Deployment and Operations

Run reliably. Recover fast.

This page condenses deployment modes, release readiness, observability priorities, and incident runbook patterns for production Locus usage.

Environment matrix

EnvironmentStorageHost ProfilePrimary Goal
Local DevelopmentIn-memory or embedded SurrealSingle processFast iteration
CI ValidationIn-memory fixturesTest jobs onlyRegression detection
StagingRemote SurrealDBGateway and MCP parity checksRelease rehearsal
ProductionRemote SurrealDB restricted accessManaged gateway and MCP deploysReliability and auditability
Release readiness

Checklist

  1. Workspace compile and tests pass.
  2. SDK examples execute in smoke path.
  3. Host integrations confirm contract compatibility.
  4. Changelog and migration notes updated.
  5. Security checklist reviewed for secrets/logging hygiene.
cargo check --workspace --examples
cargo test -p locus-core-rs --lib
cargo test -p locus-sdk
Health signals

Observe first

  • Parse success ratio
  • Validation failure ratio
  • Recall latency percentiles
  • Transform success/failure rates
  • Embedding backfill throughput
Incident runbook A

Parser spike

  1. Capture failing payloads.
  2. Validate four-layer ordering and key format.
  3. Compare behavior with last known-good release.
  4. Rollback host version if release-correlated.
Incident runbook B

Retrieval regression

  1. Capture request payload + scoring settings + path.
  2. Run explain workflow with same request.
  3. Compare fallback behavior against baseline.
  4. Patch defaults only with regression tests.
Incident runbook C

Transform instability

  1. Switch to dry-run.
  2. Validate selected nodes and provider capabilities.
  3. Reduce batch size and rerun with checkpoints.
  4. Resume full run only after parity.