failure(
sender: Callable[[Mapping[str, Any]], None],
payload: Mapping[str, Any],
spool_dir: Path,
event_id: str,
) -> bool:
validate_metadata_only(payload)
try:
sender(payload)
return True
except Exception:
atomic_json_write(spool_dir / f"{event_id}.json", {"event_id": event_id, "payload": dict(payload)})
return False
def poll_file_inbox(store: QuorumStore, inbox_dir: Path) -> int:
if not inbox_dir.exists():
return 0
validate_secure_directory(inbox_dir, expected_uid=os.geteuid() if os.name == "posix" else None)
count = 0
for path in sorted(inbox_dir.glob("*.json")):
if path.is_symlink():
raise ControllerError(f"symlink inbox entry rejected: {path}")
item = read_secure_json(path, expected_uid=os.geteuid() if os.name == "posix" else None)
event_id = item.get("event_id")
payload = item.get("payload")
if not isinstance(event_id, str) or not isinstance(payload, dict):
raise ControllerError(f"invalid inbox envelope: {path}")
if store.ingest_event(event_id, payload, "file-inbox"):
count += 1
path.unlink()
return count
def resolve_git_ref(repo: Path, ref_name: str) -> str:
if not repo.is_absolute():
raise ControllerError("Git repository path must be absolute")
info = repo.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode):
raise ControllerError("Git repository must be a non-symlink directory")
if os.name == "posix" and repo.resolve(strict=True) != Path(os.path.abspath(repo)):
raise ControllerError("Git repository has a symlinked path component")
if os.name == "posix" and info.st_mode & 0o002:
raise ControllerError("world-writable Git repository rejected")
try:
completed = subprocess.run(
[
"git",
"-c",
f"safe.directory={repo}",
"-C",
str(repo),
"rev-parse",
"--verify",
ref_name,
],
check=True,
stdout=subprocess.PIPE,
stderr=subprocess.DEVNULL,
text=True,
timeout=10,
)
except (OSError, subprocess.CalledProcessError, subprocess.TimeoutExpired) as exc:
raise ControllerError("Git ref unavailable") from exc
oid = completed.stdout.strip()
if len(oid) != 40 or any(character not in "0123456789abcdef" for character in oid.lower()):
raise ControllerError("invalid Git object id")
return oid
def poll_git_refs(
store: QuorumStore,
repo: Path,
refs: Iterable[str],
resolver: Callable[[Path, str], str] = resolve_git_ref,
) -> int:
changed = 0
for ref_name in refs:
if not ref_name.startswith("refs/") or ".." in ref_name or ref_name.endswith("/"):
raise ControllerError(f"invalid Git ref: {ref_name}")
oid = resolver(repo, ref_name)
if store.observe_ref(ref_name, oid):
event_id = f"gitref:{ref_name}:{oid}"
if store.ingest_event(event_id, {"ref": ref_name, "oid": oid}, "git-ref-poll"):
changed += 1
return changed
def responses_fallback_allowed(
workspace_agent_ok: bool,
receipt: Mapping[str, Any],
*,
now: float | None = None,
) -> bool:
validate_metadata_only(receipt)
if workspace_agent_ok:
return False
responses = receipt.get("responses_api")
return (
validate_health_receipt(receipt, "central_gpt", now=now) is Health.HEALTHY
and isinstance(responses, Mapping)
and responses.get("validated_live") is True
)
def _validate_gateway_url(url: str) -> None:
parsed = urllib.parse.urlparse(url)
if (
parsed.scheme != "http"
or parsed.hostname != "127.0.0.1"
or parsed.port != 18920
or parsed.path != "/v1/orchestrator/events"
or parsed.params
or parsed.query
or parsed.fragment
):
raise ControllerError("gateway endpoint must be the fixed loopback endpoint")
def get_gateway_state(url: str, project_key: str, token: str, timeout: float = 5.0) -> int:
_validate_gateway_url(url)
state_url = url.removesuffix("/events") + "/state/" + urllib.parse.quote(project_key, safe="")
request = urllib.request.Request(
state_url,
headers={"Authorization": f"Bearer {token}"},
method="GET",
)
try:
with urllib.request.urlopen(request, timeout=timeout) as response:
payload = json.loads(response.read(262_145))
except urllib.error.HTTPError as exc:
if exc.code == 404:
return 0
raise ControllerError(f"gateway state rejected HTTP {exc.code}") from exc
except (urllib.error.URLError, TimeoutError, json.JSONDecodeError) as exc:
raise ControllerError("gateway state unavailable") from exc
version = payload.get("state_version") if isinstance(payload, Mapping) else None
if not isinstance(version, int) or version < 0:
raise ControllerError("gateway state version invalid")
return version
def post_json(url: str, payload: Mapping[str, Any], token: str, timeout: float = 5.0) -> None:
validate_metadata_only(payload)
_validate_gateway_url(url)
request = urllib.request.Request(
url,
data=canonical_json(payload).encode("utf-8"),
headers={"Content-Type": "application/json", "Authorization": f"Bearer {token}"},
method="POST",
)
try:
with urllib.request.urlopen(request, timeout=timeout) as response:
if response.status < 200 or response.status >= 300:
raise ControllerError(f"gateway rejected HTTP {response.status}")
except urllib.error.HTTPError as exc:
raise ControllerError(f"gateway rejected HTTP {exc.code}") from exc
except (urllib.error.URLError, TimeoutError) as exc:
raise ControllerError("gateway unavailable") from exc
def run_once(config_path: Path) -> int:
config = read_secure_json(config_path, expected_uid=0 if os.name == "posix" else None)
gateway_url = config["gateway_event_url"]
token = read_gateway_token(Path(config["gateway_token_file"]))
state_version = get_gateway_state(gateway_url, config["project_key"], token)
store = QuorumStore(Path(config["database_path"]))
try:
receipts: dict[str, Health] = {}
for subject in ("central_gpt", "openclaw", "codex"):
receipt_path = Path(config["health_receipts"][subject])
try:
receipt = read_secure_json(receipt_path, expected_uid=0 if os.name == "posix" else None)
receipts[subject] = validate_health_receipt(
receipt,
subject,
max_age_seconds=int(config.get("max_health_age_seconds", 60)),
)
except (FileNotFoundError, ControllerError, json.JSONDecodeError):
receipts[subject] = Health.UNKNOWN
decision = evaluate_quorum(
receipts["central_gpt"], receipts["openclaw"], receipts["codex"], config.get("default_workload", "large_repetitive_infrastructure_monitoring")
)
quorum_event_id = f"quorum-{time.time_ns()}"
store.enqueue(
quorum_event_id,
"gateway",
{
"schema_version": 1,
"project_key": config["project_key"],
"contract_commit": config["contract_commit"],
"event_id": quorum_event_id,
"source_agent": "system",
"event_type": "executor_quorum_evaluated",
"work_unit_id": "A0-AUTOMATION-QUORUM",
"status": "READY" if decision.allowed else "BLOCKED",
"state_version": state_version,
"implementation_commit": None,
"result_commit": None,
"blocker": None if decision.allowed else decision.status,
"evidence_refs": [],
"payload": {
"allowed": decision.allowed,
"priority": decision.priority,
"quorum_status": decision.status,
"active_executor": decision.active_executor,
"standby_executor": decision.standby_executor,
},
},
)
poll_file_inbox(store, Path(config["file_inbox_path"]))
git_polling = config.get("git_ref_polling")
if isinstance(git_polling, Mapping):
poll_git_refs(store, Path(git_polling["repository"]), git_polling["refs"])
store.deliver_outbox(
lambda payload: post_json(gateway_url, payload, token),
max_attempts=int(config["blocker_retry_threshold"]),
)
return 0 if decision.allowed else 2
finally:
store.close()
def build_parser() -> argparse.ArgumentParser:
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("--config", type=Path, required=True)
parser.add_argument("--once", action="store_true", required=True)
return parser
def main(argv: Iterable[str] | None = None) -> int:
args = build_parser().parse_args(argv)
try:
return run_once(args.config)
except (ControllerError, KeyError, OSError, sqlite3.Error, json.JSONDecodeError) as error:
print(f"automation-quorum-controller: fail-closed: {type(error).__name__}", file=os.sys.stderr)
return 3
if __name__ == "__main__":
raise SystemExit(main())
+146
View File
@@ -0,0 +1,146 @@
#!/usr/bin/env python3
"""Strict metadata-only helpers shared by root health probes."""
from __future__ import annotations
import json
import os
from pathlib import Path
import stat
import tempfile
from typing import Any
import urllib.error
import urllib.parse
import urllib.request
class ProbeError(RuntimeError):
pass
def secure_regular(path: Path, *, max_mode: int, owners: set[int]) -> os.stat_result:
if not path.is_absolute():
raise ProbeError("path_not_absolute")
info = path.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
raise ProbeError("unsafe_file_type")
if stat.S_IMODE(info.st_mode) & ~max_mode:
raise ProbeError("unsafe_file_mode")
if os.name == "posix" and info.st_uid not in owners:
raise ProbeError("unsafe_file_owner")
if os.name == "posix" and path.resolve(strict=True) != Path(os.path.abspath(path)):
raise ProbeError("symlinked_path_component")
return info
def load_config(path: Path) -> dict[str, Any]:
secure_regular(path, max_mode=0o600, owners={0})
value = json.loads(path.read_text(encoding="utf-8"))
if not isinstance(value, dict):
raise ProbeError("config_not_object")
return value
def read_token(path: Path) -> str:
info = secure_regular(path, max_mode=0o640, owners={0, os.geteuid()})
if info.st_size <= 0 or info.st_size > 16_384:
raise ProbeError("invalid_token_size")
fd = os.open(path, os.O_RDONLY | getattr(os, "O_NOFOLLOW", 0))
try:
opened = os.fstat(fd)
if not stat.S_ISREG(opened.st_mode) or opened.st_ino != info.st_ino or opened.st_dev != info.st_dev:
raise ProbeError("token_changed_during_validation")
data = os.read(fd, 16_385)
if os.read(fd, 1):
raise ProbeError("invalid_token_size")
finally:
os.close(fd)
if b"\n" in data or b"\r" in data:
raise ProbeError("invalid_token_format")
try:
return data.decode("utf-8")
except UnicodeDecodeError as exc:
raise ProbeError("invalid_token_encoding") from exc
def gateway_base(endpoint: str) -> str:
parsed = urllib.parse.urlparse(endpoint)
if (
parsed.scheme != "http"
or parsed.hostname != "127.0.0.1"
or parsed.port != 18920
or parsed.path != "/v1/orchestrator/events"
or parsed.query
or parsed.fragment
):
raise ProbeError("invalid_gateway_endpoint")
return endpoint.removesuffix("/events")
def request_json(
url: str,
token: str,
*,
method: str = "GET",
payload: dict[str, Any] | None = None,
timeout: float = 10.0,
) -> tuple[int, dict[str, Any]]:
data = None if payload is None else json.dumps(payload, sort_keys=True, separators=(",", ":")).encode()
headers = {"Authorization": f"Bearer {token}"}
if data is not None:
headers["Content-Type"] = "application/json"
request = urllib.request.Request(url, data=data, headers=headers, method=method)
try:
with urllib.request.urlopen(request, timeout=timeout) as response:
body = json.loads(response.read(262_145))
if not isinstance(body, dict):
raise ProbeError("gateway_response_not_object")
return response.status, body
except urllib.error.HTTPError as exc:
if exc.code in {401, 403, 409, 422}:
raise ProbeError(f"gateway_fail_closed_{exc.code}") from exc
raise ProbeError(f"gateway_http_{exc.code}") from exc
except (urllib.error.URLError, TimeoutError, json.JSONDecodeError) as exc:
raise ProbeError("gateway_unavailable") from exc
def state_version(endpoint: str, project_key: str, token: str) -> int:
url = gateway_base(endpoint) + "/state/" + urllib.parse.quote(project_key, safe="")
try:
_, value = request_json(url, token)
except ProbeError as exc:
if str(exc) == "gateway_http_404":
return 0
raise
version = value.get("state_version")
if not isinstance(version, int) or version < 0:
raise ProbeError("invalid_state_version")
return version
def atomic_receipt(path: Path, value: dict[str, Any]) -> None:
if not path.is_absolute() or path.parent != Path("/run/aios-health"):
raise ProbeError("invalid_receipt_path")
parent = path.parent
info = parent.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISDIR(info.st_mode):
raise ProbeError("unsafe_receipt_directory")
if os.name == "posix" and (info.st_uid != 0 or stat.S_IMODE(info.st_mode) != 0o750):
raise ProbeError("unsafe_receipt_directory_metadata")
fd, temporary = tempfile.mkstemp(prefix=f".{path.name}.", dir=parent)
try:
# Receipts contain availability metadata only. Give the dedicated
# aios-quorum group read access so its unprivileged controller can
# consume them, while retaining root ownership and no public access.
if os.name == "posix":
os.fchown(fd, 0, info.st_gid)
os.fchmod(fd, 0o640)
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(value, handle, sort_keys=True, separators=(",", ":"))
handle.write("\n")
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary, path)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
+76
View File
@@ -0,0 +1,76 @@
#!/usr/bin/env python3
from __future__ import annotations
import argparse
import json
import os
from pathlib import Path
import sys
import tempfile
import time
from services.aios_orchestrator_gateway.config import Config
from services.aios_orchestrator_gateway.remote_codex import RemoteCodexBroker
def atomic_receipt(path: Path, value: dict) -> None:
path.parent.mkdir(parents=True, exist_ok=True, mode=0o700)
fd, temporary = tempfile.mkstemp(prefix=".remote-codex-health-", dir=path.parent)
try:
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(value, handle, separators=(",", ":"), sort_keys=True)
handle.flush()
os.fsync(handle.fileno())
os.chmod(temporary, 0o600)
os.replace(temporary, path)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def run(config_path: Path, receipt: Path, healthy_within: int) -> int:
config = Config.from_file(config_path)
if config.central_channel != "remote_codex":
raise RuntimeError("remote_codex_not_selected")
broker = RemoteCodexBroker(
config.database_path,
contract_commit=config.contract_commit,
lease_seconds=config.remote_job_lease_seconds,
max_retries=config.max_retries,
)
try:
snapshot = broker.health_snapshot(healthy_within_seconds=healthy_within)
finally:
broker.close()
if snapshot["healthy_runners"] < 1:
raise RuntimeError("remote_codex_runner_unhealthy")
atomic_receipt(receipt, {
"subject": "remote_codex",
"health": "HEALTHY",
"contract_commit": config.contract_commit,
"healthy_runners": snapshot["healthy_runners"],
"runner_count": snapshot["runner_count"],
"job_counts": snapshot["job_counts"],
"observed_at_epoch": int(time.time()),
})
print("REMOTE_CODEX_HEALTH=PASS")
return 0
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--gateway-config", required=True, type=Path)
parser.add_argument("--receipt", required=True, type=Path)
parser.add_argument("--healthy-within-seconds", type=int, default=300)
args = parser.parse_args()
try:
return run(args.gateway_config, args.receipt, args.healthy_within_seconds)
except Exception as exc:
reason = str(exc)
safe = reason if reason.replace("_", "").isalnum() and len(reason) <= 64 else type(exc).__name__
print(f"REMOTE_CODEX_HEALTH=BLOCKED reason={safe}", file=sys.stderr)
return 1
if __name__ == "__main__":
raise SystemExit(main())
+12 -24
View File
@@ -1,25 +1,13 @@
{
"current_baseline": "V2R1C2",
"previous_baseline": "V2R1C1",
"promoted_at": "2026-07-22T01:40:51Z",
"git": "3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24",
"promoted_by": "codex/owner-approved-gitops",
"promotion_reason": "Owner-authorized V2R1C2 Unified Portal production release after exact tag binding, complete evidence verification, and production acceptance PASS.",
"version": "V2R1C2",
"release_oid": "3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24",
"tag_oid": "a3bd84ebec9c705f7c9c5c03f7f4ec55f8854e36",
"target_oid": "3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24",
"release_run_id": "AIOS-V2R1C2-PRODUCTION-20260722T011836Z",
"status": "PRODUCTION_DEPLOYED",
"production_service": "aios-v2r1c2-production.service",
"production_port": 18920,
"rollback_version": "V2R1C1",
"constraints": {
"product_code_changed": true,
"production_redeployed": true,
"production_traffic_changed": true,
"tag_rewritten": false,
"force_push_allowed": false,
"rollback_assets_preserved": true
}
}
"current_baseline": "V1R4C6",
"previous_baseline": "V1R4C5",
"promoted_at": "2026-07-11T21:00:00+08:00",
"git": "e204a90398bf646d46245b3088226749fa056edd",
"promoted_by": "openclaw/admin",
"promotion_reason": "V1R4C6 admin search agent integration, search gateway, agent platform, VERSION bump, merge tests deploy pipeline — not yet deployed to production",
"constraints": {
"memory_search_enabled": false,
"force_push_allowed": false,
"gitea_db_mutation_requires_owner_approval": true
}
}
@@ -0,0 +1,24 @@
{
"$schema": "../schemas/automation-quorum-controller.schema.json",
"project_key": "v2r1-personal-weixin-production",
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"database_path": "/var/lib/aios-automation-quorum/controller.sqlite3",
"file_inbox_path": "/var/lib/aios-automation-quorum/inbox",
"default_workload": "large_repetitive_infrastructure_monitoring",
"health_receipts": {
"central_gpt": "/run/aios-health/central-gpt.json",
"openclaw": "/run/aios-health/openclaw.json",
"codex": "/run/aios-health/codex.json"
},
"git_ref_polling": {
