PR #71971: fix(doctor): exclude trajectory sidecars from orphan transcript sweep (#71960)

• Repository: openclaw/openclaw
• Author: Sanjays2402
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/71971

Description (problem / solution / changelog)

Fixes #71960.

What

isPrimarySessionTranscriptFileName matches any *.jsonl that isn't a compaction checkpoint or archive artifact. Trajectory sidecars (<sessionId>.trajectory.jsonl) therefore look like primary session transcripts to doctor's orphan sweep, which only cross-references sessions.json. Sidecars are referenced via the sibling *.trajectory-path.json pointer instead, so an active session's sidecar gets reported as Found 1 orphan transcript file and offered for archival on every openclaw doctor run.

Fix

Add isTrajectorySidecarFileName (matches *.trajectory.jsonl) and short-circuit it from isPrimarySessionTranscriptFileName. Mirrors the existing isCompactionCheckpointTranscriptFileName carve-out.

Side effect (intentional)

The same classifier feeds src/config/sessions/disk-budget.ts's removable-file queue. Trajectory sidecars no longer end up there either — which is correct: sidecars belong to live sessions and should be cleaned up via the regular session lifecycle, not the budget sweep.

isUsageCountedSessionTranscriptFileName and parseUsageCountedSessionIdFromFileName also exclude sidecars now, which matches the prior intent (sidecars don't have user messages and shouldn't be billed as session transcripts).

Tests

• src/config/sessions/artifacts.test.ts: assert sidecar names are excluded from the primary classifier and that the new helper recognizes them; existing tests stay green.
• src/commands/doctor-state-integrity.test.ts: regression — drop <id>.trajectory.jsonl + <id>.trajectory-path.json into the sessions dir, run doctor, assert neither the orphan warning nor the archive prompt fires and the file survives.

pnpm test src/config/sessions src/commands/doctor-state-integrity.test.ts -- --run
# Test Files: 16 passed   Tests: 125 passed

Reporter's three suggested directions were "exclude *.trajectory.jsonl from primary orphan-transcript checks" / "include sidecars referenced through *.trajectory-path.json" / "soften the warning text". Excluding from the primary classifier is the smallest correct change and naturally covers both detection sites (doctor + disk-budget).

Changed files

• src/commands/doctor-state-integrity.test.ts (modified, +25/-0)
• src/config/sessions/artifacts.test.ts (modified, +16/-0)
• src/config/sessions/artifacts.ts (modified, +15/-0)

PR #72047: fix(doctor): exclude trajectory sidecar files from orphan transcript detection [AI-assisted]

• Repository: openclaw/openclaw
• Author: luyao618
• State: closed | merged: False
• Link: https://github.com/openclaw/openclaw/pull/72047

Description (problem / solution / changelog)

🤖 AI-assisted (built with Claude Code via Hermes orchestration). Test level: fully tested. Prompt summary available on request.

Summary

• Problem: openclaw doctor state-integrity check falsely flags active *.trajectory.jsonl sidecar files as orphan transcripts, suggesting they can be safely archived while the session is still active.
• Why it matters: Users may archive active trajectory data, potentially losing session trajectory history for a running session.
• What changed: Added isTrajectoryJournalFileName() helper in artifacts.ts and excluded trajectory journal files from isPrimarySessionTranscriptFileName(), so they are never considered primary transcripts and never enter the orphan detection path.
• What did NOT change (scope boundary): The orphan transcript detection logic itself is unchanged. Only the file classification predicate was tightened. No changes to trajectory recording, session management, or other doctor checks.

Change Type (select all)

• Bug fix

Scope (select all touched areas)

• CLI commands (src/commands/)
• Session config (src/config/sessions/)

Linked Issue/PR

• Closes #71960
• This PR fixes a bug or regression

Root Cause

• Root cause: isPrimarySessionTranscriptFileName() matched any *.jsonl file that wasn't a compaction checkpoint or archive artifact. Trajectory sidecar files (*.trajectory.jsonl) are a distinct artifact type but were not excluded, so they passed the primary transcript filter. Meanwhile, referencedTranscriptPaths only contained paths from sessions.json entries — not trajectory sidecar paths referenced via *.trajectory-path.json.
• Missing detection / guardrail: No test verified that trajectory sidecar files would not be classified as primary transcripts.
• Contributing context: Trajectory sidecars were added after the orphan transcript detection, and the classification predicate was not updated to account for the new file type.

Regression Test Plan

• Coverage level that should have caught this:
  • Unit test
• Target test or file: src/commands/doctor-state-integrity.test.ts, src/config/sessions/artifacts.test.ts
• Scenario the test should lock in: A *.trajectory.jsonl file in the sessions directory must NOT be reported as an orphan transcript by doctor state-integrity.
• Why this is the smallest reliable guardrail: Direct unit test on the classification helper + integration test on the doctor check — no runtime dependencies needed.
• Existing test that already covers this: N/A

User-visible / Behavior Changes

openclaw doctor no longer warns about active trajectory sidecar files as orphan transcripts.

Diagram (if applicable)

N/A

Security Impact (required)

• New permissions/capabilities? No
• This fix tightens file classification to exclude trajectory sidecars from orphan detection. No new data exposure, no security-sensitive paths touched.

Changed files

• src/commands/doctor-state-integrity.test.ts (modified, +14/-0)
• src/config/sessions/artifacts.test.ts (modified, +10/-0)
• src/config/sessions/artifacts.ts (modified, +7/-0)