Ingest a local docs folder
docmancer ingest ./docs
Mixed file formats are fine; install docmancer[local] first if you have PDF, DOCX, RTF, or HTML.
Add a docs site
docmancer add https://docs.pytest.org
Docmancer detects the platform and chooses the fetch strategy:
Detected platform: generic
Indexed sections with SQLite FTS5 + Qdrant (dense + sparse)
Add a GitHub repo
docmancer add https://github.com/vercel/next.js
Extracts the README and docs/ directory.
Query with hybrid mode and explain
docmancer query "How do I parametrize a pytest fixture?" --mode hybrid --explain
Each result is annotated with the per-source rank that contributed (lexical#1, dense#3, sparse#2).
Query with JSON output
docmancer query "pytest fixture scope" --format json
Useful for piping into other tools.
Expand context to full pages
docmancer query "Next.js middleware" --expand
docmancer query "Next.js middleware" --expand page
--expand includes adjacent sections; --expand page returns full pages within the token budget.
Increase the token budget
docmancer query "OAuth 2.0 flow" --budget 3000
The default is 2400 tokens.
Skip vectors for a quick lexical-only run
docmancer ingest ./docs --no-vectors
docmancer query "exact error message text" --mode lexical
Useful when you want fast FTS5 without the embedding pass.
Update a source
docmancer update https://docs.pytest.org
docmancer update
The second form re-fetches every source. Unchanged sections skip re-embedding via the content-hash cache.
Install and call an API MCP pack
docmancer install-pack open-meteo@v1
docmancer mcp doctor
Then ask your agent:
Get the current weather in New York Central Park.
The agent calls open_meteo__v1__forecast through docmancer mcp serve.
Manage the local Qdrant
docmancer qdrant status
docmancer qdrant down
docmancer qdrant up
docmancer qdrant logs
Remove and clear
docmancer remove https://docs.pytest.org
docmancer remove --all
--all clears the index but keeps your config.