"repository": "/data/aios-gitops/repo.git",
"refs": [
"refs/heads/task/governance/v2r1-personal-weixin-production-contract-20260717"
]
},
"blocker_retry_threshold": 3,
"max_health_age_seconds": 60,
"lease_ttl_seconds": 120,
"gateway_event_url": "http://127.0.0.1:18920/v1/orchestrator/events",
"gateway_token_file": "/etc/aios-orchestrator-gateway/inbound.token"
}
+17
View File
@@ -0,0 +1,17 @@
{
"active_dir": "/data/aios-gitops/workspace/agents/commands/active",
"completed_dir": "/data/aios-gitops/workspace/agents/commands/completed",
"dead_letter_dir": "/data/aios-gitops/workspace/agents/commands/dead-letter",
"database_path": "/var/lib/aios-codex-watcher/state.sqlite3",
"audit_log_path": "/var/log/aios-codex-watcher/audit.jsonl",
"gateway_endpoint": "http://127.0.0.1:18920/v1/orchestrator/events",
"gateway_token_file": "/etc/aios-orchestrator-gateway/inbound.token",
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"lease_seconds": 60,
"retry_seconds": 10,
"max_attempts": 5,
"poll_seconds": 5,
"executor_path": null,
"executor_args": [],
"executor_timeout_seconds": 1800
}
+17
View File
@@ -0,0 +1,17 @@
{
"project_key": "v2r1-personal-weixin-production",
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"gateway_event_url": "http://127.0.0.1:18920/v1/orchestrator/events",
"gateway_token_file": "/etc/aios-orchestrator-gateway/inbound.token",
"gateway_database_path": "/var/lib/aios-orchestrator-gateway/state.sqlite3",
"probe_clean_base_commit": "c3984ca2b4388875d2875b4fdfdf95a2984f6ab3",
"central_receipt_path": "/run/aios-health/central-gpt.json",
"codex_receipt_path": "/run/aios-health/codex.json",
"codex_command_queue": "/data/aios-gitops/workspace/agents/commands/active",
"codex_database_path": "/var/lib/aios-codex-watcher/state.sqlite3",
"codex_audit_log": "/var/log/aios-codex-watcher/audit.jsonl",
"central_timeout_seconds": 90,
"codex_timeout_seconds": 90,
"cycle_timeout_seconds": 180,
"poll_interval_seconds": 1
}
@@ -0,0 +1,65 @@
{
"host": "127.0.0.1",
"port": 18920,
"database_path": "/var/lib/aios-orchestrator-gateway/state.sqlite3",
"event_inbox_dir": "/var/lib/aios-orchestrator-gateway/inbox/events",
"decision_inbox_dir": "/var/lib/aios-orchestrator-gateway/inbox/decisions",
"dispatch_outbox_dir": "/var/lib/aios-orchestrator-gateway/outbox",
"openclaw_sync_dir": "/var/lib/aios-orchestrator-gateway/openclaw-sync",
"command_queue_dir": "/var/lib/aios-orchestrator-gateway/commands",
"inbound_token_file": "/etc/aios-orchestrator-gateway/inbound.token",
"callback_token_file": "/etc/aios-orchestrator-gateway/callback.token",
"contract_commit": "cbe33ecab8e9794ba5f318bea43aeaf447a56fbc",
"central_channel": "remote_codex",
"codex_local": {
"queue_dir": "/var/lib/aios-codex-brain/queue",
"timeout_seconds": 300
},
"remote_codex": {
"request_max_bytes": 262144,
"job_lease_seconds": 120,
"runners": [
{
"runner_id": "codex-brain-192-168-88-6",
"token_file": "/etc/aios-orchestrator-gateway/remote-codex-88-6.token",
"scopes": ["claim", "heartbeat", "decision", "fail"]
}
]
},
"workspace_agent": {
"trigger_id": null,
"access_token_file": null
},
"responses_fallback": {
"api_key_file": null,
"model": null
},
"polling": {
"interval_seconds": 10,
"summary_interval_seconds": 300
},
"loop_controls": {
"callback_timeout_seconds": 900,
"lease_seconds": 60,
"retry_base_seconds": 10,
"max_retries": 5,
"max_same_blocker_retries": 3,
"max_hops_per_gate": 20
},
"openclaw_sync": {
"endpoint": null,
"token_file": null
},
"project_key": "v2r1-personal-weixin-production",
"git_poll": {
"enabled": true,
"bare_repo": "/data/aios-gitops/repo.git",
"ref_prefixes": [
"refs/heads/main",
"refs/heads/task/governance/",
"refs/heads/patch/",
"refs/heads/result/"
],
"interval_seconds": 30
}
}
-103
View File
@@ -1,103 +0,0 @@
{
"schema_version": 4,
"automatic": [
"build",
"test",
"artifact",
"canary",
"evidence"
],
"owner": [
"signed_release_plan"
],
"forbidden": [
"force_push",
"tag_rewrite",
"direct_main_push",
"baseline_mutation",
"registry_mutation"
],
"runtime_deployment": {
"approval_action": "signed_release_plan",
"required_capability": "runtime_switch",
"tag_operations_allowed": false,
"baseline_mutation_allowed": false,
"registry_mutation_allowed": false,
"writer_default": false,
"rollback_required": true
},
"tag_release": {
"approval_action": "signed_release_plan",
"required_capability": "tag_publish",
"separate_controller": true,
"duplicate_conflict": "fail_closed",
"rewrite_allowed": false,
"force_push_allowed": false
},
"baseline_registry_finalize": {
"approval_action": "signed_release_plan",
"required_capability": "baseline_registry_finalize",
"separate_controller": true,
"remote_tag_verification_required": true,
"gitea_api_verification_required": true,
"registry_baseline_consistency_required": true
},
"approval": {
"algorithm": "Ed25519",
"trust_domain": "aios.production",
"maximum_lifetime_seconds": 3600,
"release_plan_sha256_required": true,
"not_before_required": true,
"replay_protection": true,
"revocation_fail_closed": true,
"key_rotation_supported": true,
"capabilities": [
"runtime_switch",
"tag_publish",
"baseline_registry_finalize"
]
},
"execution": {
"environment_variable": "AIOS_SILENT_RELEASE_EXECUTION_ENABLED",
"default_enabled": false,
"shadow_side_effects_allowed": false,
"durable_saga_required": true,
"pre_switch_revalidation_required": true,
"trusted_candidate_builder_required": true,
"actual_plan_regeneration_required": true,
"startup_fail_closed_on_missing_component": true,
"recovery_worker_shadow_default": true,
"production_approval_backend_required": true,
"production_builder_signer": "kms-hsm-secret-manager-only",
"production_execution_without_backend": "unavailable"
},
"durability": {
"journal": "sha256-hash-chain",
"probe_journal": "sha256-hash-chain",
"exclusive_capability_claims": true,
"side_effect_probe_before_lease_takeover": true,
"side_effect_probe_before_failed_recoverable_retry": true,
"probe_results": [
"NOT_APPLIED",
"APPLIED_VERIFIED",
"INDETERMINATE",
"CONFLICT"
],
"probe_operation_digest_required": true,
"probe_freshness_required": true,
"tamper_result": "MANUAL_OWNER_REVIEW_REQUIRED"
},
"tag_pending": {
"required_observations": [
"remote_git_ref",
"annotated_tag_object",
"peeled_target",
"gitea_api",
"gitea_listing",
"repository_identity"
],
"indeterminate_result": "RECONCILIATION_REQUIRED",
"conflict_result": "MANUAL_OWNER_REVIEW_REQUIRED",
"automatic_reconciliation": false
}
}
-59
View File
@@ -1,59 +0,0 @@
version: 1
policy_id: gitops-isolated-worktree-governance-20260714T232948UTC8
runtime_identity:
user: aiman
group: gitops
umask: "0002"
prohibit_root_workspace_writes: true
repository:
source_worktree: /data/aios-gitops/workspace
bare_repo: /data/aios-gitops/repo.git
worktree_root: /data/aios-gitops/worktrees
remote: origin
worktrees:
controller: /data/aios-gitops/worktrees/controller
task_pattern: /data/aios-gitops/worktrees/{agent}/{task_id}
allowed_agents:
- openclaw
- chatgpt
- controller
require_clean_start: true
remove_after_successful_push: true
retain_on_failure: true
permissions:
owner: aiman
group: gitops
directory_mode: "2775"
file_mode: "0664"
default_acl:
user:aiman: rwx
group:gitops: rwx
apply_to:
- /data/aios-gitops
- /data/aios-gitops/worktrees
preflight:
required_user: aiman
required_group: gitops
required_umask: "0002"
require_writable_git_dir: true
require_writable_task_dir: true
require_writable_result_dir: true
reject_shared_source_worktree: true
reject_existing_git_lock: true
notifications:
owner_on_received: required
owner_on_terminal: required
preferred_channel: weixin
redact_secrets: true
rollback:
preserve_source_worktree: true
remove_created_worktrees: true
restore_service_dropins: true
restore_service_state: true
@@ -1,14 +0,0 @@
# Install as a systemd drop-in for GitOps/OpenClaw writer services.
# Example: /etc/systemd/system/<service>.service.d/10-runtime-identity.conf
[Service]
User=aiman
Group=gitops
SupplementaryGroups=gitops
UMask=0002
Environment=AIOS_GITOPS_ROOT=/data/aios-gitops
Environment=AIOS_BARE_REPO=/data/aios-gitops/repo.git
Environment=AIOS_WORKTREE_ROOT=/data/aios-gitops/worktrees
Environment=AIOS_RUNTIME_USER=aiman
Environment=AIOS_RUNTIME_GROUP=gitops
Environment=AIOS_RUNTIME_UMASK=0002
@@ -1,27 +0,0 @@
# AIOS V2R1C2 Unified Portal production metadata.
# No secret values belong in this file; runtime integrations use references.
AIOS_V2_HOST=127.0.0.1
AIOS_V2_PORT=18920
AIOS_V2_STATE_DIR=/var/lib/aios-v2r1c2-production
AIOS_PROVENANCE_STORE_FILE=/var/lib/aios-v2r1c2-production/run-provenance.sqlite3
AIOS_EXTERNAL_TASK_STORE_FILE=/var/lib/aios-v2r1c2-production/external-task-truth.sqlite3
AIOS_EXECUTOR_CATALOG_FILE=/var/lib/aios-v2r1c2-production/executor-catalog.json
AIOS_EXECUTION_PLAN_FILES=/opt/aios-v2r1c2/plans/AIOS-V2R1C2-UNIFIED-PORTAL-RELEASE.json:/opt/aios-v2r1c2/plans/AIOS-V2R1C2-UNIFIED-PORTAL-TAKEOVER.json
AIOS_EXECUTION_PLAN_FILE=/opt/aios-v2r1c2/plans/AIOS-V2R1C2-UNIFIED-PORTAL-TAKEOVER.json
AIOS_RELEASE_VERSION=V2R1C2
AIOS_TARGET_VERSION=V2R1C2
AIOS_BASELINE_VERSION=V2R1C1
AIOS_PREVIOUS_VERSION=V2R1C1
AIOS_RELEASE_STATUS=PRODUCTION
AIOS_RELEASE_SOURCE_OID=3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24
AIOS_RELEASE_OID=3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24
AIOS_RELEASE_TAG=V2R1C2
AIOS_RELEASE_EVIDENCE_SHA256=2e27bb73b8e6297bd3077238f73a96d697fc3c3611c601c4dc9357e5fdcd8eee
AIOS_OPENCLAW_CONTRACT_VERSION=aios.openclaw.command.v2
AIOS_OPENCLAW_SIGNATURE_ALGORITHM=Ed25519
AIOS_OPENCLAW_PLUGIN_VERSION=2.1.0-rc.4
AIOS_OPENCLAW_SIGNING_KEY_ID=openclaw-v2r1-candidate-ed25519
AIOS_OWNER_WECHAT_TARGET_REF=AIOS_OWNER_WECHAT_TARGET_REF
@@ -1,44 +0,0 @@
[Unit]
Description=AIOS V2R1C2 Unified Portal Production
After=network-online.target
Wants=network-online.target
Conflicts=aios-orchestrator-gateway.service
ConditionPathExists=/opt/aios-v2r1c2/active-runtime/app/aios_web/release_application.py
ConditionPathExists=/opt/aios-v2r1c2/plans/AIOS-V2R1C2-UNIFIED-PORTAL-TAKEOVER.json
[Service]
Type=simple
User=aios
Group=aios
WorkingDirectory=/opt/aios-v2r1c2/active-runtime/app
EnvironmentFile=/etc/aios-v2/aios-v2r1c2-production.env
Environment=PYTHONUNBUFFERED=1
Environment=PYTHONDONTWRITEBYTECODE=1
ExecStartPre=/usr/bin/test "$AIOS_V2_PORT" = "18920"
ExecStartPre=/usr/bin/test "$AIOS_RELEASE_SOURCE_OID" = "3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24"
ExecStartPre=/usr/bin/test "$AIOS_RELEASE_TAG" = "V2R1C2"
ExecStart=/usr/bin/python3 -m aios_web.release_application
Restart=on-failure
RestartSec=3
TimeoutStartSec=60
TimeoutStopSec=30
KillSignal=SIGTERM
NoNewPrivileges=true
PrivateTmp=true
ProtectSystem=strict
ProtectHome=true
ProtectKernelTunables=true
ProtectKernelModules=true
ProtectControlGroups=true
RestrictSUIDSGID=true
RestrictRealtime=true
LockPersonality=true
MemoryDenyWriteExecute=true
ReadWritePaths=/var/lib/aios-v2r1c2-production
StateDirectory=aios-v2r1c2-production
StateDirectoryMode=0750
UMask=0027
LimitNOFILE=65536
[Install]
WantedBy=multi-user.target
+1 -2
View File
@@ -11,10 +11,9 @@ Environment=AIOS_GITOPS_WORKSPACE=/data/aios-gitops/workspace
Environment=AIOS_GITOPS_DB=/data/aios-gitops/runtime/gitops_model.sqlite3
Environment=AIOS_GITOPS_TOKEN_FILE=/data/aios-gitops/runtime/secrets/gitops.tokens
Environment=AIOS_GITOPS_BARE_REPO=/data/aios-gitops/runtime/gitea-bin/data/repositories/aios/aios-v23.git
Environment=AIOS_GITOPS_AUTHORITATIVE_GIT_DIR=/data/aios-gitops/repo.git
Environment=AIOS_GITOPS_REPOSITORY_IDENTITY=aios/aios
Environment=AIOS_GITOPS_HOST=127.0.0.1
Environment=AIOS_GITOPS_PORT=18890
EnvironmentFile=-/etc/aios-gitops/orchestrator-callback.env
ExecStart=/usr/bin/python3 /data/aios-gitops/workspace/active-runtime/app/aios_chatgpt_actions_api.py
Restart=always
RestartSec=3
@@ -1,25 +0,0 @@
[Unit]
Description=AIOS Silent Release Shadow Controller
After=network-online.target
Wants=network-online.target
[Service]
Type=simple
User=aiman
Group=gitops
UMask=0077
Environment=AIOS_GITOPS_WRITES_ENABLED=false
Environment=AIOS_SILENT_RELEASE_EXECUTION_ENABLED=false
Environment=PYTHONDONTWRITEBYTECODE=1
ExecStart=/usr/bin/python3 /usr/libexec/aios-silent-release-shadow/current/active-runtime/app/silent_release_execution.py --state-root /var/lib/aios-silent-release-shadow --repository /data/aios-gitops/repo.git --host 127.0.0.1 --port 18891 --poll-seconds 30
Restart=on-failure
RestartSec=5
NoNewPrivileges=true
PrivateTmp=true
ProtectSystem=strict
ProtectHome=true
ReadWritePaths=/var/lib/aios-silent-release-shadow
ReadOnlyPaths=/data/aios-gitops/repo.git /usr/libexec/aios-silent-release-shadow
[Install]
WantedBy=multi-user.target
+137
View File
@@ -0,0 +1,137 @@
# AIOS GPT + one-executor automation quorum
## Authority and release boundary
The Gateway is the only authoritative state writer. It must hold the SQLite writer lease before changing task state. A model response, service process, configured credential, or agent presence does not establish health. Release remains paused until all live acceptance rows are backed by logs and exit codes.
Minimum quorum:
```text
central_gpt=HEALTHY AND (openclaw=HEALTHY OR codex=HEALTHY)
```
`UNKNOWN` never counts as healthy. If one executor is proven healthy, an unhealthy or unknown standby is P3 telemetry only and cannot block the active gate. If no executor is proven healthy, the controller blocks. Central GPT `UNKNOWN` or `UNHEALTHY` always blocks.
## Live health receipts
The controller reads three root-owned metadata-only JSON receipts from `/run/aios-health`. Files must be regular files, must not be symlinks, and neither files nor parent directories may be group/world writable. The controller rejects fields named `secret`, `token`, `password`, `api_key`, `authorization`, `credential`, `private_key`, `environment`, `env`, `argv`, `stdout`, or `stderr` at any depth.
A receipt claiming `HEALTHY` must contain:
```json
{
"subject": "codex",
"health": "HEALTHY",
"live_probe_succeeded": true,
"observed_at_epoch": 1784347200
}
```
The central GPT receipt additionally requires either `workspace_agent.validated_live=true` or `responses_api.validated_live=true`. Responses fallback is allowed only after a real Responses decision probe succeeded; credential/configuration presence is insufficient.
The controller and probes use only the fixed loopback endpoint
`http://127.0.0.1:18920/v1/orchestrator/events`. The bearer is read from
`/etc/aios-orchestrator-gateway/inbound.token`; the file must be absolute,
root-owned, regular, non-symlinked, no wider than `0640`, nonempty, and contain
no CR/LF. It is never accepted through argv, environment, logs, receipts, or
Git. HTTP `401`, `403`, `409`, and `422` are terminal fail-closed results for a
probe attempt.
The authenticated metadata query
`GET /v1/orchestrator/events/<event_id>` returns only the event ID, public
status, selected central channel, whether the decision was applied, and the
resulting state version. It never returns the submitted event, model output,
credential, or work-unit instruction.
## Durable behavior
SQLite persists ingress event digests, optimistic task state versions, leases, checkpoints, lifecycle events, effect IDs, outbox messages, blocker counts, observed Git refs, and project summaries. Reusing an event/effect/outbox ID with different content fails closed. An expired lease may be reclaimed; a live lease cannot be stolen.
On active-executor failure, a healthy standby receives the same `contract_commit`, `event_id`, state version, and checkpoint. The former executor is recorded as `INTERRUPTED_AGENT_FAILOVER`. Unique effect IDs prevent replay of completed side effects.
The Gateway persists these events even when OpenClaw is offline:
```text
TASK_DISPATCHED
TASK_STARTED
TASK_ENDED
DECISION_APPLIED
PROJECT_SUMMARY
PROJECT_COMPLETED
```
OpenClaw mirroring is optional and uses the durable outbox with bounded retry and dead-letter. HTTP ingress failure spools a metadata-only file for polling; the configured local Git refs are also polled and changes missed by webhooks become idempotent events. Repeated blockers beyond the configured threshold enter `DEAD_LETTER`.
## Install, verify, and rollback
Installation is versioned and does not overwrite an existing release:
```bash
sudo scripts/install_automation_quorum_controller.sh "$PWD" 20260718.1
sudo systemctl enable --now aios-automation-quorum-controller.timer
sudo systemctl enable --now aios-central-gpt-health-probe.timer
sudo systemctl enable --now aios-codex-executor-health-probe.timer
```
Run deterministic verification before enablement:
```bash
scripts/verify_automation_quorum_controller.sh "$PWD"
```
The two root probes are fixed, bounded, side-effect-free programs. They write
receipts atomically with mode `0600` below the root-owned `0750`
`/run/aios-health`. The Codex probe may write only one `acceptance_probe`
command into the controlled watcher queue; it cannot accept arbitrary command
text from configuration.
