Operations and Gates

This page defines the production operating baseline.

What this page is for

1. Decide whether a deployment is safe enough to release.
2. Define the minimum operating evidence for production.
3. Use replay to debug incidents and verify fixes.

Daily checks

Health:

curl -sS http://localhost:${PORT:-3001}/health | jq

Core production gate:

npm run -s gate:core:prod -- --base-url "http://localhost:${PORT:-3001}" --scope default

Policy sanity:

curl -sS http://localhost:${PORT:-3001}/v1/memory/rules/evaluate \
  -H 'content-type: application/json' \
  -d '{"tenant_id":"default","scope":"default","context":{"intent":"support_triage"}}' | jq '{matched}'

Go-live minimum

Before production traffic, confirm all of the following:

1. Health checks pass from the real deployment environment.
2. Core production gate completes without blocking failures.
3. Auth, rate limiting, and tenant isolation are enabled.
4. One end-to-end workflow produces a complete replay chain.
5. Operators know where to find logs, metrics, and rollback procedure.

Replay Execution

Replay identifiers:

1. request_id
2. run_id
3. decision_id
4. commit_uri

Operational replay workflow:

1. Extract failing request_id from logs.
2. Rebuild run/decision chain with run_id and decision_id.
3. Resolve affected objects via POST /v1/memory/resolve and commit_uri.
4. Replay workflow and compare with expected behavior.

Release evidence checklist

1. Core gate summary.
2. Health and consistency outputs.
3. Performance snapshot.
4. Replay evidence chain for at least one workflow.

Regression checks

npm run -s e2e:replay-learning-fault-smoke
npm run -s e2e:replay-learning-retention-smoke
RUN_REPLAY_LEARNING_SMOKES=true npm run -s regression:oneclick

What to archive after a release

1. Gate outputs and health evidence.
2. The deployment version or commit linked to the release.
3. At least one replayable workflow trace from the new release.
4. Any policy changes introduced with the release.