Library Documentation Ingestor

Ingest and learn the API of any Python library so you can use it
correctly in scientific analysis code.

When to Use

• The user asks you to work with a library you don't have detailed
  knowledge of (e.g. a niche scientific package).
• You need to know the exact constructor signatures, method names,
  parameter defaults, or return types for a library.
• You want to avoid hallucinating API details — ingest first, then code.
• The user explicitly asks you to "learn" or "look up" a library.

Procedure

Step 1 — Check for Existing Docs

Before ingesting, check whether a reference already exists:

read_doc("<package>_api")

If a doc is returned, skip to Step 4.

Step 2 — Ingest the Library

Call the ingestion tool with the PyPI package name:

ingest_library_docs(package_name="<package>")

Optionally provide a github_url for deeper source-code crawling:

ingest_library_docs(package_name="<package>", github_url="https://github.com/org/repo")

This will:

1. Crawl PyPI metadata, ReadTheDocs/Sphinx API pages, GitHub README,
   source code signatures, and docs/examples folders
2. Feed all scraped content to an LLM documentation analyst
3. Extract Core Classes, Key Functions, Common Pitfalls,
   and Quick-Start Recipes into a structured reference
4. Save it to your docs directory as <package>_api.md

Step 3 — Verify the Result

The tool returns a status dict with doc_name and path. Confirm
the document was created:

read_doc("<package>_api")

Step 4 — Use the Reference

Consult the ingested reference when writing analysis code:

• Look up exact class constructors and method signatures
• Check the Common Pitfalls section before using unfamiliar APIs
• Use Quick-Start Recipes as starting points for common tasks
• Cite the reference when explaining your code choices

What It Produces

A library_api.md-format reference document with four sections:

Section	Contents
Core Classes	Main classes with constructors, methods, parameter tables, return types
Key Functions	Standalone functions with signatures and parameter tables
Common Pitfalls	Gotchas, naming conflicts, unit mismatches, common mistakes
Quick-Start Recipes	Copy-paste code snippets for common tasks

Requirements

• Requires sciagent[wizard] to be installed:
  pip install sciagent[wizard]
• Needs network access to crawl documentation sources
• Works best with well-documented packages on PyPI / ReadTheDocs / GitHub

Domain Customization