Roadmap¶
For the full story backlog with acceptance criteria, see STORIES.md in the repo.
Current milestones¶
| Milestone | Status | Key stories |
|---|---|---|
| M1 — Core ingest + file store | ✅ shipped | STORY-001–004 |
| M2 — BM25 selection | ✅ shipped | STORY-005 |
| M3 — Web UI | ✅ shipped | STORY-006–008 |
| M4 — MCP server | ✅ shipped | STORY-009 |
| M5 — Semantic search + spelling tolerance | ✅ shipped | STORY-010 |
| M6 — Telegram bot | ✅ shipped | STORY-018 |
| M7 — Multi-turn conversation | ✅ shipped | STORY-019–020 |
| M8 — Quality scoring feedback loop | planned | STORY-038 |
| M9 — Memory tiers + consolidation | planned | STORY-040–042 |
| M10 — Episode graph nodes | planned | STORY-040 |
| M11 — Serendipity agent | planned | STORY-043 |
| M12 — Graph schema versioning | planned | STORY-049 |
Phase 2B (in progress)¶
Active distribution and UX work. See FEATURES.md for the full list.
Phase 3 (planned)¶
- Episode nodes + consolidation pipeline
- Serendipity / ambient enrichment agent
- User taste profile
- Server-side session management + context management
- Telemetry opt-in + debug export
Out of scope¶
- Hosted/cloud version — Lattice is local-first by design
- Atom deletion — quality scoring + archival achieves the retrieval quality goal without irreversible deletion
- LLM fine-tuning — RAG is strictly better for this use case; fine-tuning is static and has catastrophic forgetting risk