Phase 5 — Goals

Goal 5.1: Observability production stack

Description: Deploy Prometheus, Grafana, Loki, Tempo (or equivalent) per TECH_STACK.md; wire exporters from all services.

Acceptance criteria:

• Dashboards for proxy golden signals and dependency health
• Alert rules for SLO burn (error rate, latency)
• No high-cardinality labels (no org_id in Prometheus labels)

Validation: Staging soak; alert fire drill

---

Goal 5.2: Resilience and performance validation

Description: Load, chaos, and benchmark tests on proxy and context assembly.

Acceptance criteria:

• Documented results vs PERFORMANCE.md budgets
• Circuit breaker thresholds tuned with metrics evidence
• No memory leaks under sustained streaming load

Validation: CI nightly or manual gate before release

---

Goal 5.3: CI and release discipline

Description: Harden CI, release process, and dependency governance.

Acceptance criteria:

• Go/Python/TS CI matrices where applicable
• Release checklist in DEPLOYMENT.md followed for v0.x
• Dependabot policy expanded beyond github-actions

Validation: Dry-run release from main

---

Goal 5.4: Security and compliance baseline

Description: Threat model review, pen-test fixes, GDPR deletion path for memories.

Acceptance criteria:

• SECURITY.md checklist items traced to tests or runbooks
• Deletion flow audited; no cross-tenant deletion possible
• Secrets rotation documented

Validation: Security review sign-off; table-top incident exercise