fix(noema): add price-attested provider fallback - #1410
Conversation
|
Warning Review limit reachedNext included review available in 21 minutes. View limit detailsLimit details: You’ve used the included review currently available. You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository. Review configuration: ⚙️ Run configurationConfiguration used: Organization UI Review profile: CHILL Plan: Pro Plus Run ID: 📒 Files selected for processing (3)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
| OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }} | ||
| OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }} | ||
| CONTEXTUAL_ORCHESTRATOR_REQUIRE_ZDR: ${{ steps.target_visibility.outputs.require_zdr }} | ||
| CONTEXTUAL_ORCHESTRATOR_POOL: auto |
There was a problem hiding this comment.
🟡 Noema reviews consume paid models
When free providers fail, CONTEXTUAL_ORCHESTRATOR_POOL admits priced routes for required Noema reviews. These reviews can incur unapproved provider charges.
Prompt for agents
Restore the binding zero-cost Noema contract while addressing free-route outages. AGENTS.md and docs/adr/0003-contextual-orchestrator-vendored-free-zdr.md require Noema to use orchestrator/free; orchestrator/auto admits fully price-attested paid routes after free-route preflight exhaustion. Keep Noema fail-closed on free-route exhaustion, or update the governing policy through a separately authorized decision before enabling paid fallback. Align the model exported later in this workflow and all affected contract tests with the selected governed route.
Was this helpful? React with 👍 or 👎 to provide feedback.
| python -m coverage run -m pytest \ | ||
| tests/test_noema_orchestrator_workflow_contract.py \ | ||
| tests/test_contextual_orchestrator_review_sidecar_contract.py \ | ||
| -q |
Incident
DiskSage Required Noema Review run
33255007404, job99107057131, reached the trusted contextual-orchestrator sidecar but exited before/healthzbecause every selectedorchestrator/freeroute failed the bounded plain-chat preflight. The visible413 request_too_largeline was the expected request-envelope self-test, not the crash.Root cause
The integrated auto-pool repair is free-first and permits a one-time fallback only to routes with complete published prompt/completion prices and currency evidence. Required Noema Review still provisions the sidecar without
CONTEXTUAL_ORCHESTRATOR_POOL, so it defaults tofree, and then hardcodesNOEMA_LLM_MODEL=orchestrator/free. It therefore cannot use the already-governed fallback path when all free routes are temporarily unavailable.TDD state
Current head is intentionally RED. The test-only commit requires Noema to provision
orchestrator/autoand invoke the matching virtual model while preserving ZDR enforcement and direct-provider exclusion. Production workflow changes follow only after the focused test is observed failing for the expected missing behavior.Intended repair
The DiskSage failed job was separately re-run as a tactical recovery attempt; this PR is the durable central repair.