Home

---

Fonky is a Python library that exposes retrieval, loading, scraping, document-processing, and text-processing functionality to AI agents through provider-specific tool integrations.

Purpose

Fonky separates canonical implementations from agent-framework integration. Retrieval, loading, scraping, and processing logic remains in shared modules while provider packages expose those operations through the conventions required by OpenAI Agents SDK, Anthropic Claude, Google ADK, xAI, Mistral AI, and LangChain.

Core capabilities

Area	Module	Purpose
Retrieval	fonky.fetchers	Retrieve data from APIs, search services, public data sources, geospatial services, environmental services, and web sources.
Loading	fonky.loaders	Load documents, files, cloud objects, mail, notebooks, and structured data.
Web extraction	fonky.scrapers	Extract structured and unstructured content from web pages and crawled sources.
Processing	fonky.processors	Normalize, clean, tokenize, transform, chunk, vectorize, and search text and datasets.
Models	fonky.models	Shared models and response structures used by canonical implementations.
Configuration	fonky.config	Runtime configuration, credentials, and defaults.
Error handling	fonky.boogr	Shared error wrapping and logging.

Provider integrations

Package	Framework	Tool exposure
fonky.gpt.tools	OpenAI Agents SDK	@function_tool
fonky.claude.tools	Anthropic Claude SDK	@beta_tool
fonky.gemini.tools	Google ADK	Plain Python callables wrapped by ADK
fonky.grok.tools	xAI SDK	Executable callables plus explicit *_tool declarations
fonky.mistral.tools	Mistral AI SDK	Executable callables plus JSON *_tool declarations
fonky.langchain.tools	LangChain Core	@tool(parse_docstring=True)

The architecture and workflow diagrams identify Claude and Mistral separately because they use different declaration objects and tool-result messages even though both execute Fonky callables locally.

Each provider package exposes the same logical Fonky operations while preserving the provider-specific tool contract. Every provider integration is a peer adapter that delegates directly to the canonical to the canonical Fonky implementation modules rather than depending on another provider package.

Documentation layout

1. Architecture defines canonical modules, provider adapters, naming rules, and execution flow.
2. User Guide provides task-oriented examples for selecting and using Fonky tools.
3. API Reference renders source documentation with MkDocs and mkdocstrings.
4. Development defines validation, documentation, and extension requirements.

Build the documentation

python -m pip install -r requirements.txt
python -m pip install mkdocs-material mkdocstrings[python]
mkdocs serve

Build the static site:

mkdocs build