Run the strict two-cycle closure only after both services are installed from an
immutable committed release:
```bash
sudo /opt/aios-automation-quorum/current/bin/a0-unattended-loop-probe \
--config /etc/aios-automation-quorum/health-probes.json
```
Each cycle requires a live central decision that dispatches the named Codex
acceptance work unit through the Gateway. The verifier then requires exactly
one claim/start/execute/end/post marker, both lifecycle events in the Gateway,
and a persisted `PROJECT_SUMMARY`. Directly queuing a command cannot satisfy
this two-cycle closure.
Rollback accepts only an existing non-symlink directory below `/opt/aios-automation-quorum/releases`. Repeating rollback to the current target is a no-op:
```bash
sudo scripts/rollback_automation_quorum_controller.sh \
/opt/aios-automation-quorum/releases/20260718.0
```
## Live acceptance matrix
Each row requires captured logs, timestamps, exit codes, committed OIDs, Gitleaks output, and where applicable live Gateway DB evidence. Static unit tests are necessary but cannot close the gate.
1. GPT + OpenClaw healthy, Codex stopped: two unattended cycles.
2. GPT + Codex healthy, OpenClaw stopped: two unattended cycles.
3. Both executors healthy: OpenClaw gets repetitive infrastructure; Codex gets complex code/security.
4. Active stops mid-task: standby continues from checkpoint with no duplicate effect.
5. Standby alone stops: active flow continues and no repair gate is created.
6. GPT stops: execution blocks even with two healthy executors.
7. GPT healthy and both executors stop: `BLOCKED_NO_HEALTHY_EXECUTOR`.
8. `UNKNOWN` is never counted as healthy.
9. HTTP ingress failure recovers through file polling.
10. Missed webhook recovers through Git ref polling.
11. Gateway restart restores events, leases, and outbox.
12. Repeated event ID does not create a second dispatch; changed payload dead-letters.
13. Repeated blocker beyond threshold dead-letters.
14. Workspace Agent failure falls back only with a live-validated Responses channel.
15. OpenClaw offline: Codex lifecycle and project summary remain complete in Gateway.
16. Codex offline: OpenClaw lifecycle and project summary remain complete in Gateway.
Until all rows pass against the live service, retain:
```text
AUTOMATION_CLOSED=NOT_STARTED
BASELINE=V1R4C6
RELEASE=paused
PRODUCTION_RELEASE=NOT_STARTED
```
@@ -1,268 +0,0 @@
# AIOS Silent Release Controller Design Report
状态:隔离分支实现;生产 mutation 禁止
目标:自动准备 → 自动验证 → 单次 Owner 签名授权 → 分控制器自动执行与验收
## 1. 当前痛点
现有流程需要多轮人工传递 commit、tree、artifact、manifest、测试、canary、drop-in
和回滚证据。证据在不同消息里重复抄写,容易出现:
- 构建后 artifact、manifest 或 policy 已变化,但旧批准仍被使用;
- Git、Gitea、systemd 与回滚结果由人工拼接,缺少同一一致性快照;
- runtime switch、tag、Baseline、Registry 的授权边界不清;
- 重启后只剩工作树内临时 JSON,无法可靠恢复;
- approval replay、并发 writer、key revoke 和 secret inheritance 缺少统一防线。
## 2. 新架构
```text
Build / Test / Canary
|
v
ReleaseCandidate + DeploymentPlan
|
v
ReleaseAcceptanceEngine
|
v
READY_FOR_APPROVAL
|
v
SignedReleasePlanApproval (Ed25519)
|
+--> runtime_switch capability
| DeploymentController
|
+--> tag_publish capability
| TagController
|
+--> baseline_registry_finalize capability
BaselineRegistryFinalizeController
Durable ReleaseSagaStore + CapabilityLedger + Event Journal
贯穿三个控制器;控制器互相不拥有对方的 mutation API。
```
一次 Owner 批准是同一个不可变 release plan 的签名,不是一个万能执行入口。三个 capability
分别消费、分别留痕、分别幂等恢复。`force_push`、`tag_rewrite`、`direct_main_push`
永远不是可签发 capability。
## 3. 状态机
```text
PREPARING
-> VERIFYING
-> READY_FOR_APPROVAL
-> APPROVED
-> PRE_SWITCH_REVALIDATING
-> SWITCH_PENDING
-> SWITCHED
-> RUNTIME_VERIFYING
-> TAG_PENDING
-> TAG_VERIFIED
-> BASELINE_REGISTRY_PENDING
-> COMPLETED
任一可恢复故障 -> RECONCILIATION_REQUIRED
对象冲突 -> MANUAL_OWNER_REVIEW_REQUIRED
切换后失败 -> ROLLBACK_PENDING -> ROLLED_BACK
```
状态快照使用 revision compare-and-set;事件同时写入 append-only journal。writer 崩溃后根据
状态确定唯一恢复动作。并发控制器中只有一个 CAS winner。
## 4. 数据模型
### ReleaseCandidate
```text
candidate_id
source_commit
tree_oid
artifact_sha256
manifest_sha256
tests
canary
rollback_ready
git_consistency
gitea_consistency
approval_state
```
### DeploymentPlan
```text
service
release_path / release_root
dropin_path / dropin_sha256
rollback_path / rollback_sha256
source_commit
artifact_sha256
manifest_sha256
policy_sha256
health_contract
sha256
```
### SignedReleasePlanApproval
```text
release_plan_id
candidate_id
source_commit
tree_oid
artifact_sha256
manifest_sha256
deployment_plan_sha256
policy_sha256
approval_plan_id
evidence_closure_oid
owner_approval_record_oid
capabilities
expires_at
nonce
key_id
signature
fingerprint
```
签名内容使用 canonical JSON,`signature` 不参与被签名字段。验证包括 Ed25519 签名、过期时间、
公钥状态、approval revoke、candidate/plan/policy 逐字段绑定和 capability allowlist。
## 5. API
```text
POST /createReleaseCandidate
GET /getReleaseCandidate?candidate_id=...
GET /getApprovalEnvelope?candidate_id=...
POST /submitOwnerApproval
POST /startDeployment
GET /getReleaseSaga?candidate_id=...
```
写接口要求 Bearer scope。`AIOS_SILENT_RELEASE_EXECUTION_ENABLED=false` 是默认值;此时可以
持久化 candidate、生成 approval envelope、保存经验证的 Owner approval 和读取 Saga,
但 `/startDeployment` 只返回 `SHADOW_MODE` 与空 `side_effects`。
## 6. PreSwitchValidator
切换前重新收集并比较:
- source commit 与 tree;
- artifact、manifest、policy 和 deployment plan SHA-256;
- rollback readiness;
- local、authoritative bare 与 Gitea refs;
- repository identity;
- task writer/production writer 边界;
- health contract。
任一变化都抛出:
```text
APPROVAL_STALE_REAPPROVAL_REQUIRED
```
旧批准不会被“最佳努力”继续使用。
## 7. systemd 防线
- service 与 drop-in 均为显式 allowlist;
- release 与 drop-in resolved path 必须位于 release root 内;
- 拒绝普通文件被替换为 symlink,拒绝 symlink escape;
- deployment lock 覆盖 switch、restart 和首次 health check;
- restart 前后记录 MainPID 与 ActiveEnterTimestampMonotonic;
- artifact、drop-in、rollback 和 manifest 均按 plan hash 复验;
- 子进程只接收固定 allowlist 环境,不继承 token、password、credential、askpass、
auth socket、`LD_PRELOAD`、`PYTHONPATH` 或数据库连接串。
## 8. TagController 分离
TagController 必须独立完成 approval、execute、verify,并留下:
```text
evidence_closure_oid
approval_plan_id
repository_identity
remote_tag_oid
remote_peeled_target_oid
gitea_api_tag_oid
gitea_listing_visible
```
同名 tag 的 tag OID 或 peeled target 不同,进入
`MANUAL_OWNER_REVIEW_REQUIRED`;自动流程不 push、不删除、不重写、不 force push。
## 9. Baseline / Registry 门禁
只有以下全部一致时,`baseline_registry_finalize` capability 才能完成:
```text
Registry = Baseline = remote tag = Gitea API = expected target
```
UNKNOWN、API 不可达、repository identity 不符或 listing 不可见均 fail closed。
## 10. Policy as Code
```text
automatic:
build, test, artifact, canary, evidence
owner:
signed_release_plan
capabilities:
runtime_switch
tag_publish
baseline_registry_finalize
forbidden:
force_push
tag_rewrite
direct_main_push
baseline_mutation
registry_mutation
```
最后两个 forbidden 项指“脱离受治理 finalize controller 的直接 mutation”。
## 11. Security model
- Ed25519 公钥 ring 支持 rotation 和 revoke,私钥不进入 runtime;
- approval plan 可撤销,过期、未知 key、错误签名均 fail closed;
- nonce 与 capability operation 持久化,重复相同请求幂等,不同 fingerprint 重放拒绝;
- Saga snapshot 原子写、journal append-only、CAS 防并发 writer;
- credential 只通过受控 reference 使用,不进入 Git、API 响应或异常;
- 子进程环境从空集合构建,不复制 `os.environ`;
- Shadow 模式默认开启,execution adapter 缺失时不能降级执行;
- runtime、tag、Baseline/Registry 分属三个能力最小化控制器。
## 12. Migration plan
1. Shadow:生成 candidate、plan、acceptance、approval package,仅比较现有人工结果。
2. Signed approval:接入 Owner 公钥 ring、撤销源和审计 record;仍不接生产 executor。
3. Canary execution:使用非生产 service/release root 验证锁、PID、health、崩溃恢复和回滚。
4. Controlled production switch:另行 Owner 授权启用 execution flag,只开放 runtime capability。
5. Tag/finalize:分别完成 Gitea 与 Registry/Baseline 生产验收后再启用对应 capability。
每个阶段均可退回上一个阶段;不开启下一能力不会影响已经启用的控制器。
## 13. Rollback plan
- switch 前保存旧 symlink target、PID、启动时间、artifact 与 plan fingerprint;
- restart 或 health 任一失败进入 `ROLLBACK_PENDING`;
- 原子恢复旧 target,只 restart 指定 allowlist service;
- 重新验证旧 runtime commit、writer=false 与 repository identity;
- 成功进入 `ROLLED_BACK`,失败进入 `MANUAL_OWNER_REVIEW_REQUIRED`;
- rollback 不修改 Git main、production tag、Baseline 或 Registry;
- 保留脱敏 journal、approval fingerprint 和失败证据。
## 14. 测试与外部依赖边界
自动测试覆盖签名伪造、字段篡改、过期/revoke/key rotation、nonce replay、CAS race、
崩溃恢复、artifact/policy/plan mutation、tag conflict、secret leakage、runtime without
`.git`、rollback 和 shadow no-side-effects。
本隔离实现不创建生产 KMS/key、不安装 systemd unit、不部署 Action runtime、不创建 tag,
也不修改 Baseline/Registry。上述动作必须在单独生产批准中执行。
@@ -1,107 +0,0 @@
# AIOS Silent Release Execution Readiness
## Scope and stop point
This change makes the controller execution-ready while deploying only a
localhost-bound Shadow control plane. Both production gates remain false:
```text
AIOS_GITOPS_WRITES_ENABLED=false
AIOS_SILENT_RELEASE_EXECUTION_ENABLED=false
```
Shadow completion is evidence preparation, not a production release. It cannot
write a production tag, switch a business runtime, or mutate Baseline/Registry.
## Recovery state machine
Every expired or `FAILED_RECOVERABLE` capability requires fresh, operation-bound
probe evidence before another adapter may run:
```text
NOT_APPLIED -> one new exclusive claim may retry
APPLIED_VERIFIED -> mark COMPLETED without adapter replay
INDETERMINATE -> RECONCILIATION_REQUIRED; no automatic writes
CONFLICT -> MANUAL_OWNER_REVIEW_REQUIRED; no automatic writes
```
Probe evidence binds the immutable operation digest, is freshness checked, and
is appended to a separate SHA-256 hash-chained journal. Completion still
requires the active claim ID. File locks and leases ensure that concurrent
recovery workers cannot execute the same adapter.
`TAG_PENDING` uses independent remote ref, annotated object, peeled target,
Gitea API, Gitea listing, and repository identity evidence. Exact absence may
roll back a switched runtime. Exact publication continues to finalization.
Unavailable/divergent evidence stops at reconciliation; a different OID or
target stops for Owner review. No branch can force, delete, or rewrite a tag.
## Trusted build and approval
Callers submit only an event ID and exact source commit. Test, canary, artifact,
manifest, rollback, secret scan, Git and Gitea evidence are generated by the
trusted builder. The builder emits an Ed25519 attestation binding:
- builder attestation/key IDs;
- source commit and tree OID;
- artifact, manifest and package-closure SHA-256;
- test and canary evidence SHA-256;
- issuance time and signature.
The controller contains only trusted public keys. Production signing is exposed
through a KMS/HSM/secret-manager adapter; a local private-key adapter is marked
test-only and is used only for ephemeral Shadow acceptance.
Owner approvals are immutable, time-bounded, revocable, trust-domain bound, and
single-use per capability ledger. If the production approval backend reference
is absent, production execution is unavailable while Shadow remains available.
Caller-supplied `owner_id` is never a fallback.
## Persistent controller and API
`aios-silent-release-controller.service` runs independently from the business
GitOps API. It uses a single leader lock, durable intents/candidates/Sagas,
startup journal verification, idempotent main events, one notification record,
and recovery reporting. The final deployed service is report-only Shadow mode.
The Action contract exposes:
- `createReleaseIntent`
- `getReleaseCandidate`
- `getReleasePlan`
- `getReleaseAcceptance`
- `submitSignedApproval`
- `getReleaseSaga`
- `getCapabilityStatus`
- `startDeployment`
- `getRecoveryStatus`
Every Shadow response declares `execution_enabled=false`, `shadow_mode=true`
and `side_effects_executed=false`. `startDeployment` returns a simulated
`SHADOW_COMPLETED` result and never claims production capabilities.
## Migration and rollback
1. Merge through a governed pull request.
2. Build from a clean detached worktree at the exact merge commit.
3. Verify artifact, manifest, package closure, tests, live fixture, no-.git
canary, rollback archive and secret scan.
4. Install a versioned Shadow release without changing the business runtime.
5. Start the separate localhost-only controller with both gates false.
6. Run Shadow candidates and freeze production PID/tag/Baseline/Registry hashes.
7. Remove ephemeral test keys and temporary repositories.
Rollback stops and disables only the Shadow controller and restores/removes its
versioned `current` link. The existing production API, tag set, Baseline and
Registry are not involved.
## Future activation envelope
Activation requires a new one-time Owner approval binding the exact controller
commit, artifact/manifest/policy/unit hashes, trusted key IDs, KMS/secret
references, release and rollback roots, authoritative Git/repository identity,
Shadow candidate/evidence IDs, test results, activation plan and rollback plan.
That approval may only flip
`AIOS_SILENT_RELEASE_EXECUTION_ENABLED=false` to `true`; each real release still
requires its own signed ReleasePlan. This change does not request or consume
that activation approval.
@@ -1,84 +0,0 @@
# Silent Release Controller Production Contract
Status: review-only implementation. Production execution remains disabled.
## Trust boundary
The caller submits only release intent. A configured trusted builder owns ref
validation, build, tests, artifact and manifest hashing, package closure,
canary, rollback readiness, Git/Gitea consistency, and secret scanning. Caller
supplied success evidence is rejected.
The resulting immutable `ReleasePlan` contains deployment, tag, and
finalization plans. One Ed25519 owner signature binds its SHA-256 digest,
candidate evidence, policy digest, owner identity, trust domain, validity
window, evidence lineage, and three narrow capabilities.
## Durable execution
The Saga journal is the source of truth. Every record is sequence-numbered and
SHA-256 hash-chained from a fixed genesis hash. Snapshots are rebuildable
caches. Journal corruption, truncation, reordering, missing records, or hash
mismatch moves the release to `MANUAL_OWNER_REVIEW_REQUIRED`.
Capability execution uses an exclusive, leased claim. A second worker receives
`BUSY`; an expired claim is not reassigned until a capability-specific
side-effect probe reports that the operation was not applied. Runtime, tag,
and finalization remain separately claimed and recoverable.
## Execution order
```text
APPROVED
-> PRE_SWITCH_REVALIDATING
-> SWITCH_PENDING
-> SWITCHED
-> RUNTIME_VERIFYING
-> RUNTIME_VERIFIED
-> TAG_PENDING
-> TAG_VERIFIED
-> BASELINE_REGISTRY_PENDING
-> COMPLETED
```
The actual adapter configuration is regenerated before execution and must
match the signed plan. Trusted allowlists come from controlled policy, not the
request.
A failure before verified tag publication rolls the runtime back. A failure
after the exact tag is verified enters `RECONCILIATION_REQUIRED`; the tag is
never deleted or rewritten. A same-name/different-OID conflict enters
`MANUAL_OWNER_REVIEW_REQUIRED`. Registry/Baseline finalization requires exact
remote and Gitea tag proof, and Baseline is recorded last.
## Action and recovery contract
`submitOwnerApproval` verifies and durably stores the signed plan, then moves
the Saga to `APPROVED`. `startDeployment` loads that durable approval and
ignores caller-supplied approval content. Enabling execution without the
trusted builder, verifier, validator, policy, and orchestrator fails during
service construction.
The recovery worker scans incomplete journals and resumes through the same
claims and verification gates. It never automatically resumes a manual-review
Saga. In shadow mode it only reports state and performs no adapter side
effects.
The deployed Action factory does not enable the legacy caller-evidence path.
If the trusted builder is absent, candidate creation is rejected even while
execution remains in shadow mode. The compatibility path exists only through
the explicit test-only `SilentReleaseActionService.shadow()` constructor.
Finalization is composed from separate `RegistryController` and
`BaselineController` claims. Registry is verified first and Baseline is the
last mutation; a retry reuses completed Registry evidence and never moves
Baseline ahead of tag verification.
## Governance
- `force_push`: forbidden
- `tag_rewrite`: forbidden
- `direct_main_push`: forbidden
- production execution default: disabled
- production tag operations in this change: not executed
- Baseline/Registry mutations in this change: not executed
@@ -1,68 +0,0 @@
# Gitea Tag Publication Atomicity
Release publication is a fail-closed Saga bound to one immutable governance
fingerprint:
- `version`
- `target_oid`
- `evidence_closure_oid`
- `approval_plan_id`
The normal progression is:
`PENDING` → `LOCAL_TAG_CREATED` → `REMOTE_PUSH_CONFIRMED` →
`GITEA_API_CONFIRMED` → `BASELINE_UPDATE_PENDING` → `BASELINE_UPDATED` →
`PRODUCTION_DEPLOYED`
An unavailable remote uses `REMOTE_PUSH_PENDING` or
`RECONCILIATION_REQUIRED`. A same-name tag object mismatch, peeled-target
mismatch, changed governance fingerprint, or immutable deployment-evidence
conflict enters `MANUAL_OWNER_REVIEW_REQUIRED`. That state blocks automatic
push and reconciliation; recovery never deletes, rewrites, or force-pushes a
tag.
