Implement full Lore roadmap: 9 features across 4 phases

[amitpaz1]

Summary

Full implementation of the Lore roadmap — 9 features across 4 phases, shipping full-stack per feature (REST API + CLI + MCP tools + Web UI).

Phase 1: Foundation

• F1: Guided Bootstrap — lore bootstrap validates Python, Postgres, pgvector, Docker, runs migrations, verifies health
• F2: Enhanced Setup Wizard — Config validation, test-connection, backup/rollback support with --validate, --test-connection, --dry-run flags
• F3: SLO Dashboard + Alerting — SLO CRUD, background checker (60s interval), webhook/email alert channels, timeseries API, UI panel

Phase 2: Intelligence

• F4: Adaptive Retrieval Profiles — Named profiles in Postgres (coding/incident-response/research presets), per-request ?profile= param, 60s cache
• F5: Graph Approval Inbox — Risk scoring via SQL CTE, sortable params, batch operations, reviewer notes, review_decisions audit table
• F6: Policy-Based Retention — Declarative lifecycle policies, cron-based snapshot schedules, restore drills with timing metrics, compliance checks

Phase 3: Platform

• F7: Multi-Tenant Workspaces — Workspace isolation within orgs, RBAC with workspace:manage/workspace:read, audit log with filtering
• F8: Plugin SDK — LorePlugin ABC with 5 hooks, entry_point discovery, hot-reload, scaffold CLI (lore plugin create), test harness
• F9: Proactive Recommendations — Multi-signal scoring (context similarity, entity overlap, temporal patterns, access patterns), explainability, feedback loop, MCP suggest() tool

Infrastructure

• 6 new migrations (012–017)
• 118 new tests (2081 total, all passing)
• 51 files changed, ~5500 lines added
• Background tasks: SLO checker + policy scheduler in FastAPI lifespan

Test plan

• All 2081 tests pass (118 new + 1963 existing)
• Bootstrap checks tested with mocked subprocess/env
• Setup wizard validation tested with real filesystem ops
• SLO threshold evaluation + alert dispatch tested
• Profile cache TTL + model validation tested
• Plugin registry, hooks, scaffold, harness tested
• Recommendation signals tested individually (context similarity, entity overlap, temporal, access patterns)
• Feedback recorder weight clamping tested
• Integration test with running Postgres (manual)

🤖 Generated with Claude Code