How vikasjha2003/fathom-research works
Fathom Research turns a user’s topic into a sourced report through search, page reading, drafting, and critic-guided revision. The Streamlit interface streams workflow progress and presents the report, research, review, and download. In main.py, a LangGraph stateful workflow coordinates specialized agent nodes, tool dispatch, result extraction, and quality-gated routing. The README documents the model providers and product capabilities; sampled code shows the principal workflow and tool implementations. The available UI excerpt is partial, so some presentation details are based on the README.
Main entry points: main.py StateGraph orchestration · streamlit_app.py Streamlit interface · prompts.py Agent prompts · tools.py Web search tool
Components 17
User interface
- Streamlit interface
streamlit_app.py - Report and results
streamlit_app.py
Research workflow
- StateGraph orchestration
main.py - Research state
main.py - Agent prompts
prompts.py - Searcher agent
main.py - Search result extraction
main.py - Reader agent
main.py - Reading result extraction
main.py - Report writer
main.py - Critic and quality gate
main.py
Research tools
- Web search tool
tools.py - Page scraper
tools.py
External services
- LLM providers
- Tavily search
- Web pages
Other components
- Researcher
Connections 27
- Researcher → Streamlit interface: enters topic
streamlit_app.py - Streamlit interface → StateGraph orchestration: streams workflow
streamlit_app.py - StateGraph orchestration → Research state: tracks state
main.py - StateGraph orchestration → Searcher agent: starts search
main.py - Searcher agent → Agent prompts: loads prompt
main.py - Searcher agent → LLM providers: invokes model
README.md - Searcher agent → Web search tool: dispatches queries
main.py - Web search tool → Tavily search: searches web
tools.py - Web search tool → Web pages: returns source links
- StateGraph orchestration → Search result extraction: routes results
main.py - Search result extraction → Reader agent: passes findings
main.py - Reader agent → Agent prompts: loads prompt
main.py - Reader agent → LLM providers: invokes model
README.md - Reader agent → Page scraper: dispatches URLs
main.py - Page scraper → Web pages: fetches pages
tools.py - StateGraph orchestration → Reading result extraction: routes readings
main.py - Reading result extraction → Report writer: passes evidence
main.py - Report writer → Agent prompts: loads prompt
main.py - Report writer → LLM providers: invokes model
README.md - Report writer → Critic and quality gate: submits draft
main.py - Critic and quality gate → Agent prompts: loads prompt
main.py - Critic and quality gate → LLM providers: invokes model
README.md - Critic and quality gate → Report writer: routes revision
main.py - Critic and quality gate → StateGraph orchestration: ends when approved
main.py - StateGraph orchestration → Streamlit interface: streams updates
streamlit_app.py - Streamlit interface → Report and results: presents results
README.md - Report and results → Researcher: offers download
README.md