## Legacy state migration
Readers normalize the historical `TAG_CREATED_LOCAL` value to
`LOCAL_TAG_CREATED` in memory. Read-only status responses therefore expose only
the canonical value. The next governed Saga write persists the canonical name;
reading legacy state alone does not mutate the tag, Baseline, Registry, or Saga
file. Any unrecognized persisted state fails closed as
`RECONCILIATION_REQUIRED`.
`BASELINE_UPDATE_PENDING` is persisted before either `baseline/current.json` or
`versions/<version>.json` changes. Recovery revalidates the governance
fingerprint, evidence lineage, remote annotated tag, peeled target, Gitea API,
listing, and repository identity before idempotently completing matching
partial records. A different tag or target enters
`MANUAL_OWNER_REVIEW_REQUIRED`.
## Durable state
Production writers set `AIOS_GITOPS_PUBLICATION_STATE_DIR` to storage outside
their disposable worktree. Recreating a writer must reuse that directory.
Worktree JSON is not the authoritative Saga store.
## Deployment gate
`mark_production_deployed` performs fresh Git remote, peeled-target, Gitea ref,
Gitea tag-object, listing, repository-identity, evidence-lineage, Registry, and
Baseline checks. Deployment is allowed only when:
`Registry = Baseline = remote tag = Gitea API = expected target`
Deployment evidence must be an immutable `sha256:<64 hex>` or `git:<40 hex>`
identifier.
## Read-only acceptance
- `getTagPublicationStatus` reads durable Saga state.
- `getReleaseConsistency` compares every release authority.
- `runReleaseAcceptance` returns `accepted: true` only for `PASS`.
Unavailable evidence returns `UNKNOWN` with `fail_closed: true`; it is never
treated as acceptance.
The CLI, HTTP Action, and release script all enter the same publisher core for
evidence-lineage verification. The release script stages Registry and Baseline
records only; durable Saga state remains outside Git.
@@ -1,68 +0,0 @@
# Isolated Writer Exact-Path Policy V1
This policy is an additional, service-owned authorization layer for
`POST /api/commit-file`. It never expands a Token Policy V2 grant.
An atomic write is permitted only when all of the following agree:
1. socket peer, audience, operation, branch, base and exact path in the
Token Policy V2 grant;
2. service audience, repository identity, branch, operation and exact path
in `AIOS_GITOPS_PATH_POLICY_FILE`;
3. lexical path, resolved containment and symlink checks;
4. the configured content-size and content-profile checks;
5. operation digest, expected branch HEAD and durable operation ledger.
Any failed layer returns a fail-closed policy error. Coarse `/api/file` and
`/api/commit` routes remain forbidden.
## File contract
The configured path must be absolute. The parent directory must be
`root:gitops 0750`; the regular, non-symlink file must be
`root:gitops 0640`. JSON must use UTF-8 canonical encoding with sorted keys,
compact separators and one trailing newline. Unknown fields, unsupported
operations or profiles, duplicate paths, oversized policies and
group/world-writable files are rejected during Writer startup.
Example:
```json
{"allowed_branch":"patch/codex-review-test-worker-20260724","allowed_operations":{"commit_file.atomic":{"content_profile":"systemd-codex-review-worker-v1","exact_paths":["deploy/systemd/aios-codex-review-test-worker.service"],"max_content_bytes":32768}},"audience":"aios-codex-review-worker-writer.service","policy_id":"codex-review-worker-path-policy-v1","repository_identity":"aios/aios","schema_version":1}
```
The SHA-256 of the canonical policy is exposed in the isolated runtime
status and written to the operation audit record. The request body cannot
select or override the policy file.
## Content profile
`systemd-codex-review-worker-v1` is an exact, fail-closed unit grammar. The
only sections are `[Unit]`, `[Service]` and `[Install]`. Each section has a
directive allowlist; every single-value directive and every repeated
`Environment`, `ReadOnlyPaths` and `ReadWritePaths` entry must match the
governed Shadow unit contract exactly.
The profile fixes the Python module
`task_plane.workers.codex_review_test`, `User=aiman`, `Group=gitops`,
the controlled `/usr/libexec/aios-codex-review-test-worker/current`
working directory, all hardening switches and the three permitted
environment variables. It grants read access only to the authoritative
bare repository and the versioned worker artifact, and write access only
to the worker runtime, state and cache directories.
Unknown sections or directives are rejected, including lifecycle hooks,
credential injection, environment imports, root images, bind mounts,
capability expansion, devices, sockets and top-level writable paths. The
parser also rejects duplicate sections, duplicate single-value directives,
empty resets, continuation lines, CRLF, NUL, non-UTF-8 surrogate content,
overlong lines and Unicode lookalike directive names. It does not use
`ConfigParser(strict=False)` or silently merge repeated values.
Production credentials, if later authorized, must be supplied by a
separately governed deployment drop-in. They are not permitted in the
source unit or this content profile.
Policy changes require a separate maintenance branch and Owner-reviewed
ordinary merge commit. A task-scoped Writer must not edit its own policy
file.
+3 -324
View File
@@ -13,27 +13,6 @@ components:
scheme: bearer
bearerFormat: opaque-token
schemas:
PublicationState:
type: string
description: Canonical durable publication state. Legacy TAG_CREATED_LOCAL is normalized on read and is never emitted by new writes.
enum:
- PENDING
- LOCAL_TAG_CREATED
- REMOTE_PUSH_PENDING
- REMOTE_PUSH_CONFIRMED
- GITEA_API_CONFIRMED
- BASELINE_UPDATE_PENDING
- BASELINE_UPDATED
- PRODUCTION_DEPLOYED
- RECONCILIATION_REQUIRED
- MANUAL_OWNER_REVIEW_REQUIRED
PublicationStatus:
type: object
required: [version, state]
properties:
version: {type: string}
state: {$ref: '#/components/schemas/PublicationState'}
fail_closed: {type: boolean}
BranchRequest:
type: object
required: [branch]
@@ -51,31 +30,6 @@ components:
path: {type: string, example: docs/gitops/chatgpt-change.md}
content: {type: string}
message: {type: string}
AtomicFileCommitRequest:
type: object
additionalProperties: false
required: [branch, path, content, message, expected_head_oid, idempotency_key]
properties:
branch:
type: string
pattern: '^patch/[A-Za-z0-9._/-]+$'
path:
type: string
minLength: 1
maxLength: 512
content:
type: string
maxLength: 1048576
message:
type: string
minLength: 1
maxLength: 200
expected_head_oid:
type: string
pattern: '^[0-9a-f]{40}$'
idempotency_key:
type: string
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{7,127}$'
CommitRequest:
type: object
properties:
@@ -96,20 +50,9 @@ components:
message: {type: string}
TagRequest:
type: object
required: [version, target_oid, evidence_closure_oid, approval_plan_id]
required: [version]
properties:
version: {type: string, example: V1R1C1}
target_oid:
type: string
pattern: '^[0-9a-fA-F]{40}$'
description: Full approved commit OID. The tag is created explicitly at this commit, never implicitly at workspace HEAD.
evidence_closure_oid:
type: string
pattern: '^[0-9a-fA-F]{40}$'
description: Required evidence closure commit. target_oid must be its ancestor.
approval_plan_id:
type: string
description: Audited owner-approved release or tag plan identifier.
message: {type: string}
execute:
type: boolean
@@ -160,216 +103,7 @@ components:
properties:
text: {type: string}
branch: {type: string}
DeploymentPlan:
type: object
required: [service, release_path, release_root, cutover_link, dropin_path, dropin_sha256, systemd_unit_sha256, rollback_path, rollback_sha256, source_commit, tree_oid, artifact_sha256, manifest_sha256, package_closure_sha256, authoritative_git_dir, repository_identity, writer_expected, baseline_before_sha256, registry_before_sha256, production_tag_refset_before_sha256, health_contract_version, timeout_policy, retry_policy, policy_sha256, health_contract]
properties:
service: {type: string}
release_path: {type: string}
release_root: {type: string}
cutover_link: {type: string}
dropin_path: {type: string}
dropin_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
systemd_unit_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
rollback_path: {type: string}
rollback_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
source_commit: {type: string, pattern: '^[0-9a-f]{40}$'}
tree_oid: {type: string, pattern: '^[0-9a-f]{40}$'}
artifact_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
manifest_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
package_closure_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
authoritative_git_dir: {type: string}
repository_identity: {type: string}
writer_expected: {type: boolean, enum: [false]}
baseline_before_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
registry_before_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
production_tag_refset_before_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
health_contract_version: {type: string}
timeout_policy: {type: object}
retry_policy: {type: object}
policy_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
health_contract: {type: object, additionalProperties: true}
ReleaseCandidateInput:
type: object
required: [source_commit, version]
additionalProperties: false
properties:
candidate_id: {type: string, pattern: '^[A-Za-z0-9][A-Za-z0-9_.-]{2,127}$'}
source_commit: {type: string, pattern: '^[0-9a-f]{40}$'}
version: {type: string, pattern: '^V[0-9]+R[0-9]+C[0-9]+(?:[-.][A-Za-z0-9][A-Za-z0-9.-]*)?$'}
ReleaseIntentInput:
type: object
required: [event_id, source_commit]
additionalProperties: false
properties:
event_id: {type: string, pattern: '^[A-Za-z0-9][A-Za-z0-9_.:-]{2,255}$'}
source_commit: {type: string, pattern: '^[0-9a-f]{40}$'}
SignedReleasePlanApproval:
type: object
required: [release_plan_id, candidate_id, source_commit, tree_oid, artifact_sha256, manifest_sha256, deployment_plan_sha256, release_plan_sha256, policy_sha256, approval_plan_id, evidence_closure_oid, owner_approval_record_oid, capabilities, owner_identity, trust_domain, issued_at, not_before, expires_at, nonce, key_id, signature]
properties:
release_plan_id: {type: string}
candidate_id: {type: string}
source_commit: {type: string, pattern: '^[0-9a-f]{40}$'}
tree_oid: {type: string, pattern: '^[0-9a-f]{40}$'}
artifact_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
manifest_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
deployment_plan_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
release_plan_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
policy_sha256: {type: string, pattern: '^[0-9a-f]{64}$'}
approval_plan_id: {type: string}
evidence_closure_oid: {type: string, pattern: '^[0-9a-f]{40}$'}
owner_approval_record_oid: {type: string, pattern: '^[0-9a-f]{40}$'}
owner_identity: {type: string}
trust_domain: {type: string}
issued_at: {type: string, format: date-time}
not_before: {type: string, format: date-time}
capabilities:
type: array
uniqueItems: true
items:
type: string
enum: [runtime_switch, tag_publish, baseline_registry_finalize]
expires_at: {type: string, format: date-time}
nonce: {type: string, minLength: 16}
key_id: {type: string}
signature: {type: string, format: byte, writeOnly: true}
StartDeploymentRequest:
type: object
required: [candidate_id]
additionalProperties: false
properties:
candidate_id: {type: string}
paths:
/createReleaseIntent:
post:
operationId: createReleaseIntent
summary: Create an idempotent intent; all build evidence comes from the trusted builder
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/ReleaseIntentInput'}
responses:
'200': {description: Intent created or idempotently returned}
'400': {description: Caller evidence or invalid intent rejected}
x-openai-isConsequential: true
/createReleaseCandidate:
post:
operationId: createReleaseCandidate
summary: Submit release intent to the trusted builder; caller evidence is rejected
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/ReleaseCandidateInput'}
responses:
'200': {description: Candidate created or idempotently returned}
'400': {description: Invalid or conflicting candidate evidence}
'401': {description: Unauthorized}
x-openai-isConsequential: true
/getReleaseCandidate:
get:
operationId: getReleaseCandidate
summary: Read a durable release candidate
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Candidate}
'404': {description: Candidate not found}
/getReleasePlan:
get:
operationId: getReleasePlan
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Immutable release plan}
'404': {description: Candidate not found}
/getReleaseAcceptance:
get:
operationId: getReleaseAcceptance
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Builder attestation and acceptance checks}
'404': {description: Candidate not found}
/getApprovalEnvelope:
get:
operationId: getApprovalEnvelope
summary: Generate the immutable owner-signature payload
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Approval envelope fingerprint inputs}
'404': {description: Candidate not found}
/submitOwnerApproval:
post:
operationId: submitOwnerApproval
summary: Verify and durably store a signed owner release plan
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/SignedReleasePlanApproval'}
responses:
'200': {description: Signature verified and approval stored}
'400': {description: Invalid approval}
'403': {description: Signature, key, expiry, or revocation failure}
'409': {description: Conflicting release plan}
x-openai-isConsequential: true
/submitSignedApproval:
post:
operationId: submitSignedApproval
summary: Alias for immutable signed owner approval submission
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/SignedReleasePlanApproval'}
responses:
'200': {description: Signature verified and approval stored}
'403': {description: Production approval backend unavailable or signature invalid}
x-openai-isConsequential: true
/startDeployment:
post:
operationId: startDeployment
summary: Resume the durable approved Saga; caller cannot replace approval
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/StartDeploymentRequest'}
responses:
'200': {description: Shadow result or deployment Saga result}
'403': {description: Execution disabled or approval invalid}
'409': {description: Saga conflict}
x-openai-isConsequential: true
/getReleaseSaga:
get:
operationId: getReleaseSaga
summary: Read durable Saga state and event journal
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Saga snapshot}
'404': {description: Saga not found}
/getCapabilityStatus:
get:
operationId: getCapabilityStatus
parameters:
- {in: query, name: candidate_id, required: true, schema: {type: string}}
responses:
'200': {description: Durable single-use capability records}
/getRecoveryStatus:
get:
operationId: getRecoveryStatus
responses:
'200': {description: Fail-closed recovery classification}
/api/base:
get:
operationId: getBaseline
@@ -453,25 +187,6 @@ paths:
schema: {$ref: '#/components/schemas/FileCommitRequest'}
responses: {'200': {description: File committed}, '400': {description: Secret rejected or invalid path}, '401': {description: Unauthorized}}
x-openai-isConsequential: true
/api/commit-file:
post:
operationId: commitFileAtomic
summary: Atomically write one allowlisted file, commit, push, and verify using Token Policy V2
security: [{bearerAuth: []}]
requestBody:
required: true
content:
application/json:
schema: {$ref: '#/components/schemas/AtomicFileCommitRequest'}
responses:
'200': {description: Completed or idempotently replayed}
'400': {description: Invalid canonical request}
'401': {description: Missing, invalid, expired, or revoked Token Policy V2 grant}
'403': {description: Audience, source, branch, operation, path, or policy denied}
'409': {description: Idempotency conflict or manual review required}
'413': {description: Request too large}
'503': {description: Push or remote verification unavailable; recoverable}
x-openai-isConsequential: true
/api/commit:
post:
operationId: commitChanges
@@ -484,39 +199,6 @@ paths:
schema: {$ref: '#/components/schemas/CommitRequest'}
responses: {'200': {description: Commit result}, '401': {description: Unauthorized}}
x-openai-isConsequential: true
/api/tag-publication-status:
get:
operationId: getTagPublicationStatus
summary: Read durable publication Saga status without mutation
parameters:
- {in: query, name: version, required: true, schema: {type: string, pattern: '^V\d+R\d+C\d+(?:-[A-Za-z0-9_.-]+)?$'}}
responses:
'200':
description: Canonical publication state or fail-closed UNKNOWN
content:
application/json:
schema: {$ref: '#/components/schemas/PublicationStatus'}
'400': {description: Version required}
/api/release-consistency:
get:
operationId: getReleaseConsistency
summary: Compare Registry, Baseline, remote tag, Gitea API, listing, repository identity, and expected target
parameters:
- {in: query, name: version, required: true, schema: {type: string}}
- {in: query, name: target_oid, required: true, schema: {type: string, pattern: '^[0-9a-fA-F]{40}$'}}
- {in: query, name: evidence_closure_oid, required: true, schema: {type: string, pattern: '^[0-9a-fA-F]{40}$'}}
- {in: query, name: approval_plan_id, required: true, schema: {type: string}}
responses: {'200': {description: PASS, FAIL, or fail-closed UNKNOWN}}
/api/release-acceptance:
get:
operationId: runReleaseAcceptance
summary: Run read-only release acceptance; UNKNOWN is not accepted
parameters:
- {in: query, name: version, required: true, schema: {type: string}}
- {in: query, name: target_oid, required: true, schema: {type: string, pattern: '^[0-9a-fA-F]{40}$'}}
- {in: query, name: evidence_closure_oid, required: true, schema: {type: string, pattern: '^[0-9a-fA-F]{40}$'}}
- {in: query, name: approval_plan_id, required: true, schema: {type: string}}
responses: {'200': {description: Acceptance result with accepted boolean}}
/api/tag:
post:
operationId: requestOrCreateTag
@@ -527,7 +209,7 @@ paths:
content:
application/json:
schema: {$ref: '#/components/schemas/TagRequest'}
responses: {'200': {description: Gitea-confirmed tag and fail-closed baseline result}, '400': {description: Missing governance evidence or invalid OID}, '403': {description: Forbidden}, '409': {description: Existing tag target conflict or evidence lineage mismatch}, '502': {description: Gitea reconciliation required}, '503': {description: Remote push pending}}
responses: {'200': {description: Tag request committed or tag created}, '403': {description: Forbidden}}
x-openai-isConsequential: true
/api/rollback:
post:
@@ -563,7 +245,4 @@ paths:
content:
application/json:
schema: {$ref: '#/components/schemas/NLRequest'}
responses:
'200': {description: Safe GitOps plan}
'401': {description: Unauthorized}
x-openai-isConsequential: true
responses: {'
-56
View File
@@ -1,56 +0,0 @@
# V2R1C1 中央 Baseline 治理收口
## 收口范围
本次变更只登记已经完成生产验证并上线的 `V2R1C1`。它新增中央版本 Manifest、把中央 Baseline Pointer 从 `V1R4C6` 切换为 `V2R1C1`,并建立 Release、Companion、Artifact 与最终生产 Evidence 的追溯关系。
本次不修改产品实现,不移动或重建 `V2R1C1` Tag,不改变 Release OID,不重新部署生产,不切换生产流量,也不重写历史证据。
## 权威标识
- Release:`release/AIOS-V2R1C1-20260721`
- Release OID:`4e0e6dcb90e6732aa9838fe9d0af0b34746829b0`
- Tag / Tag OID:`V2R1C1` / `4e0e6dcb90e6732aa9838fe9d0af0b34746829b0`
- SemVer:`2.1.0-rc.1`
- OpenClaw Companion:`2.1.0-rc.4`
- Companion Source OID:`83f8f878c7da5dd8263f2af2efa80750492e5cb3`
- Companion OID:`12d1c13170e0c5a691d67a9408e739e7a17064a3`
- Artifact SHA256:`aadd5f3cfa0d8e796a21f6368f5c95e8c60e9063d6e9fcd9c4947cfa0c013671`
- Contract:`aios.openclaw.command.v2` / `Ed25519`
- Evidence Root:`/var/lib/aios-gitops/acceptance/aios-v2r1c1-openclaw-v2-companion-20260721T094442Z`
- Final Audit Manifest SHA256:`2cdb884abf195969de3cdaa44c76aaac52a9165fb4918436855291e46b2f8d55`
- Production Status:`PRODUCTION_DEPLOYED`
## 治理路径
- Governance Branch:`patch/chatgpt/AIOS-V2R1C1-BASELINE-REGISTRY-20260721`
- Base Ref:`refs/heads/main`
- Base OID:`3cf9958cb94614449e093639cbf902216d73447f`
- Manifest Commit:`a56caf46f6bc335ec2395b674ead4b2a427c5edd`
- Baseline Commit:`6dbd5cbe07d2960ca653049fefbcf3a3e18f9ab1`
- Reviewer:`gpt-5.3-codex-spark validator and docs_reporter`
- Review:`PASS`(`2026-07-21T14:23:00Z`)
## 命令与退出码
| 阶段 | 命令 | 开始 UTC | 结束 UTC | Exit Code |
| --- | --- | --- | --- | ---: |
| 中央只读核对 | `GET /api/run; GET /api/base; GET /api/ver` | 2026-07-21T14:09:40Z | 2026-07-21T14:09:40Z | 0 |
| 创建治理分支 | `bash /tmp/registry-create-branch.sh` | 2026-07-21T14:17:05Z | 2026-07-21T14:17:05Z | 0 |
| 核心元数据验证 | `bash /tmp/registry-validate-core.sh` | 2026-07-21T14:19:56Z | 2026-07-21T14:19:56Z | 0 |
完整机读记录位于 `.agent/results/releases/V2R1C1-BASELINE-CLOSURE.json`,治理运行证据位于 `/var/lib/aios-gitops/acceptance/aios-v2r1c1-baseline-registry-20260721T141705Z`。
## 元数据回滚
若中央 Registry 解析或治理服务异常,保留 `versions/V2R1C1.json`,通过新的受审治理提交把 `baseline/current.json` 和 `VERSION` 临时恢复到 `V1R4C6`。该操作不得回滚生产系统、移动 Tag、修改 Release OID 或重写历史。
## 范围证明
```text
product_code_changed=false
tag_changed=false
production_redeployed=false
production_traffic_changed=false
secret_material_recorded=false
```
@@ -1,35 +0,0 @@
# V2R1C2 Production Release
- Release Run ID: `AIOS-V2R1C2-PRODUCTION-20260722T011836Z`
- Previous production baseline: `V2R1C1`
- Tag: `V2R1C2`
- Tag object: `a3bd84ebec9c705f7c9c5c03f7f4ec55f8854e36`
- Approved target: `3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24`
- Evidence closure: `8cdca463aff9707edee00e20a3b56884130f6978`
- Initial registry status: `TAG_CREATED_NOT_PROMOTED`
- Final status: `PRODUCTION_DEPLOYED`
- Production service: `aios-v2r1c2-production.service`
- Production port: `18920`
- Production acceptance SHA256: `a6c35e12cab7907ab6bcfd37d781b3d216f05a5ae002aa54a3112c70f43b991a`
- Production online smoke SHA256: `fe9c6964e440f8871f2eb225f0132f70c634923389f7b8dfa85f615b100980a4`
- Production health SHA256: `ee8084e8b57d188dc0039a276f4f3f1ed7025ff58ce91b111cc765627eb52d4c`
- Live evidence directory: `/var/lib/aios-gitops/acceptance/AIOS-V2R1C2-PRODUCTION-20260722T011836Z`
- Evidence hash index: `/var/lib/aios-gitops/acceptance/AIOS-V2R1C2-PRODUCTION-20260722T011836Z/production-evidence.sha256`
## Production gates
Promotion may proceed only when the 18921 candidate remains healthy, the committed evidence checksum file verifies without drift, the 18920 baseline is healthy, required secret references are available without exposing secret values, and the V2R1C1 rollback service/configuration remains intact.
