Skip to content

v0.0.6

Choose a tag to compare

View all tags
@github-actions github-actions released this 21 Sep 07:42
· 26 commits to main since this release
v0.0.6
8dffb5d
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Added

  • Configurable embedding model per client via the --model <name> CLI flag and CDS_MCP_MODEL environment variable
  • Capire-versioned embeddings with local manifest resolution
  • Per-chunk metadata in searchMarkdownDocs output

Changed

  • Changed the default embedding model from Xenova/all-MiniLM-L6-v2 to sentence-transformers/all-MiniLM-L6-v2
  • Switched to CAP AI SQLite VECTOR_EMBEDDING
  • Improved search_docs query param description
  • Replaced background model polling and CDS_MCP_REFRESH_MS with request-driven model and configuration refresh

Fixed

  • Blank Cache-Control/Pragma so conditional GET can 304, avoiding a full re-download of the embeddings bundle on every startup
  • Isolated project model caches and CAP compiler/global state across requests
  • Preserved transitive model imports for explicitly trusted direct CLI projects
  • Prevented rejected MCP model compilations from exposing out-of-root diagnostics
  • Refreshed cached models after changes to CAP project configuration
  • Restricted MCP model access to advertised workspace roots, including symlink and resolved-source validation
Assets 2
Loading