Skip to content

contextpull 0.1.0

Choose a tag to compare

View all tags
@mi2arun mi2arun released this 14 Sep 13:41
· 21 commits to main since this release
v0.1.0
95a2f61

First release. Pull, don't push: an always-in-context index plus five tools (index, search, read, grep, neighbours) that return exact document sections on demand, for Claude Code, any MCP host, or your own client.

  • Zero-dependency Python core over SQLite + FTS5; heading-aware sections with stable ids; incremental ingest; flat/compact/hierarchical index under a token budget
  • MCP server (pip install "contextpull[mcp]"): stdio and streamable HTTP; index delivered via server instructions, tool and resource
  • PDF ingest ([pdf] extra), optional embeddings with hybrid search, model-written summaries
  • Node package contextpull on npm: store-native reader and MCP server, identical results to Python (23/23 conformance)
  • ragbisect adapters that put the agentic loop in the same benchmark table as bm25, dense and hybrid; first results in benchmarks/

PyPI: https://pypi.org/project/contextpull/0.1.0/ · npm: https://www.npmjs.com/package/contextpull/v/0.1.0

Assets 2
Loading