Skip to content

fix: persist pipeline state to disk and restore on server restart#1585

Merged
OneStepAt4time merged 1 commit intodevelopfrom
fix/1424-pipeline-persist
Apr 10, 2026
Merged

fix: persist pipeline state to disk and restore on server restart#1585
OneStepAt4time merged 1 commit intodevelopfrom
fix/1424-pipeline-persist

Conversation

@OneStepAt4time
Copy link
Copy Markdown
Owner

@OneStepAt4time OneStepAt4time commented Apr 10, 2026

Summary

  • Fixes [enhancement] E4-3: Pipeline state not persisted — server restart discards orchestrations #1424 — Pipeline state was persisted to pipelines.json but the implementation had three bugs that made it ineffective on server restart
  • Stale file leak: persistPipelines() returned early when no running pipelines remained, leaving a stale file that hydrate() would restore as completed/failed entries on next startup
  • Missing persist after pipeline completion: advancePipeline() could transition a pipeline to completed/failed without calling persistPipelines(), so the file still showed running status
  • Fragile constructor: stateDir was set via hydrate() instead of the constructor, creating an ordering dependency with startup

Changes

File Change
src/pipeline.ts Delete state file when no running pipelines; persist after advancePipeline transitions; use static unlink import
src/server.ts Pass config.stateDir to PipelineManager constructor
src/__tests__/pipeline.test.ts Add 10 persistence tests (persist, hydrate, orphan detection, corrupt/missing file, null dir, non-existent dir)

Test plan

  • npx tsc --noEmit — passes
  • npm run build — passes
  • npm test — 2635 passed, 0 failed
  • Manual: create pipeline → verify pipelines.json written → simulate restart → verify hydration restores state

Aegis version

Developed with: v0.3.2-alpha

Generated by Hephaestus (Aegis dev agent)

@OneStepAt4time
fix: persist pipeline state to disk and restore on server restart (#1424
1a57135 
)

- Delete pipelines.json when no running pipelines remain (was leaking
  stale file, causing hydrate() to restore completed/failed entries)
- Persist after advancePipeline transitions pipeline to completed/failed
  (was only persisting at stage-level transitions)
- Pass stateDir to PipelineManager constructor instead of relying on
  hydrate() to set it (removes fragile ordering dependency)
- Replace dynamic import('node:fs/promises').unlink with static import
- Add 10 tests covering persist, hydrate, orphan detection, and edge cases

Closes #1424

Generated by Hephaestus (Aegis dev agent)
@OneStepAt4time
Copy link
Copy Markdown
Owner Author

OneStepAt4time commented Apr 10, 2026

🔧 PR #1585 ready for review: fix(pipeline): persist pipeline state across restarts (#1424). CI CLEAN. Please review.

aegis-gh-agent[bot]
aegis-gh-agent bot approved these changes Apr 10, 2026
View reviewed changes
Copy link
Copy Markdown
Contributor

@aegis-gh-agent aegis-gh-agent bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM. CI green. Approved by Argus.

@OneStepAt4time OneStepAt4time merged commit e6277d8 into develop Apr 10, 2026
9 checks passed
@OneStepAt4time OneStepAt4time deleted the fix/1424-pipeline-persist branch April 10, 2026 07:06
@OneStepAt4time OneStepAt4time mentioned this pull request Apr 10, 2026
Closed
OneStepAt4time pushed a commit that referenced this pull request Apr 10, 2026
docs: remove Consensus Review and Model Router from documentation (PR #…
1a29064 
…1583)

Remove all documentation references to deleted features (PR #1583).

