feat: add git-consistent memory gateway architecture

testdata/docs/disaster-recovery.md

@@ -0,0 +1,6 @@
+# Disaster Recovery Policy
+
+- Recovery Time Objective (RTO): 2 hours.
+- Recovery Point Objective (RPO): 15 minutes.
+- Test failover monthly.
+- Keep indexed meeting notes for decision traceability.

testdata/docs/incident-runbook.md

@@ -0,0 +1,8 @@
+# Incident Runbook
+
+When the QMD HTTP MCP service is degraded:
+
+- Check `GET /health`.
+- Validate vector index status with `qmd status`.
+- Restart container only after collecting logs.
+- Escalate if query quality drops after embedding drift.

testdata/meetings/2026-03-01-architecture.md

@@ -0,0 +1,8 @@
+# Architecture Sync 2026-03-01
+
+Decisions:
+
+- Use QMD as internal shared MCP endpoint.
+- Keep HTTP endpoint on port 8181.
+- Persist cache and index to avoid repeated cold starts.
+- Confirm disaster recovery RTO target remains 2 hours.

testdata/meetings/2026-03-02-ops.md

@@ -0,0 +1,6 @@
+# Ops Review 2026-03-02
+
+Action items:
+
+- Add smoke test covering health, collection add, embed, search, and query.
+- Validate that deep query can locate runbook and recovery policy.

testdata/notes/ops-checklist.md

@@ -0,0 +1,6 @@
+# Ops Checklist
+
+1. Verify MCP service health endpoint before release.
+2. Confirm embeddings are up to date.
+3. Keep cache volume persistent to avoid repeated model downloads.
+4. Keep runbook links in docs/incident-runbook.md.

testdata/notes/personal-productivity.md

@@ -0,0 +1,6 @@
+# Personal Productivity Notes
+
+- Daily review at 09:00.
+- Focus blocks: 90 minutes.
+- Keep a short incident checklist for local services.
+- Search keyword target: lightweight knowledge retrieval.