The obsolete release-plan `DECLARED` rows are not authoritative for this run. The production gate must explicitly select `AIOS-V2R1C2-UNIFIED-PORTAL-TAKEOVER-20260722` and require its three required stages to be `SUCCEEDED`; an unqualified aggregate across stale plan rows fails closed.
## Synthetic WeChat scope
WeChat-facing acceptance for this release used synthetic/mock receipts and transcripts only. It did not impersonate a real user and does not claim a real-user WeChat session. The relevant committed evidence entries are `wechat_simulated_receipts` and `wechat_simulated_transcript`; the production online smoke also records `simulation: true` and `real_user_impersonated: false`.
## Rollback rule
If Health, API, UI, Role Matrix, Tenant Isolation, Runtime, or Audit verification fails after the 18920 switch, stop the V2R1C2 production unit and restart the preserved V2R1C1 18920 service. Do not delete or rewrite either release tag or rollback asset.
## Result
The production switch completed at `2026-07-22T01:40:51Z`. Full acceptance and the production online smoke passed, including Health, API, UI, Role Matrix, Tenant Isolation, Runtime, Audit and the selected takeover release gate. The previous `aios-orchestrator-gateway.service` unit, configuration and Release Run backup remain available for rollback.
@@ -0,0 +1,65 @@
# A0 callback transport and pending-decision recovery
This runbook separates code acceptance from live callback acceptance. Passing
unit tests does not prove that a cloud Workspace Agent can reach the installed
tool endpoint.
## Transport configuration
Choose exactly one mode in the owner-managed remote-tool configuration:
- `remote_https_action`: `endpoint` must be the real public HTTPS Action URL.
- `secure_mcp_tunnel`: `endpoint` must exactly match an entry in
`approved_mcp_tunnel_endpoints`.
The configuration stores only `access_secret_ref` and `callback_secret_ref`.
It must not contain resolved credentials. Startup rejects HTTP, loopback,
unspecified/private literal addresses, URL credentials, query/fragment data,
unapproved tunnel endpoints, and known placeholder markers.
Generate the cloud-facing manifest only after owner configuration exists:
```bash
PYTHONPATH=/opt/aios-orchestrator-gateway/current \
python3 /opt/aios-orchestrator-gateway/current/scripts/generate_orchestrator_tool_manifest.py \
--transport-config /etc/aios-orchestrator-gateway/remote-tool.json \
--output /var/lib/aios-orchestrator-gateway/orchestrator-tool-manifest.json
```
The generated Action manifest exposes only
`submit_orchestration_decision`. The MCP manifest lists the same single tool
with write-side-effect and idempotency annotations.
## Recovery
Plan mode is read-only:
```bash
PYTHONPATH=/opt/aios-orchestrator-gateway/current \
python3 /opt/aios-orchestrator-gateway/current/scripts/recover_pending_decisions.py \
--config /etc/aios-orchestrator-gateway/config.json --plan
```
Application handles at most one leased record and never resends a model prompt
or replays a decision:
```bash
PYTHONPATH=/opt/aios-orchestrator-gateway/current \
python3 /opt/aios-orchestrator-gateway/current/scripts/recover_pending_decisions.py \
--config /etc/aios-orchestrator-gateway/config.json --apply-one
```
## Required live acceptance
Do not mark the live Gate PASS until metadata-only evidence proves, in order:
1. Workspace trigger returned HTTP 202.
2. The remote callback tool received one authenticated request.
3. The bridge returned HTTP 202 from the loopback 18920 decision endpoint.
4. The event reached `DECIDED`, `callback_state=APPLIED`, and
`decision_applied=true`.
5. Recovery planning and one apply cycle did not create a second outbox side
effect for the same event.
Never include Authorization, token values, resolved SecretRefs, prompts, or raw
model output in logs or RESULT files.
-42
View File
@@ -1,42 +0,0 @@
# AIOS Task Plane
该模块是与任务发起方、Worker 类型和通知后端解耦的持久化任务控制平面。当前交付仅用于隔离开发与 Shadow 验收;默认 `execution_enabled=false`,不会修改生产、tag、Baseline 或 Registry。
## 架构
- `TaskPlaneService`:统一服务边界和 HTTP 路由。
- `TaskPlaneStore`:SQLite WAL、revision/CAS、串行事务、租约、hash-chained journal、evidence 和 durable outbox。
- `WorkerRegistry` / `LeaseManager`:身份绑定、capability、并发额度、心跳、过期探测后恢复。
- `DelegationEngine`:父子 DAG、全局深度、capability/risk/approval/budget 单调约束和幂等重放。
- `PolicyEngine`:默认关闭执行;身份、证据、结果摘要和可插拔签名验签均 fail-closed。
- `WorkerAdapter`:OpenClaw、Codex、Hermes、A2A 和 systemd Worker 的统一契约。默认实现只返回 `SHADOW_VERIFIED`,不执行外部命令。
- `DurableOutboxDispatcher`:Webhook、NATS JetStream、GPT callback、Dashboard 和 WeChat 的回执式通知接口。没有真实回执时始终保留 `NOT_DELIVERED`。
## 状态与恢复
正常路径为 `CREATED → QUEUED/CLAIMED → RUNNING → VERIFYING/COMPLETED`;失败路径使用 `FAILED_RECOVERABLE`、`FAILED_FINAL`、`CANCELLED` 或 `MANUAL_REVIEW_REQUIRED`。每次状态事件都进入同一事务中的 append-only hash chain。
租约过期不会自动重放副作用。控制平面必须先调用 Worker probe,只有明确确认原执行已停止,才可将 claim 标记为 `EXPIRED` 并把任务转为 `FAILED_RECOVERABLE`。结果必须匹配 task、claim、Worker identity、未过期 lease 和 operation digest;重复结果返回冲突。
## 安全模型
- task/worker/claim/message ID 使用 allowlist pattern,拒绝 URL 编码差异、路径分隔符和控制字符。
- HTTP body 最大 1 MB,非对象 JSON、异常 Content-Length 和非法参数返回 400/413。
- 内部异常响应不回显路径、payload 或凭据;默认 HTTP access log 不记录原始 request target。
- secret-like 字段在 durable task、event、evidence 和 outbox 写入前脱敏。
- 结果签名存在但没有配置验签器时拒绝;摘要、evidence 或 event 被篡改时验收失败。
- SQLite 文件创建为 0600;新建状态目录使用 0700。
## 验收
```bash
PYTHONPATH=active-runtime \
bash scripts/task-plane/run_task_plane_acceptance.sh
PYTHONPATH=active-runtime \
python3 scripts/task-plane/create_shadow_tasks.py \
--database reports/task-plane-shadow.sqlite3 \
--report reports/task-plane-shadow.json
```
上述脚本只使用本地 SQLite 和 Shadow adapter,不触发真实 OpenClaw、Codex、Hermes、微信或生产执行。
-314
View File
@@ -1,314 +0,0 @@
openapi: 3.0.3
info:
title: AIOS Task Plane API
version: 1.0.0
description: >
Durable control-plane APIs for task orchestration with fail-closed policy.
Every POST operation requires bearerAuth; read-only GET operations may be
exposed according to the surrounding runtime policy.
servers:
- url: /
paths:
/tasks:
get:
summary: List tasks
responses:
"200":
description: Task list
post:
security: [ { bearerAuth: [] } ]
summary: Create task
parameters:
- in: header
name: Idempotency-Key
required: false
schema:
type: string
minLength: 1
maxLength: 128
requestBody:
required: true
content:
application/json:
schema:
$ref: "#/components/schemas/TaskPlan"
responses:
"201": { description: Created }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"409": { description: Conflict }
"413": { description: Payload too large }
"423": { description: Execution or resource locked }
"503": { description: Authentication or service unavailable }
/tasks/{task_id}:
get:
summary: Get task
parameters:
- $ref: "#/components/parameters/TaskId"
responses:
"200": { description: OK }
"404": { description: Not found }
/tasks/{task_id}/claim:
post:
security: [ { bearerAuth: [] } ]
summary: Claim a task
parameters:
- $ref: "#/components/parameters/TaskId"
requestBody:
required: true
content:
application/json:
schema:
$ref: "#/components/schemas/TaskClaimRequest"
responses:
"200": { description: Claimed }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Not found }
"409": { description: Conflict }
"423": { description: Execution disabled }
/tasks/{task_id}/heartbeat:
post:
security: [ { bearerAuth: [] } ]
summary: Renew task claim lease
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Leased }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Task or claim not found }
"409": { description: Claim conflict }
"413": { description: Payload too large }
"423": { description: Lease or execution locked }
"503": { description: Service unavailable }
/tasks/{task_id}/progress:
post:
security: [ { bearerAuth: [] } ]
summary: Report task progress
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Progress saved }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Task or claim not found }
"409": { description: Claim conflict }
"413": { description: Payload too large }
"423": { description: Lease or execution locked }
"503": { description: Service unavailable }
/tasks/{task_id}/result:
post:
security: [ { bearerAuth: [] } ]
summary: Submit task result
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Result accepted }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Result policy rejected }
"404": { description: Task or claim not found }
"409": { description: Duplicate or mismatched result }
"413": { description: Payload too large }
"423": { description: Lease or execution locked }
"503": { description: Verification service unavailable }
/tasks/{task_id}/fail:
post:
security: [ { bearerAuth: [] } ]
summary: Mark task failed
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Failure recorded }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Task or claim not found }
"409": { description: Claim conflict }
"413": { description: Payload too large }
"423": { description: Lease or execution locked }
"503": { description: Service unavailable }
/tasks/{task_id}/cancel:
post:
security: [ { bearerAuth: [] } ]
summary: Cancel task
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Cancelled }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Task not found }
"409": { description: Revision or state conflict }
"413": { description: Payload too large }
"423": { description: Resource locked }
"503": { description: Service unavailable }
/tasks/{task_id}/delegate:
post:
security: [ { bearerAuth: [] } ]
summary: Delegate task
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Delegated }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Parent task not found }
"409": { description: Conflict }
"413": { description: Payload too large }
"423": { description: Execution locked }
"503": { description: Service unavailable }
/workers/register:
post:
security: [ { bearerAuth: [] } ]
summary: Register worker
responses:
"201": { description: Worker registered }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Worker identity mismatch }
"409": { description: Registration conflict }
"413": { description: Payload too large }
"423": { description: Resource locked }
"503": { description: Service unavailable }
/workers/{worker_id}/heartbeat:
post:
security: [ { bearerAuth: [] } ]
summary: Worker heartbeat
parameters: [ $ref: "#/components/parameters/WorkerId" ]
responses:
"200": { description: Heartbeat accepted }
"400": { description: Bad request }
"401": { description: Unauthorized }
"403": { description: Forbidden }
"404": { description: Worker not found }
"409": { description: Worker state conflict }
"413": { description: Payload too large }
"423": { description: Worker locked }
"503": { description: Service unavailable }
/workers:
get:
summary: List workers
responses:
"200": { description: Worker list }
/workers/{worker_id}:
get:
summary: Get worker
parameters: [ $ref: "#/components/parameters/WorkerId" ]
responses:
"200": { description: Worker details }
/tasks/{task_id}/events:
get:
summary: Task events
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Event list }
/tasks/{task_id}/evidence:
get:
summary: Task evidence
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Evidence list }
/tasks/{task_id}/children:
get:
summary: Task child tasks
parameters: [ $ref: "#/components/parameters/TaskId" ]
responses:
"200": { description: Child list }
components:
securitySchemes:
bearerAuth:
type: http
scheme: bearer
bearerFormat: opaque
parameters:
TaskId:
name: task_id
in: path
required: true
schema:
type: string
minLength: 1
maxLength: 128
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
WorkerId:
name: worker_id
in: path
required: true
schema:
type: string
minLength: 1
maxLength: 128
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
schemas:
TaskPlan:
type: object
additionalProperties: false
properties:
schema_version: { type: integer, enum: [1] }
task_id:
type: string
minLength: 1
maxLength: 128
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
task_type:
type: string
minLength: 1
maxLength: 128
pattern: '^[A-Za-z0-9][A-Za-z0-9._-]{0,127}$'
issuer: { type: object }
intent: { type: string, maxLength: 16384 }
required_capabilities:
type: array
maxItems: 128
items: { type: string }
preferred_workers:
type: array
maxItems: 64
items: { type: string }
risk: { type: string, enum: [low, medium, high] }
approval_policy:
type: string
maxLength: 128
dependencies:
type: array
maxItems: 128
items: { type: string }
max_delegation_depth: { type: integer, minimum: 0, maximum: 32 }
budgets:
type: object
additionalProperties: false
properties:
time_seconds: { type: integer, minimum: 0 }
attempts: { type: integer, minimum: 0 }
tokens: { type: integer, minimum: 0 }
cost: { type: number, minimum: 0 }
evidence_requirements:
type: array
maxItems: 128
items: { type: string }
parent_task_id:
type: string
maxLength: 128
correlation_id:
type: string
maxLength: 128
metadata:
type: object
TaskClaimRequest:
type: object
additionalProperties: false
required: [worker_id, actor_identity]
properties:
worker_id:
type: string
maxLength: 128
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
lease_seconds:
type: integer
minimum: 1
maximum: 3600
actor_identity:
type: string
minLength: 1
maxLength: 512
@@ -0,0 +1,228 @@
openapi: 3.0.3
info:
title: AIOS Orchestrator Gateway metadata API
version: 0.2.0
servers:
- url: http://127.0.0.1:18920
paths:
/v1/orchestrator/central/jobs/claim:
post:
summary: Claim one remote Codex decision job
security:
- bearerAuth: []
parameters:
- $ref: '#/components/parameters/RunnerId'
- $ref: '#/components/parameters/IdempotencyKey'
requestBody:
required: true
content:
application/json:
schema:
type: object
additionalProperties: false
required: [contract_commit]
properties:
contract_commit:
type: string
pattern: '^[0-9a-f]{40}$'
responses:
'200':
description: Job lease or no available job
'401':
description: Runner or token scope rejected
'409':
description: Contract or idempotency conflict
/v1/orchestrator/central/jobs/{job_id}/heartbeat:
post:
summary: Extend an active remote Codex job lease
security:
- bearerAuth: []
parameters:
- $ref: '#/components/parameters/JobId'
- $ref: '#/components/parameters/RunnerId'
- $ref: '#/components/parameters/IdempotencyKey'
requestBody:
$ref: '#/components/requestBodies/RemoteLeaseRequest'
responses:
'202': {description: Lease extended}
'401': {description: Runner, scope, or lease credential rejected}
'409': {description: Lease is stale or owned by another runner}
/v1/orchestrator/central/jobs/{job_id}/decision:
post:
summary: Submit one strict decision for a leased job
security:
- bearerAuth: []
parameters:
- $ref: '#/components/parameters/JobId'
- $ref: '#/components/parameters/RunnerId'
- $ref: '#/components/parameters/IdempotencyKey'
requestBody:
required: true
content:
application/json:
schema:
type: object
additionalProperties: false
required: [lease_token, claim_idempotency_key, decision]
properties:
lease_token: {type: string, pattern: '^[0-9a-f]{64}$'}
claim_idempotency_key: {type: string}
decision: {$ref: '#/components/schemas/Decision'}
responses:
'202': {description: Decision accepted and applied exactly once}
'401': {description: Runner, scope, or lease credential rejected}
'409': {description: State, contract, lease, or idempotency conflict}
'422': {description: Decision schema or allowed work-unit violation}
/v1/orchestrator/central/jobs/{job_id}/fail:
post:
summary: Fail a leased job into retry or dead-letter
security:
- bearerAuth: []
parameters:
- $ref: '#/components/parameters/JobId'
- $ref: '#/components/parameters/RunnerId'
- $ref: '#/components/parameters/IdempotencyKey'
requestBody:
required: true
content:
application/json:
schema:
type: object
additionalProperties: false
required: [lease_token, claim_idempotency_key, category]
properties:
lease_token: {type: string, pattern: '^[0-9a-f]{64}$'}
claim_idempotency_key: {type: string}
category: {type: string, pattern: '^[A-Za-z0-9_]{1,64}$'}
responses:
'202': {description: Failure recorded}
'401': {description: Runner, scope, or lease credential rejected}
'409': {description: Lease or idempotency conflict}
/v1/orchestrator/events:
post:
security:
- bearerAuth: []
requestBody:
required: true
content:
application/json:
schema:
type: object
responses:
'202':
description: Event accepted
'409':
description: Idempotency or state conflict
'422':
description: Invalid event
/v1/orchestrator/events/{event_id}:
get:
security:
- bearerAuth: []
parameters:
- in: path
name: event_id
required: true
schema:
type: string
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
responses:
'200':
description: Metadata-only event status
content:
application/json:
schema:
type: object
additionalProperties: false
required: [event_id, status, trigger_http_status, trigger_error_category, trigger_channel, remote_id, callback_state, callback_deadline, callback_attempts, decision_applied, state_version_after]
properties:
event_id:
type: string
status:
enum: [RECEIVED, AWAITING_DECISION, DECIDED, TRIGGER_FAILED, DEAD_LETTER]
trigger_channel:
type: string
nullable: true
enum: [workspace_agent, responses_api, codex_local, remote_codex]
trigger_http_status:
type: integer
nullable: true
minimum: 100
maximum: 599
trigger_error_category:
type: string
nullable: true
enum: [UNAUTHORIZED, FORBIDDEN, NOT_FOUND, NOT_RUNNABLE, NETWORK_ERROR]
remote_id:
type: string
nullable: true
callback_state:
enum: [PENDING, SENT, AWAITING_CALLBACK, CALLBACK_RECEIVED, APPLIED, STALE_SUPERSEDED, DEAD_LETTER]
callback_deadline:
type: integer
nullable: true
minimum: 0
callback_attempts:
type: integer
minimum: 0
decision_applied:
type: boolean
state_version_after:
type: integer
nullable: true
minimum: 0
'401':
description: Missing or invalid bearer
'404':
description: Event not found
components:
parameters:
JobId:
in: path
name: job_id
required: true
schema:
type: string
pattern: '^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$'
RunnerId:
in: header
name: X-AIOS-Runner-ID
required: true
schema: {type: string, maxLength: 128}
IdempotencyKey:
in: header
name: Idempotency-Key
required: true
schema: {type: string, minLength: 16, maxLength: 128}
requestBodies:
RemoteLeaseRequest:
required: true
content:
application/json:
schema:
type: object
additionalProperties: false
required: [lease_token, claim_idempotency_key]
properties:
lease_token: {type: string, pattern: '^[0-9a-f]{64}$'}
claim_idempotency_key: {type: string}
schemas:
Decision:
type: object
additionalProperties: false
required: [schema_version, project_key, contract_commit, event_id, expected_state_version, verdict, state_updates, next_work_units, requires_human_input, stop_reason]
properties:
schema_version: {type: integer, enum: [1]}
project_key: {type: string}
contract_commit: {type: string, pattern: '^[0-9a-f]{40}$'}
event_id: {type: string}
expected_state_version: {type: integer, minimum: 0}
verdict: {type: string, enum: [BLOCKED, COMPLETE, DISPATCH, NOOP]}
state_updates: {type: array, items: {type: object}}
next_work_units: {type: array, items: {type: object}}
requires_human_input: {type: boolean}
stop_reason: {type: string, nullable: true}
securitySchemes:
bearerAuth:
type: http
scheme: bearer
-153
View File
@@ -1,153 +0,0 @@
#!/usr/bin/env bash
set -Eeuo pipefail
VERSION=V2R1C2
EXPECTED_TARGET=3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24
TAG_OBJECT=a3bd84ebec9c705f7c9c5c03f7f4ec55f8854e36
REPO=/data/aios-gitops/repo.git
SOURCE=/opt/aios-v2r1c2
TAKEOVER=/data/aios-gitops/worktrees/codex/AIOS-V2R1C2-UNIFIED-PORTAL-TAKEOVER-20260722
CANDIDATE_SERVICE=aios-v2r1c2-portal.service
OLD_SERVICE=aios-orchestrator-gateway.service
NEW_SERVICE=aios-v2r1c2-production.service
CANDIDATE_STATE=/var/lib/aios-v2r1c2-portal
PRODUCTION_STATE=/var/lib/aios-v2r1c2-production
RUN_ID=${AIOS_RELEASE_RUN_ID:-AIOS-V2R1C2-PRODUCTION-20260722T011836Z}
EVIDENCE_DIR=/var/lib/aios-gitops/acceptance/$RUN_ID
SCRIPT_DIR=$(cd -- "$(dirname -- "${BASH_SOURCE[0]}")" && pwd)
RELEASE_ROOT=$(cd -- "$SCRIPT_DIR/../.." && pwd)
UNIT_SOURCE=$RELEASE_ROOT/deploy/aios-v2/systemd/aios-v2r1c2-production.service
ENV_SOURCE=$RELEASE_ROOT/deploy/aios-v2/systemd/aios-v2r1c2-production.env.example
LOCK=/run/lock/aios-v2r1c2-production-release.lock
require_root() {
test "$(id -u)" -eq 0 || { echo 'root required' >&2; exit 77; }
}
health_old() {
curl -fsS --max-time 10 http://127.0.0.1:18920/healthz
}
health_new() {
curl -fsS --max-time 10 http://127.0.0.1:18920/api/v2/health
}
health_candidate() {
curl -fsS --max-time 10 http://127.0.0.1:18921/api/v2/health
}
wait_candidate() {
for _ in $(seq 1 30); do
if health_candidate; then
return 0
fi
sleep 1
done
return 1
}
preflight() {
require_root
test "$(git --git-dir="$REPO" rev-parse refs/tags/$VERSION)" = "$TAG_OBJECT"
test "$(git --git-dir="$REPO" rev-list -n 1 "$VERSION")" = "$EXPECTED_TARGET"
test -f "$SOURCE/active-runtime/app/aios_web/release_application.py"
test -f "$UNIT_SOURCE"
test -f "$ENV_SOURCE"
command -v sqlite3 >/dev/null
systemctl is-active --quiet "$CANDIDATE_SERVICE"
systemctl is-active --quiet "$OLD_SERVICE"
health_candidate >/dev/null
health_old >/dev/null
(
cd "$TAKEOVER/artifacts/v2r1c2"
sha256sum -c checksums.sha256 >/dev/null
)
diff -qr --exclude=__pycache__ "$TAKEOVER/active-runtime/app" "$SOURCE/active-runtime/app" >/dev/null
diff -qr "$TAKEOVER/plans" "$SOURCE/plans" >/dev/null
echo PREFLIGHT=PASS
}
prepare() {
preflight
install -d -m 0750 "$EVIDENCE_DIR" "$EVIDENCE_DIR/backup"
health_old >"$EVIDENCE_DIR/production-health-before.json"
health_candidate >"$EVIDENCE_DIR/candidate-health-before.json"
systemctl cat "$OLD_SERVICE" >"$EVIDENCE_DIR/backup/$OLD_SERVICE.unit"
cp -a /etc/aios-orchestrator-gateway "$EVIDENCE_DIR/backup/"
cp -a /etc/systemd/system/aios-orchestrator-gateway.service "$EVIDENCE_DIR/backup/"
if test -e "$PRODUCTION_STATE"; then
mv "$PRODUCTION_STATE" "$EVIDENCE_DIR/backup/aios-v2r1c2-production-state.preexisting"
fi
systemctl stop "$CANDIDATE_SERVICE"
trap 'systemctl start "$CANDIDATE_SERVICE" || true' EXIT
while IFS= read -r -d '' database; do
sqlite3 "$database" 'PRAGMA wal_checkpoint(TRUNCATE);' >/dev/null
test "$(sqlite3 "$database" 'PRAGMA quick_check;')" = ok
done < <(find "$CANDIDATE_STATE" -maxdepth 1 -type f -name '*.sqlite3' -print0)
install -d -o aios -g aios -m 0750 "$PRODUCTION_STATE"
cp -a "$CANDIDATE_STATE/." "$PRODUCTION_STATE/"
chown -R aios:aios "$PRODUCTION_STATE"
while IFS= read -r -d '' database; do
test "$(sqlite3 "$database" 'PRAGMA quick_check;')" = ok
done < <(find "$PRODUCTION_STATE" -maxdepth 1 -type f -name '*.sqlite3' -print0)
install -o root -g root -m 0644 "$UNIT_SOURCE" /etc/systemd/system/aios-v2r1c2-production.service
install -d -o root -g aios -m 0750 /etc/aios-v2
install -o root -g aios -m 0640 "$ENV_SOURCE" /etc/aios-v2/aios-v2r1c2-production.env
systemctl daemon-reload
systemctl start "$CANDIDATE_SERVICE"
wait_candidate >"$EVIDENCE_DIR/candidate-health-after-state-copy.json"
trap - EXIT
echo PREPARE=PASS
}
rollback() {
require_root
systemctl stop "$NEW_SERVICE" || true
systemctl disable "$NEW_SERVICE" >/dev/null 2>&1 || true
systemctl enable "$OLD_SERVICE" >/dev/null 2>&1 || true
systemctl start "$OLD_SERVICE"
health_old >"$EVIDENCE_DIR/rollback-health.json"
echo ROLLBACK=PASS
}
promote() {
require_root
test -f /etc/systemd/system/aios-v2r1c2-production.service
test -f /etc/aios-v2/aios-v2r1c2-production.env
! systemctl is-active --quiet "$NEW_SERVICE"
systemctl is-active --quiet "$CANDIDATE_SERVICE"
health_candidate >/dev/null
health_old >"$EVIDENCE_DIR/production-health-switch-input.json"
trap 'trap - ERR; rollback' ERR
systemctl stop "$OLD_SERVICE"
systemctl start "$NEW_SERVICE"
for _ in $(seq 1 30); do
if health_new >"$EVIDENCE_DIR/production-health-after-switch.json" 2>/dev/null; then
echo PROMOTE=PASS
trap - ERR
return 0
fi
sleep 1
done
return 1
}
finalize() {
require_root
systemctl is-active --quiet "$NEW_SERVICE"
health_new >"$EVIDENCE_DIR/production-health-final.json"
systemctl disable "$OLD_SERVICE" >/dev/null
systemctl enable "$NEW_SERVICE" >/dev/null
echo FINALIZE=PASS
}
exec 9>"$LOCK"
flock -n 9 || { echo 'release operation already running' >&2; exit 75; }
case "${1:-}" in
preflight) preflight ;;
prepare) prepare ;;
promote) promote ;;
rollback) rollback ;;
finalize) finalize ;;
*) echo "usage: $0 {preflight|prepare|promote|rollback|finalize}" >&2; exit 64 ;;
esac
-69
View File
@@ -1,69 +0,0 @@
#!/usr/bin/env bash
set -Eeuo pipefail
RUN_ID=AIOS-V2R1C2-PRODUCTION-20260722T011836Z
CONTROL=/data/aios-gitops/worktrees/controller/ops/aios-v2/v2r1c2-production-control.sh
REPO=/data/aios-gitops/repo.git
EVIDENCE_DIR=/var/lib/aios-gitops/acceptance/$RUN_ID
DRIVER=/data/aios-gitops/worktrees/controller/tools/run_online_smoke_18920.py
TEMP_SOURCE=
cleanup_temp() {
case "$TEMP_SOURCE" in
/tmp/aios-v2r1c2-production-acceptance.*)
test -d "$TEMP_SOURCE" && rm -rf -- "$TEMP_SOURCE"
;;
'') ;;
*) echo "refusing to remove unexpected temporary path: $TEMP_SOURCE" >&2; return 1 ;;
esac
}
rollback_on_error() {
code=$?