REMOVED FEATURES:
- Consensus Review endpoints (/v1/sessions/:id/consensus, /v1/consensus/:id)
- Model Router config and tiered routing

FILES CHANGED:
- README.md: Remove consensus/model router from features list
- docs/advanced.md: Remove Consensus Review section
- docs/api-reference.md: Remove consensus endpoints; remove consensus.completed SSE event
- docs/architecture.md: Remove consensus.ts and model-router.ts from module overview
- docs/getting-started.md: Update advanced features reference
- docs/enterprise.md: Remove modelRouter from config example
- docs/enterprise/01-architecture.md: Remove consensus/model-router; mark findings RESOLVED
- docs/enterprise/03-testing-observability.md: Remove model-router.ts from thin-tests list
- docs/enterprise/05-enterprise-roadmap.md: Remove E4-1 (consensus); update M-E4 scope
- docs/enterprise/index.md: Remove consensus reliability finding

RESOLVED IN v0.3.3:
- P-3 (PipelineManager persistence): PR #1585
- P-1 (Pipeline stage timeout): PR #1606
- CON-1 (Consensus): PR #1583 (feature removed)

NOTE: openapi.yaml still contains /consensus endpoints — separate update needed.
This was referenced Apr 10, 2026
Closed
Closed
OneStepAt4time pushed a commit that referenced this pull request Apr 12, 2026
docs: remove Consensus Review and Model Router from documentation (PR #…
7a6cfc2 
…1583)

Remove all documentation references to deleted features (PR #1583).

REMOVED FEATURES:
- Consensus Review endpoints (/v1/sessions/:id/consensus, /v1/consensus/:id)
- Model Router config and tiered routing

FILES CHANGED:
- README.md: Remove consensus/model router from features list
- docs/advanced.md: Remove Consensus Review section
- docs/api-reference.md: Remove consensus endpoints; remove consensus.completed SSE event
- docs/architecture.md: Remove consensus.ts and model-router.ts from module overview
- docs/getting-started.md: Update advanced features reference
- docs/enterprise.md: Remove modelRouter from config example
- docs/enterprise/01-architecture.md: Remove consensus/model-router; mark findings RESOLVED
- docs/enterprise/03-testing-observability.md: Remove model-router.ts from thin-tests list
- docs/enterprise/05-enterprise-roadmap.md: Remove E4-1 (consensus); update M-E4 scope
- docs/enterprise/index.md: Remove consensus reliability finding

RESOLVED IN v0.3.3:
- P-3 (PipelineManager persistence): PR #1585
- P-1 (Pipeline stage timeout): PR #1606
- CON-1 (Consensus): PR #1583 (feature removed)

NOTE: openapi.yaml still contains /consensus endpoints — separate update needed.
@OneStepAt4time OneStepAt4time mentioned this pull request Apr 12, 2026
Merged
OneStepAt4time added a commit that referenced this pull request Apr 12, 2026
@OneStepAt4time
docs: remove Consensus Review and Model Router from documentation (PR #…
b4d4179 
…1583) (#1709)

Remove all documentation references to deleted features (PR #1583).

REMOVED FEATURES:
- Consensus Review endpoints (/v1/sessions/:id/consensus, /v1/consensus/:id)
- Model Router config and tiered routing

FILES CHANGED:
- README.md: Remove consensus/model router from features list
- docs/advanced.md: Remove Consensus Review section
- docs/api-reference.md: Remove consensus endpoints; remove consensus.completed SSE event
- docs/architecture.md: Remove consensus.ts and model-router.ts from module overview
- docs/getting-started.md: Update advanced features reference
- docs/enterprise.md: Remove modelRouter from config example
- docs/enterprise/01-architecture.md: Remove consensus/model-router; mark findings RESOLVED
- docs/enterprise/03-testing-observability.md: Remove model-router.ts from thin-tests list
- docs/enterprise/05-enterprise-roadmap.md: Remove E4-1 (consensus); update M-E4 scope
- docs/enterprise/index.md: Remove consensus reliability finding

RESOLVED IN v0.3.3:
- P-3 (PipelineManager persistence): PR #1585
- P-1 (Pipeline stage timeout): PR #1606
- CON-1 (Consensus): PR #1583 (feature removed)

NOTE: openapi.yaml still contains /consensus endpoints — separate update needed.

Co-authored-by: Argus <argus@openclaw.ai>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@aegis-gh-agent aegis-gh-agent[bot] aegis-gh-agent[bot] approved these changes

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@OneStepAt4time