Roadmap¶
Planned improvements to this list and its automation. This is a living document: items are unordered within each horizon, carry no date commitments, and may change as the ecosystem evolves. Contributions toward any item are welcome — see CONTRIBUTING.md. Completed work is recorded in the Changelog.
Near-term¶
- Diagnose and restore the weekly discovery output cadence —
.github/PROPOSED_UPDATES.mdwas last regenerated on 2026-05-11 despite the weekly schedule. - Triage the current discovery candidates with explicit accept / reject notes against the production-infrastructure scope (see the FAQ).
- Add a CI check that validates every
benchmarks.mdrow against the six-field schema (Metric, Value, Tag, Source, Date, Methodology) and a valid evidence tag.
Mid-term¶
- Unit tests for
scripts/discovery_engine.pycovering the remaining URL parsing and discovery markdown generation (the verified-date audit is now tested). - Extend the decision tree beyond frameworks and vector databases to embedding selection, reranking, and chunking strategy.
Longer-term¶
- A postmortem / failure case studies section, cross-linked with rag-pitfalls.md.
- Evaluate splitting the largest README sections (for example, Agentic RAG and Multimodal RAG) into side documents as they grow.
Out of Scope¶
These are deliberate decisions, not gaps: