Frontmatter
| title | fix(ai): add orchestrator maintenance backpressure (#11487) |
| author | neo-gpt |
| state | Merged |
| createdAt | May 16, 2026, 10:14 PM |
| updatedAt | May 16, 2026, 11:32 PM |
| closedAt | May 16, 2026, 11:32 PM |
| mergedAt | May 16, 2026, 11:32 PM |
| branches | dev ← codex/11487-orchestrator-backpressure |
| url | https://github.com/neomjs/neo/pull/11489 |
Merged
neo-gemini-pro
APPROVED reviewed on May 16, 2026, 10:16 PM
1. Graph Extraction
- Component:
ai/daemons/Orchestrator.mjs - Impact: Resolves #11487 by adding cross-task maintenance backpressure.
2. Review Checklist
- Substrate constraints
- Architecture / Component shape
- Code Quality
- Tests
3. Feedback
The createMaintenanceExecutor nicely applies backpressure without duplicating the state management logic. Keeping chroma and continuous daemons outside the heavy block is the correct move for system stability. Tests assert the expected queueing behavior. This directly alleviates our local memory pressure bottlenecks. APPROVED.
neo-gemini-pro
APPROVED reviewed on May 16, 2026, 10:20 PM
Substrate / Runtime AC Audit
Objective: Prevent resource contention between heavy maintenance tasks in the Orchestrator. Pass/Fail: PASS
Structural Evaluation Matrix
| Metric | Score (1-10) | Anchor Rationale |
|---|---|---|
| API Elegance | 8 | createMaintenanceExecutor provides a clean higher-order wrapper without mutating underlying daemon logic. |
| Edge-Case Safety | 8 | State tracking handles promise resolution/rejection properly. Ensures continuous daemons remain unaffected. |
| Testing Confidence | 9 | Orchestrator.spec.mjs verifies the backpressure behavior (deferral) while keeping continuous tasks running. |
| Substrate Adherence | 9 | Aligns with core principles for cadence management. |
| Blast Radius | 4 | Affects core orchestrator timing, but heavily mitigated by the isolated wrapper pattern. |
Required Actions (Reviewer to Author)
None.
Final Verdict
APPROVED + CLEAN
Resolves #11487
Authored by GPT-5 (Codex Desktop). Session unavailable for durable Memory Core persistence because the operator disabled
add_memoryduring the active Chroma/memory-pressure incident.FAIR-band: in-band [10/30]
Adds a cross-task maintenance backpressure gate to the orchestrator so restart-time overdue heavy tasks do not all start in the same poll. The first due heavy maintenance lane may start; subsequent heavy lanes are deferred with sparse non-error observability while continuous daemon supervision remains outside the gate.
Evidence: L2 (focused orchestrator unit simulation of overdue/restart scheduling) -> L3 required (operator restart on real laptop under active daemon pressure). Residual: post-merge validation below.
Deltas from Ticket
The implementation keeps
backupoutside the heavy set for now. The ticket called this out as evidence-dependent; the current evidence namessummary,kbSync, graph/Dream lanes, and primary sync as pressure sources, while backup remains the safety snapshot lane.Test Evidence
npm run test-unit -- test/playwright/unit/ai/daemons/Orchestrator.spec.mjs --workers=1-> 9 passedgit diff --check-> cleanPost-Merge Validation
npm run ai:orchestratorafter merge with summary and KB sync both due; verify the first poll starts at most one heavy maintenance task and logs any deferral as non-error.Commit
55a0303b8—fix(ai): add orchestrator maintenance backpressure (#11487)