Roadmap

This page describes future intent. The current state lists what works.

Make 0.1 trustworthy in real repositories and publish it.

Publish the first release so generated CI can install @entwine/cli from npm, then run the generated GitHub, GitLab, and Bitbucket pipelines on real repositories and fix what they reveal.
Resolve the gaps found by running entwine check on ForgeCMS, Flowview, and Agentyx. Links from docs to repository files outside docs/ (such as ../CLAUDE.md or ../apps/...), links to raw-HTML anchors (<a id="...">), and links to a directory without an index page are all errors today and are the most common blockers. A docs/README.md entry point is not yet recognized as the project role. Decide each deliberately; the regression tests in crates/entwine-engine/tests/compilation.rs pin current behavior.
Improve graph readability beyond roughly 50 documents, where the fixed radial layout produces a wide canvas that the text relationship list serves better.

Later

Directions that depend on evidence from real use. None is promised, and none commits Entwine to a specific provider or integration.

Documentation drift and stale-document hints.
Change-aware recommendations after code changes.
Better project-state assistance for people maintaining state.md.
Optional integration with coding agents through structured context, which stays the stable boundary.
Semantic context selection, so tools can request the relevant slice of knowledge.
Additional knowledge roles, if the convention proves too small.
Search, small opt-in configuration, and broader distribution, evaluated separately. Preserve the canonical model boundary.

Entwine will not generate documentation with AI or invent project facts; it scaffolds questions and leaves answers to people and their tools.