trap - ERR
AIOS_RELEASE_RUN_ID="$RUN_ID" "$CONTROL" rollback || true
cleanup_temp || true
exit "$code"
}
trap rollback_on_error ERR
AIOS_RELEASE_RUN_ID="$RUN_ID" "$CONTROL" promote
systemctl is-active --quiet aios-v2r1c2-production.service
! systemctl is-active --quiet aios-orchestrator-gateway.service
curl -fsS --max-time 10 http://127.0.0.1:18920/api/v2/health >"$EVIDENCE_DIR/health-after-promote.json"
curl -fsS --max-time 10 http://127.0.0.1:18921/api/v2/health >"$EVIDENCE_DIR/candidate-health-during-production.json"
TEMP_SOURCE=$(mktemp -d /tmp/aios-v2r1c2-production-acceptance.XXXXXX)
git --git-dir="$REPO" archive 8cdca463aff9707edee00e20a3b56884130f6978 | tar -x -C "$TEMP_SOURCE"
(
cd "$TEMP_SOURCE"
python3 tools/aios_v2r1c2_acceptance.py \
--source-oid 3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24 \
--online \
--base-url http://127.0.0.1:18920 \
--evidence-output "$EVIDENCE_DIR/production-acceptance.json"
)
python3 "$DRIVER" "$TEMP_SOURCE" "$EVIDENCE_DIR/production-online-smoke.json"
python3 - "$EVIDENCE_DIR/production-acceptance.json" "$EVIDENCE_DIR/production-online-smoke.json" <<'PY'
import json
import sys
for path in sys.argv[1:]:
with open(path, encoding='utf-8') as handle:
data = json.load(handle)
if data.get('passed') is not True:
raise SystemExit(f'acceptance failed: {path}')
PY
systemctl show aios-v2r1c2-production.service \
-p ActiveState -p SubState -p MainPID -p NRestarts \
>"$EVIDENCE_DIR/production-service-status.txt"
sha256sum "$EVIDENCE_DIR"/production-acceptance.json \
"$EVIDENCE_DIR"/production-online-smoke.json \
"$EVIDENCE_DIR"/health-after-promote.json \
>"$EVIDENCE_DIR/production-evidence.sha256"
cleanup_temp
TEMP_SOURCE=
trap - ERR
echo SWITCH_VALIDATED=PASS
@@ -1,15 +0,0 @@
{
"patch_name": "gitea-tag-publication-atomicity",
"branch": "patch/gitea-tag-publication-atomicity-20260723",
"base": "main",
"risk": "high",
"description": "Fail-closed, resumable Gitea release-tag publication with exact remote ref, peeled target, API and listing verification before baseline update.",
"owner_approval": "received",
"constraints": {
"force_push": false,
"tag_rewrite": false,
"direct_main_push": false,
"secret_commit": false,
"production_runtime_anchor_change": false
}
}
@@ -1,9 +0,0 @@
{
"id": "patch-1784043007770-e97206c3",
"patch_name": "gitops-isolated-worktree-governance-20260714T232948UTC8",
"branch": "patch/gitops-isolated-worktree-governance-20260714T232948UTC8",
"risk": "high",
"description": "Root-cause remediation for recurring ACL and shared-worktree failures: unified aiman:gitops runtime identity, UMask=0002, inherited default ACLs, per-task isolated Git worktrees, startup preflight, migration/rollback scripts, systemd drop-in templates, audit evidence, and an OpenClaw execution task. No production business deployment in this patch.",
"created_by": "chatgpt-gitops-bot",
"created_at": "2026-07-14T15:30:07Z"
}
-10
View File
@@ -1,10 +0,0 @@
{
"id": "nl-1784681185786-19ac0f82",
"text": "Create a safe GitOps plan for tag V2R1C2. The tag target must be commit 3dd622b5a515f1fa4ef72b9f8ae9ce48ac703f24. Evidence closure commit is 8cdca463aff9707edee00e20a3b56884130f6978. Do not tag the current workspace HEAD. Do not modify main. Do not force push. Do not promote production. If the target commit cannot be bound explicitly, keep the operation blocked.",
"intent": "unknown",
"action_plan": [],
"execute": false,
"owner_approval_required": true,
"created_by": "chatgpt-gitops-bot",
"created_at": "2026-07-22T00:46:25Z"
}
@@ -1,9 +0,0 @@
{
"type": "branch",
"branch": "patch/aios-agent-task-control-plane-20260724",
"base": "main",
"created_by": "agent-task-plane-writer",
"created_at": "2026-07-24T02:20:56Z",
"direct_main_push": false,
"owner_approval_required": true
}
@@ -1,9 +0,0 @@
{
"type": "branch",
"branch": "patch/gitea-tag-publication-atomicity-20260723",
"base": "main",
"created_by": "owner-gitea-tag-atomicity-20260723",
"created_at": "2026-07-23T04:59:13Z",
"direct_main_push": false,
"owner_approval_required": true
}
@@ -1,9 +0,0 @@
{
"type": "branch",
"branch": "task/codex/gitops-isolated-worktree-migration-20260715T001652UTC8",
"base": "patch/gitops-isolated-worktree-governance-20260714T232948UTC8",
"created_by": "chatgpt-gitops-bot",
"created_at": "2026-07-14T16:17:28Z",
"direct_main_push": false,
"owner_approval_required": true
}
+459
View File
@@ -0,0 +1,459 @@
#!/usr/bin/env python3
from __future__ import annotations
import argparse
from dataclasses import dataclass
import json
import logging
import os
from pathlib import Path
import signal
import stat
import subprocess
import tempfile
import threading
import time
from typing import Any
import urllib.error
import urllib.request
import uuid
from services.aios_orchestrator_gateway.errors import ConfigurationError, TriggerError, ValidationError
from services.aios_orchestrator_gateway.models import DECISION_SCHEMA, Event
from services.aios_orchestrator_gateway.remote_codex import validate_remote_decision
from services.aios_orchestrator_gateway.securefs import atomic_write_json, read_secure_bytes, read_secure_json
LOG = logging.getLogger("aios-remote-codex-brain")
REQUIRED_EXEC_FLAGS = (
"--json", "--output-schema", "--output-last-message", "--sandbox",
"--config", "--cd", "--ephemeral",
)
APPROVED_CODEX_ROOTS = (
Path("/usr/lib/node_modules/@openai/codex"),
Path("/usr/local/lib/node_modules/@openai/codex"),
Path("/opt/openai/codex"),
)
EXPECTED_CODEX_VERSION = "0.144.6"
@dataclass(frozen=True)
class RunnerConfig:
gateway_url: str
runner_id: str
token_file: Path
contract_commit: str
codex_binary: Path
codex_home: Path
controller_readonly: Path
state_file: Path
timeout_seconds: int
poll_seconds: int
heartbeat_seconds: int
@classmethod
def from_file(cls, path: Path) -> "RunnerConfig":
raw = read_secure_json(path, expected_uid=0)
expected = {
"gateway_url", "runner_id", "token_file", "contract_commit",
"codex_binary", "codex_home", "controller_readonly", "state_file",
"timeout_seconds", "poll_seconds", "heartbeat_seconds",
}
if set(raw) != expected:
raise ConfigurationError("invalid_runner_config_fields")
def absolute(name: str) -> Path:
value = raw.get(name)
if not isinstance(value, str) or not Path(value).is_absolute():
raise ConfigurationError(f"invalid_{name}")
return Path(value)
url = raw.get("gateway_url")
if url != "http://127.0.0.1:28920/v1/orchestrator":
raise ConfigurationError("gateway_must_use_approved_loopback_tunnel")
runner_id = raw.get("runner_id")
contract = raw.get("contract_commit")
if not isinstance(runner_id, str) or not runner_id or len(runner_id) > 128:
raise ConfigurationError("invalid_runner_id")
if not isinstance(contract, str) or len(contract) != 40 or any(c not in "0123456789abcdef" for c in contract):
raise ConfigurationError("invalid_contract_commit")
def bounded(name: str, low: int, high: int) -> int:
value = raw.get(name)
if not isinstance(value, int) or not low <= value <= high:
raise ConfigurationError(f"invalid_{name}")
return value
return cls(
gateway_url=url, runner_id=runner_id, token_file=absolute("token_file"),
contract_commit=contract, codex_binary=absolute("codex_binary"),
codex_home=absolute("codex_home"),
controller_readonly=absolute("controller_readonly"),
state_file=absolute("state_file"),
timeout_seconds=bounded("timeout_seconds", 30, 3600),
poll_seconds=bounded("poll_seconds", 1, 300),
heartbeat_seconds=bounded("heartbeat_seconds", 5, 300),
)
class RemoteCodexRunner:
def __init__(self, config: RunnerConfig):
self.config = config
self._stop = threading.Event()
def _environment(self) -> dict[str, str]:
return {
"CODEX_HOME": str(self.config.codex_home),
"HOME": str(self.config.codex_home.parent),
"PATH": f"{self.config.codex_binary.parent}:/usr/bin:/bin",
"LANG": "C.UTF-8", "LC_ALL": "C.UTF-8",
}
@staticmethod
def _safe_regular(path: Path, *, owner_uid: int, mode: int | None = None) -> os.stat_result:
info = path.lstat()
if stat.S_ISLNK(info.st_mode) or not stat.S_ISREG(info.st_mode):
raise ConfigurationError("unsafe_regular_file")
if info.st_uid != owner_uid:
raise ConfigurationError("unsafe_file_owner")
if mode is not None and stat.S_IMODE(info.st_mode) != mode:
raise ConfigurationError("unsafe_file_mode")
return info
def validate_official_cli(self) -> None:
binary = self.config.codex_binary
if not binary.is_absolute() or not binary.exists():
raise ConfigurationError("codex_cli_not_installed")
link_info = binary.lstat()
if link_info.st_uid != 0:
raise ConfigurationError("unsafe_codex_launcher")
if not stat.S_ISLNK(link_info.st_mode) and link_info.st_mode & (stat.S_IWGRP | stat.S_IWOTH):
raise ConfigurationError("unsafe_codex_launcher")
target = binary.resolve(strict=True)
target_info = target.lstat()
if not stat.S_ISREG(target_info.st_mode) or target_info.st_uid != 0:
raise ConfigurationError("unsafe_codex_target")
if target_info.st_mode & (stat.S_IWGRP | stat.S_IWOTH) or not os.access(target, os.X_OK):
raise ConfigurationError("unsafe_codex_target")
approved = next((root for root in APPROVED_CODEX_ROOTS if target.is_relative_to(root)), None)
if approved is None:
raise ConfigurationError("unapproved_codex_install_root")
if approved.name == "codex" and "node_modules" in approved.parts:
package = approved / "package.json"
value = read_secure_json(package, expected_uid=0)
if value.get("name") != "@openai/codex" or value.get("version") != EXPECTED_CODEX_VERSION:
raise ConfigurationError("unofficial_codex_package")
native = list(approved.glob("node_modules/@openai/codex-*/vendor/*/bin/codex"))
if len(native) != 1:
raise ConfigurationError("codex_native_binary_missing")
native_info = native[0].lstat()
if (
not stat.S_ISREG(native_info.st_mode) or native_info.st_uid != 0
or native_info.st_mode & (stat.S_IWGRP | stat.S_IWOTH)
or not os.access(native[0], os.X_OK)
):
raise ConfigurationError("unsafe_codex_native_binary")
result = self._run_quiet([str(binary), "exec", "--help"], 30, capture_stdout=True)
if result.returncode != 0 or any(flag not in result.stdout for flag in REQUIRED_EXEC_FLAGS):
raise ConfigurationError("codex_cli_version_incompatible")
def validate_auth_metadata(self) -> None:
auth = self.config.codex_home / "auth.json"
self._safe_regular(auth, owner_uid=os.geteuid(), mode=0o600)
def login_status(self) -> None:
result = self._run_quiet(
[str(self.config.codex_binary), "login", "status"], 30, capture_stdout=True
)
if result.returncode != 0:
raise ConfigurationError("codex_authentication_required")
normalized = result.stdout.lower()
if "oneapi" in normalized or "api key" in normalized or "chatgpt" not in normalized:
raise ConfigurationError("codex_chatgpt_authentication_required")
def validate_runtime_boundary(self) -> None:
controller = self.config.controller_readonly
if not controller.is_dir() or controller.is_symlink() or os.access(controller, os.W_OK):
raise ConfigurationError("controller_not_read_only")
if (controller / ".git").exists():
raise ConfigurationError("controller_must_not_be_git_worktree")
state_parent = self.config.state_file.parent
state_parent.mkdir(parents=True, exist_ok=True, mode=0o700)
if state_parent.is_symlink():
raise ConfigurationError("unsafe_state_directory")
def health_check(self) -> None:
self.validate_official_cli()
self.validate_auth_metadata()
self.validate_runtime_boundary()
self.login_status()
def _run_quiet(self, argv: list[str], timeout: int, *, capture_stdout: bool = False) -> subprocess.CompletedProcess[str]:
try:
return subprocess.run(
argv, cwd=self.config.controller_readonly, env=self._environment(),
stdin=subprocess.DEVNULL,
stdout=subprocess.PIPE if capture_stdout else subprocess.DEVNULL,
stderr=subprocess.STDOUT if capture_stdout else subprocess.DEVNULL,
text=True, timeout=timeout, check=False,
)
except (OSError, subprocess.SubprocessError) as exc:
raise ConfigurationError("codex_cli_probe_failed") from exc
def _token(self) -> str:
info = self._safe_regular(self.config.token_file, owner_uid=os.geteuid(), mode=0o600)
if info.st_size > 4096:
raise ConfigurationError("runner_token_too_large")
raw = read_secure_bytes(self.config.token_file, expected_uid=os.geteuid(), max_bytes=4096)
if not raw or b"\n" in raw or b"\r" in raw:
raise ConfigurationError("invalid_runner_token")
return raw.decode("utf-8")
def _request(self, path: str, scope: str, body: dict[str, Any], key: str) -> dict[str, Any]:
encoded = json.dumps(body, separators=(",", ":"), sort_keys=True).encode()
request = urllib.request.Request(
f"{self.config.gateway_url}/{path.lstrip('/')}", data=encoded, method="POST",
headers={
"Authorization": f"Bearer {self._token()}",
"Content-Type": "application/json",
"Idempotency-Key": key,
"X-AIOS-Runner-ID": self.config.runner_id,
},
)
try:
with urllib.request.urlopen(request, timeout=30) as response:
if response.status not in {200, 202}:
raise TriggerError("gateway_request_rejected")
raw = response.read(1_048_577)
except (urllib.error.URLError, TimeoutError, OSError) as exc:
raise TriggerError(f"gateway_{scope}_unavailable") from exc
if len(raw) > 1_048_576:
raise TriggerError("gateway_response_too_large")
try:
value = json.loads(raw)
except json.JSONDecodeError as exc:
raise TriggerError("gateway_invalid_response") from exc
if not isinstance(value, dict):
raise TriggerError("gateway_invalid_response")
return value
def _load_state(self) -> dict[str, Any] | None:
try:
return read_secure_json(self.config.state_file, expected_uid=os.geteuid())
except FileNotFoundError:
return None
def _save_state(self, value: dict[str, Any]) -> None:
atomic_write_json(self.config.state_file, value, mode=0o600)
def _clear_state(self) -> None:
try:
self.config.state_file.unlink()
directory_fd = os.open(self.config.state_file.parent, os.O_RDONLY | os.O_DIRECTORY)
try:
os.fsync(directory_fd)
finally:
os.close(directory_fd)
except FileNotFoundError:
pass
def _claim(self) -> dict[str, Any] | None:
claim_key = "claim-" + uuid.uuid4().hex
response = self._request(
"central/jobs/claim", "claim",
{"contract_commit": self.config.contract_commit}, claim_key,
)
if response.get("claimed") is False:
return None
required = {
"claimed", "job_id", "runner_id", "lease_version", "lease_until",
"lease_token", "event", "claim_idempotency_key",
}
if set(response) != required or response.get("runner_id") != self.config.runner_id:
raise TriggerError("invalid_claim_response")
event = Event.from_dict(response.get("event"))
if event.contract_commit != self.config.contract_commit:
raise ValidationError("contract_version_mismatch")
state = {
"phase": "CLAIMED", "job_id": response["job_id"],
"lease_token": response["lease_token"], "claim_idempotency_key": claim_key,
"event": event.to_dict(), "decision": None,
}
self._save_state(state)
return state
def _heartbeat(self, state: dict[str, Any], key: str | None = None) -> dict[str, Any]:
return self._request(
f"central/jobs/{state['job_id']}/heartbeat", "heartbeat",
{"lease_token": state["lease_token"],
"claim_idempotency_key": state["claim_idempotency_key"]},
key or ("heartbeat-" + uuid.uuid4().hex),
)
def _exec(self, event: Event, heartbeat_state: dict[str, Any]) -> dict[str, Any]:
prompt = (
"读取只读治理合同与本次事件。只输出符合Decision Schema的决策。"
"不得修改Git、部署生产、读取秘密或伪造证据;证据不足时输出BLOCKED。\n"
+ event.canonical_json()
)
runtime_root = self.config.state_file.parent / "runtime"
runtime_root.mkdir(parents=True, exist_ok=True, mode=0o700)
os.chmod(runtime_root, 0o700)
with tempfile.TemporaryDirectory(prefix="remote-decision-", dir=runtime_root) as directory:
root = Path(directory)
schema_path = root / "decision.schema.json"
output_path = root / "decision.json"
schema_path.write_text(json.dumps(DECISION_SCHEMA, separators=(",", ":")), encoding="utf-8")
schema_path.chmod(0o600)
argv = [
str(self.config.codex_binary), "exec", "--json",
"--output-schema", str(schema_path),
"--output-last-message", str(output_path),
"--sandbox", "read-only", "--config", 'approval_policy="never"',
"--cd", str(self.config.controller_readonly), "--ephemeral", "-",
]
process = subprocess.Popen(
argv, cwd=self.config.controller_readonly, env=self._environment(),
stdin=subprocess.PIPE, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL,
text=True, start_new_session=True,
)
heartbeat_error: list[Exception] = []
heartbeat_stop = threading.Event()
def heartbeat_loop() -> None:
while not heartbeat_stop.wait(self.config.heartbeat_seconds):
try:
self._heartbeat(heartbeat_state)
except Exception as exc:
heartbeat_error.append(exc)
return
thread = threading.Thread(target=heartbeat_loop, daemon=True)
thread.start()
try:
process.communicate(prompt, timeout=self.config.timeout_seconds)
except subprocess.TimeoutExpired as exc:
self._terminate_group(process)
raise TriggerError("codex_exec_timeout") from exc
finally:
heartbeat_stop.set(); thread.join(timeout=5)
if heartbeat_error:
raise TriggerError("lease_heartbeat_failed")
if process.returncode != 0:
raise TriggerError("codex_exec_nonzero")
try:
if output_path.stat().st_size > 1_048_576:
raise TriggerError("codex_output_too_large")
value = json.loads(output_path.read_text(encoding="utf-8"))
except (OSError, json.JSONDecodeError) as exc:
raise TriggerError("codex_exec_invalid_json") from exc
if not isinstance(value, dict):
raise TriggerError("codex_exec_invalid_json")
return validate_remote_decision(value, event).to_dict()
@staticmethod
def _terminate_group(process: subprocess.Popen[str]) -> None:
try:
os.killpg(process.pid, signal.SIGTERM)
process.wait(timeout=2)
except (ProcessLookupError, subprocess.TimeoutExpired):
try:
os.killpg(process.pid, signal.SIGKILL)
except ProcessLookupError:
pass
process.wait(timeout=5)
def _submit(self, state: dict[str, Any]) -> None:
key = state.get("decision_idempotency_key")
if not isinstance(key, str):
key = "decision-" + uuid.uuid4().hex
state["decision_idempotency_key"] = key
self._save_state(state)
response = self._request(
f"central/jobs/{state['job_id']}/decision", "decision",
{"lease_token": state["lease_token"],
"claim_idempotency_key": state["claim_idempotency_key"],
"decision": state["decision"]}, key,
)
if response.get("accepted") is not True or response.get("completed") is not True:
raise TriggerError("decision_not_confirmed")
self._clear_state()
def _fail(self, state: dict[str, Any], category: str) -> None:
key = state.get("fail_idempotency_key")
if not isinstance(key, str):
key = "fail-" + uuid.uuid4().hex
state["fail_idempotency_key"] = key
self._save_state(state)
self._request(
f"central/jobs/{state['job_id']}/fail", "fail",
{"lease_token": state["lease_token"],
"claim_idempotency_key": state["claim_idempotency_key"],
"category": category}, key,
)
self._clear_state()
def run_one(self) -> bool:
self.health_check()
state = self._load_state()
if state is None:
state = self._claim()
if state is None:
return False
try:
self._heartbeat(state, state.get("recovery_heartbeat_key") or "recover-" + uuid.uuid4().hex)
if state.get("phase") == "CLAIMED":
event = Event.from_dict(state.get("event"))
decision = self._exec(event, state)
state["phase"] = "DECIDED"
state["decision"] = decision
self._save_state(state)
if state.get("phase") != "DECIDED" or not isinstance(state.get("decision"), dict):
raise TriggerError("invalid_runner_state")
self._submit(state)
return True
except (ConfigurationError, TriggerError, ValidationError) as exc:
category = type(exc).__name__.upper()
try:
self._fail(state, category)
except Exception:
LOG.warning("remote_job_retained category=TRANSPORT_UNAVAILABLE")
raise
def run_forever(self) -> None:
while not self._stop.is_set():
try:
worked = self.run_one()
except Exception as exc:
LOG.warning("remote_cycle_failed category=%s", type(exc).__name__)
worked = False
self._stop.wait(0 if worked else self.config.poll_seconds)
def main() -> int:
parser = argparse.ArgumentParser()
parser.add_argument("--config", required=True, type=Path)
parser.add_argument("--once", action="store_true")
parser.add_argument("--health", action="store_true")
args = parser.parse_args()
logging.basicConfig(level=logging.INFO, format="%(levelname)s %(name)s %(message)s")
try:
runner = RemoteCodexRunner(RunnerConfig.from_file(args.config))
if args.health:
runner.health_check()
print("REMOTE_CODEX_BRAIN_HEALTH=PASS")
return 0
if args.once:
runner.run_one()
return 0
runner.run_forever()
return 0
except Exception as exc:
reason = str(exc)
safe = reason if reason.replace("_", "").isalnum() and len(reason) <= 64 else type(exc).__name__
print(f"REMOTE_CODEX_BRAIN=BLOCKED reason={safe}")
return 1
if __name__ == "__main__":
raise SystemExit(main())
+155
View File
@@ -0,0 +1,155 @@
#!/usr/bin/env bash
set -Eeuo pipefail
umask 0077
SOURCE_ROOT="${SOURCE_ROOT:-$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)}"
CONTRACT_COMMIT="${CONTRACT_COMMIT:-cbe33ecab8e9794ba5f318bea43aeaf447a56fbc}"
INSTALL_ROOT=/opt/aios-remote-codex-brain
CONFIG_ROOT=/etc/aios-remote-codex-brain
STATE_ROOT=/var/lib/aios-remote-codex-brain
SERVICE_USER=aios-remote-codex
CODEX_BINARY=/usr/bin/codex
CODEX_HOME="$STATE_ROOT/.codex"
CONTROLLER="$INSTALL_ROOT/controller-readonly"
RUNNER_ID="codex-brain-192-168-88-6"
EXPECTED_CODEX_VERSION=0.144.6
fail() { printf 'STATUS=%s\nAUTOMATION_CLOSED=NO\n' "$1" >&2; exit 2; }
[[ "$(id -u)" -eq 0 ]] || fail OWNER_INSTALL_REQUIRED
for command in python3 node npm runuser systemctl systemd-analyze ssh sha256sum; do
command -v "$command" >/dev/null 2>&1 || fail "MISSING_COMMAND_${command}"
done
id "$SERVICE_USER" >/dev/null 2>&1 || useradd --system --home-dir "$STATE_ROOT" --shell /usr/sbin/nologin "$SERVICE_USER"
install -d -o root -g root -m 0755 "$INSTALL_ROOT" "$INSTALL_ROOT/releases"
install -d -o root -g "$SERVICE_USER" -m 0750 "$CONFIG_ROOT"
install -d -o "$SERVICE_USER" -g "$SERVICE_USER" -m 0700 "$STATE_ROOT" "$CODEX_HOME" "$STATE_ROOT/runtime"
[[ -e "$CODEX_BINARY" ]] || fail BLOCKED_CODEX_CLI_NOT_INSTALLED
resolved="$(readlink -f "$CODEX_BINARY")"
case "$resolved" in
/usr/lib/node_modules/@openai/codex/*|/usr/local/lib/node_modules/@openai/codex/*|/opt/openai/codex/*) ;;
*) fail BLOCKED_UNOFFICIAL_CODEX_CLI ;;
esac
python3 - "$CODEX_BINARY" "$resolved" <<'PY' || exit 2
import os,stat,sys
launcher=os.lstat(sys.argv[1]); target=os.lstat(sys.argv[2])
if launcher.st_uid != 0: raise SystemExit(2)
if not stat.S_ISLNK(launcher.st_mode) and launcher.st_mode & (stat.S_IWGRP|stat.S_IWOTH): raise SystemExit(2)
if target.st_uid != 0 or target.st_mode & (stat.S_IWGRP|stat.S_IWOTH): raise SystemExit(2)
if not stat.S_ISREG(target.st_mode) or not os.access(sys.argv[2],os.X_OK): raise SystemExit(2)
PY
package_root="$(npm root -g)/@openai/codex"
[[ -f "$package_root/package.json" && ! -L "$package_root/package.json" ]] || fail BLOCKED_UNOFFICIAL_CODEX_CLI
[[ "$(node -e "console.log(require('$package_root/package.json').name)")" = '@openai/codex' ]] || fail BLOCKED_UNOFFICIAL_CODEX_CLI
[[ "$(node -e "console.log(require('$package_root/package.json').version)")" = "$EXPECTED_CODEX_VERSION" ]] || fail BLOCKED_CODEX_CLI_VERSION_INCOMPATIBLE
mapfile -d '' native_targets < <(find "$package_root/node_modules/@openai" -path '*/vendor/*/bin/codex' -type f -perm /111 -print0)
[[ "${#native_targets[@]}" -eq 1 ]] || fail BLOCKED_UNOFFICIAL_CODEX_CLI
python3 - "${native_targets[0]}" <<'PY' || exit 2
import os,stat,sys
s=os.lstat(sys.argv[1])
if not stat.S_ISREG(s.st_mode) or s.st_uid != 0 or s.st_mode & (stat.S_IWGRP|stat.S_IWOTH) or not os.access(sys.argv[1],os.X_OK): raise SystemExit(2)
PY
help_file="$(mktemp)"; trap 'rm -f "$help_file"' EXIT
runuser -u "$SERVICE_USER" -- env -i HOME="$STATE_ROOT" CODEX_HOME="$CODEX_HOME" PATH=/usr/bin:/bin \
"$CODEX_BINARY" exec --help >"$help_file" 2>&1 || fail BLOCKED_CODEX_CLI_VERSION_INCOMPATIBLE
for flag in --json --output-schema --output-last-message --sandbox --config --cd --ephemeral; do
grep -Fq -- "$flag" "$help_file" || fail BLOCKED_CODEX_CLI_VERSION_INCOMPATIBLE
done
rm -f "$help_file"; trap - EXIT
AUTH_JSON="$CODEX_HOME/auth.json"
[[ -e "$AUTH_JSON" ]] || fail BLOCKED_REMOTE_CODEX_AUTHENTICATION_REQUIRED
python3 - "$AUTH_JSON" "$SERVICE_USER" <<'PY' || exit 2
import os,pwd,stat,sys
s=os.lstat(sys.argv[1]); uid=pwd.getpwnam(sys.argv[2]).pw_uid
if stat.S_ISLNK(s.st_mode) or not stat.S_ISREG(s.st_mode) or s.st_uid != uid or stat.S_IMODE(s.st_mode) != 0o600: raise SystemExit(2)
PY
status_file="$(mktemp)"; trap 'rm -f "$status_file"' EXIT
runuser -u "$SERVICE_USER" -- env -i HOME="$STATE_ROOT" CODEX_HOME="$CODEX_HOME" PATH=/usr/bin:/bin \
"$CODEX_BINARY" login status >"$status_file" 2>&1 || fail BLOCKED_REMOTE_CODEX_AUTHENTICATION_REQUIRED
grep -Eqi 'chatgpt' "$status_file" || fail BLOCKED_REMOTE_CODEX_AUTHENTICATION_REQUIRED
grep -Eqi 'one[ -]?api|api key' "$status_file" && fail BLOCKED_REMOTE_CODEX_AUTHENTICATION_REQUIRED
rm -f "$status_file"; trap - EXIT
for required in \
"$CONFIG_ROOT/runner.token" \
"$CONFIG_ROOT/tunnel_ed25519" \
"$CONFIG_ROOT/known_hosts"; do
[[ -f "$required" && ! -L "$required" ]] || fail BLOCKED_REMOTE_CODEX_TUNNEL_UNAVAILABLE
chown "$SERVICE_USER:$SERVICE_USER" "$required"
chmod 0600 "$required"
done
for required in \
remote-runner/aios_remote_codex_brain.py \
remote-runner/verify_remote_codex_brain.sh \
services/aios_orchestrator_gateway/__init__.py \
systemd/aios-remote-codex-brain.service \
systemd/aios-remote-codex-brain-health.service \
systemd/aios-remote-codex-brain-health.timer \
systemd/aios-remote-codex-tunnel.service \
.agent/policies/V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT.yaml; do
[[ -f "$SOURCE_ROOT/$required" && ! -L "$SOURCE_ROOT/$required" ]] || fail INVALID_SOURCE
done
digest="$(find "$SOURCE_ROOT/remote-runner" "$SOURCE_ROOT/services/aios_orchestrator_gateway" "$SOURCE_ROOT/systemd" -type f -print0 | sort -z | xargs -0 sha256sum | sha256sum | awk '{print $1}')"
release="$INSTALL_ROOT/releases/$digest"
if [[ ! -e "$release" ]]; then
install -d -o root -g root -m 0755 "$release" "$release/remote-runner" "$release/services" "$release/systemd"
cp -a "$SOURCE_ROOT/remote-runner/." "$release/remote-runner/"
cp -a "$SOURCE_ROOT/services/aios_orchestrator_gateway" "$release/services/"
cp -a "$SOURCE_ROOT/systemd/." "$release/systemd/"
find "$release" -type d -exec chmod 0755 {} +
find "$release" -type f -exec chmod 0644 {} +
chmod 0755 "$release/remote-runner/"*.py "$release/remote-runner/"*.sh
chown -R root:root "$release"
fi
ln -sfn "$release" "$INSTALL_ROOT/.current-$digest"
mv -Tf "$INSTALL_ROOT/.current-$digest" "$INSTALL_ROOT/current"
rm -rf "$CONTROLLER"
install -d -o root -g root -m 0555 "$CONTROLLER/.agent/policies"
install -o root -g root -m 0444 \
"$SOURCE_ROOT/.agent/policies/V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT.yaml" \
"$CONTROLLER/.agent/policies/V2R1-PERSONAL-WEIXIN-PRODUCTION-EXECUTION-CONTRACT.yaml"
cat >"$CONFIG_ROOT/config.json" <<EOF
{
"gateway_url":"http://127.0.0.1:28920/v1/orchestrator",
"runner_id":"$RUNNER_ID",
"token_file":"$CONFIG_ROOT/runner.token",
"contract_commit":"$CONTRACT_COMMIT",
"codex_binary":"$CODEX_BINARY",
"codex_home":"$CODEX_HOME",
"controller_readonly":"$CONTROLLER",
"state_file":"$STATE_ROOT/runner-state.json",
"timeout_seconds":300,
"poll_seconds":5,
"heartbeat_seconds":30
}
EOF
chown root:"$SERVICE_USER" "$CONFIG_ROOT/config.json"
chmod 0640 "$CONFIG_ROOT/config.json"
cat >"$CONFIG_ROOT/tunnel.env" <<'EOF'
AIOS_GATEWAY_SSH_HOST=hw.imruyi.top
AIOS_GATEWAY_SSH_PORT=22
AIOS_GATEWAY_SSH_USER=aios-remote-codex-tunnel
EOF
chown root:"$SERVICE_USER" "$CONFIG_ROOT/tunnel.env"
chmod 0640 "$CONFIG_ROOT/tunnel.env"
for unit in aios-remote-codex-brain.service aios-remote-codex-brain-health.service aios-remote-codex-brain-health.timer aios-remote-codex-tunnel.service; do
install -o root -g root -m 0644 "$SOURCE_ROOT/systemd/$unit" "/etc/systemd/system/$unit"
done
systemd-analyze verify \
/etc/systemd/system/aios-remote-codex-brain.service \
/etc/systemd/system/aios-remote-codex-brain-health.service \
/etc/systemd/system/aios-remote-codex-brain-health.timer \
/etc/systemd/system/aios-remote-codex-tunnel.service
systemctl daemon-reload
systemctl enable --now aios-remote-codex-tunnel.service
systemctl enable --now aios-remote-codex-brain.service aios-remote-codex-brain-health.timer
printf 'REMOTE_CODEX_BRAIN_INSTALL=PASS\nRELEASE_DIGEST=%s\n' "$digest"
+20
View File
@@ -0,0 +1,20 @@
#!/usr/bin/env bash
set -euo pipefail
umask 0077
CONFIG=/etc/aios-remote-codex-brain/config.json
CURRENT=/opt/aios-remote-codex-brain/current
[[ -f "$CONFIG" && ! -L "$CONFIG" ]] || { echo 'REMOTE_CODEX_VERIFY=BLOCKED'; exit 1; }
[[ -L "$CURRENT" ]] || { echo 'REMOTE_CODEX_VERIFY=BLOCKED'; exit 1; }
resolved="$(readlink -f "$CURRENT")"
case "$resolved" in /opt/aios-remote-codex-brain/releases/*) ;; *) echo 'REMOTE_CODEX_VERIFY=BLOCKED'; exit 1;; esac
systemctl is-active --quiet aios-remote-codex-tunnel.service
systemctl is-active --quiet aios-remote-codex-brain.service
systemctl is-enabled --quiet aios-remote-codex-brain-health.timer
runuser -u aios-remote-codex -- env -i \
HOME=/var/lib/aios-remote-codex-brain \
CODEX_HOME=/var/lib/aios-remote-codex-brain/.codex \
PATH=/usr/bin:/bin \
PYTHONPATH="$CURRENT" \
/usr/bin/python3 "$CURRENT/remote-runner/aios_remote_codex_brain.py" --config "$CONFIG" --health
printf 'REMOTE_CODEX_VERIFY=PASS\n'
@@ -1,43 +0,0 @@
{
"base": "main",
"base_oid": "b739fbb5ec8f7096d2f5ade87f65a7c74b0ee870",
"commit_count": 10,
"commits": [
"fc0b9f29bf6dadfbcc8599da6f0ab11a6e19b529",
"3812e1ab0123c7da4726e51179a10746b5bbde83",
"3196d91b12d0ac5d7ea550b0474703a08736471b",
"54093d15437ce0af31acf6feeeffacf6626a59cd",
"27a0ae126b96d63c40544174e7a3859e30eba056",
"83ab67f7e11eae9da4546cab53169f5a6a47e0ae",
"f4b6ac8609faf947d2659b2c32564a26fcb3eb10",
"a3edc628a91a4af3d702aacefe0389f844615bc1",
"e40c05779c435c3726618de81e02e90953d727f3",
"edfad05d57fb3d23d8ab19d27bb6a43a7f2a298f"
],
"file_count": 15,
"files": [
"active-runtime/app/aios_chatgpt_actions_api.py",
"active-runtime/app/aios_gitops_write_guard.py",
"active-runtime/app/gitea_tag_publication.py",
"docs/gitops/gitea-tag-publication-atomicity.md",
"docs/gitops/openapi-chatgpt-actions.yaml",
"patches/gitea-tag-publication-atomicity-20260723.json",
"patches/patch_gitea-tag-publication-atomicity-20260723.branch.json",
"scripts/generate_delivery_manifest.py",
"scripts/release_version.sh",
"scripts/verify_gitea_credential_rotation.py",
"security/incidents/GITEA-TOKEN-33-EXPOSURE-20260723.redacted.json",
"tests/run_live_gitea_fixture.sh",
"tests/test_chatgpt_actions_tag.py",
"tests/test_gitea_tag_publication.py",
"tests/test_gitea_tag_publication_live.py"
],
"generated_from": {
"commit_range": "b739fbb5ec8f7096d2f5ade87f65a7c74b0ee870..edfad05d57fb3d23d8ab19d27bb6a43a7f2a298f",
"commits_command": "git log --reverse --format=%H b739fbb5ec8f7096d2f5ade87f65a7c74b0ee870..edfad05d57fb3d23d8ab19d27bb6a43a7f2a298f",
"files_command": "git diff --name-only b739fbb5ec8f7096d2f5ade87f65a7c74b0ee870..edfad05d57fb3d23d8ab19d27bb6a43a7f2a298f"
},
"manifest_path": "reports/gitea-tag-publication-atomicity-delivery.json",
"source_branch": "patch/gitea-tag-publication-atomicity-20260723",
"source_commit": "edfad05d57fb3d23d8ab19d27bb6a43a7f2a298f"
}
-10
View File
@@ -1,10 +0,0 @@
{
"execution_status": "PARTIALLY_COMPLETED",
"tests_run": 19,
"failures": 0,
"errors": 0,
"skipped": 0,
"multi_agent_shadow_verified": true,
"production_side_effects": false,
"notification_delivery_status": "NOT_DELIVERED"
}
-53
View File
@@ -1,53 +0,0 @@
{
"root_task_id": "AIOS-AGENT-TASK-CONTROL-PLANE-20260724-SHADOW",
"execution_mode": "shadow_adapter",
"production_side_effects": false,
"tasks": [
{
"task_id": "AIOS-TASK-PLANE-OPENCLAW-ACCEPTANCE",
"parent_task_id": "AIOS-AGENT-TASK-CONTROL-PLANE-20260724-SHADOW",
"task_status": "COMPLETED",
"worker_result": {
"worker": "openclaw",
"task_id": "AIOS-TASK-PLANE-OPENCLAW-ACCEPTANCE",
"status": "SHADOW_VERIFIED",
"claim_id": "claim-093cbf4c-11d5-47a3-97dc-a373a8a53940",
"execution_mode": "shadow",
"production_side_effects": false
},
"result_digest": "b3a8ed4e0a60a0a9364d1eea1a6d30a403da1fb5ff79ff38a489a52227f4116e"
},
{
"task_id": "AIOS-TASK-PLANE-CODEX-REVIEW",
"parent_task_id": "AIOS-AGENT-TASK-CONTROL-PLANE-20260724-SHADOW",
"task_status": "COMPLETED",
"worker_result": {
"worker": "codex",
"task_id": "AIOS-TASK-PLANE-CODEX-REVIEW",
"status": "SHADOW_VERIFIED",
"claim_id": "claim-3cfe4031-27c7-409c-9841-cd5c0d26218d",
"execution_mode": "shadow",
"production_side_effects": false
},
"result_digest": "65fb825869b9b32d923b387796839e57b9ca7053716720861988ceb8d7e9c93a"
},
{
"task_id": "AIOS-TASK-PLANE-HERMES-ARCHITECTURE",
"parent_task_id": "AIOS-AGENT-TASK-CONTROL-PLANE-20260724-SHADOW",
"task_status": "COMPLETED",
"worker_result": {
"worker": "hermes",
"task_id": "AIOS-TASK-PLANE-HERMES-ARCHITECTURE",
"status": "SHADOW_VERIFIED",
"claim_id": "claim-fd7ba481-ab71-47dc-8e91-3655cbef1c40",
"execution_mode": "shadow",
"production_side_effects": false
},
"result_digest": "09c9ce563424b6490b7be1e854d06de786bd243a48b099137a7a2e9cb7035265"
}
],
"event_chains_valid": true,
"notification_delivery_status": [
"NOT_DELIVERED"
]
}
-1
View File
@@ -1 +0,0 @@
{"schema_version":1,"task_id":"AIOS-AGENT-TASK-CONTROL-PLANE-20260724","writer_status":"READY","production_side_effects":false}
@@ -0,0 +1,61 @@
{
"$schema": "https://json-schema.org/draft/2020-12/schema",
"$id": "https://aios.local/schemas/automation-quorum-controller.schema.json",
"title": "AIOS GPT plus one-executor quorum controller configuration",
"type": "object",
"additionalProperties": false,
"required": [
"project_key",
"contract_commit",
"database_path",
"file_inbox_path",
"health_receipts",
"blocker_retry_threshold",
"max_health_age_seconds",
"lease_ttl_seconds",
"gateway_event_url",
"gateway_token_file"
],
"properties": {
"$schema": {"type": "string"},
"project_key": {"type": "string", "pattern": "^[A-Za-z0-9][A-Za-z0-9._:-]{0,127}$"},
"contract_commit": {"type": "string", "pattern": "^[0-9a-f]{40}$"},
"database_path": {"type": "string", "pattern": "^/"},
"file_inbox_path": {"type": "string", "pattern": "^/"},
"default_workload": {
"enum": [
"large_repetitive_infrastructure_monitoring",
"complex_code_security_architecture"
]
},
"health_receipts": {
"type": "object",
"additionalProperties": false,
"required": ["central_gpt", "openclaw", "codex"],
"properties": {
"central_gpt": {"type": "string", "pattern": "^/"},
"openclaw": {"type": "string", "pattern": "^/"},
"codex": {"type": "string", "pattern": "^/"}
}
},
"git_ref_polling": {
"type": "object",
"additionalProperties": false,
"required": ["repository", "refs"],
"properties": {
"repository": {"type": "string", "pattern": "^/"},
"refs": {
"type": "array",
"minItems": 1,
"uniqueItems": true,
"items": {"type": "string", "pattern": "^refs/"}
}
}
},
"blocker_retry_threshold": {"type": "integer", "minimum": 1, "maximum": 20},
"max_health_age_seconds": {"type": "integer", "minimum": 5, "maximum": 300},
"lease_ttl_seconds": {"type": "integer", "minimum": 15, "maximum": 3600},
"gateway_event_url": {"type": "string", "const": "http://127.0.0.1:18920/v1/orchestrator/events"},
"gateway_token_file": {"type": "string", "pattern": "^/"}
}
}

Some files were not shown because too many files have changed in this diff Show More