Phase 4: artifact provenance — model-tracking + debug endpoint

Phase 1: persist tour/diagram/README/repo-map artifacts to Qdrant

Tier 2: shareable conversation URLs backed by Qdrant

feat: persistent agent memory + README generator

refactor: split main.py into focused routers

feat(ui): Blueprint Dark — complete color system rebuild

Fix PostHog blocked by ad blockers via Vercel reverse proxy

rebrand: update all display name references to Cartographer

fix: keepalive during retrieval phase to prevent proxy timeout

Fix RAG stream hanging: add async keepalive to /query/stream

Add model selector: dropdown UI + /agent/models endpoint

fix: use queue-based keepalive for agent SSE — prevent generator corruption

feat: use Cerebras qwen-3-235b for agent tool calling (tested, confirmed working)

fix: remove Cerebras from AgentService — llama3.1-8b too small for tool use

fix: include CEREBRAS_API_KEY in agent init guard in main.py

fix: add SSE keepalive to ingest route (same proxy timeout issue)

fix: add SSE keepalive to prevent HF Space proxy from dropping agent connections

feat: Mermaid diagrams, agent fixes, deployment prep

Major feature: interactive diagrams, MCP integration, and bug fixes

checkpoint: pre-UI-overhaul state

Add ingestion progress streaming via SSE

Fix remaining architect-flagged issues — targeting 9.5+

Fix 7 architect-identified issues, add eval harness — targeting 9.5+

Cleanup + slash prompt autocomplete + CLAUDE.md file hygiene rule

Phase 7: MCP server + client — agent now discovers tools via protocol

Fix CORS: allow any localhost port via regex instead of hardcoded list

Fix agent SSE error handling — use agent_error event, never return 503

Fix duplicate graph, agent error handling, and UI clarity

Phase 6: Code Knowledge Graph (D3 force-directed call graph)

Phase 5: MCP server + Agentic RAG (ReAct loop)

Fix 5 bugs found in expert review

Phase 4: deployment config (Vercel + Render + GitHub Actions CI)

Phase 3: React UI — chat, repo management, streaming, source cards

Phase 2: retrieval, generation, and FastAPI backend