Framework integration guides
Use these guides when wiring IterableData into AI frameworks or provider SDKs. Longer provider notes stay in the repository so llms.txt and GitHub remain the single source of truth; this page summarizes what each guide covers and links to the in-docs agent APIs.
In-docs guides (start here)
| Guide | When to use |
|---|---|
| Building agents | Designing an agent that detects formats, samples rows, and converts data |
| MCP server | Exposing IterableData tools over Model Context Protocol (iterabledata[mcp]) |
| Agent tools API | Calling detect_format, read_sample, plan_conversion, and related tools from Python |
| Catalog API | Machine-readable format/capability metadata for planners |
| AI API | Native providers (openai, anthropic, gemini, azure) and doc.generate |
Provider and framework notes (GitHub)
| Guide | Summary |
|---|---|
| AI frameworks | LangChain, CrewAI, and AutoGen patterns for tools and memory |
| OpenAI | OpenAI SDK usage with iterable.ai and redaction |
| Anthropic Claude | Claude / Messages API wiring |
| Google Gemini | Gemini provider setup |
Install AI extras with pip install 'iterabledata[ai]' (add [mcp] for the MCP server). Never commit API keys; use redact_for_llm() before cloud calls.
Extending formats outside core
For a niche format that should not land in the main package, ship a small plugin package with the iterabledata.formats entry point. See Plugin system and the reference plugin walkthrough for a complete layout.