How akarshvyas/multi-agent-research-system works
ResearchMind accepts a topic through a Streamlit interface or the standalone pipeline, then searches for sources, scrapes selected content, drafts a report, and critiques it. The agent definitions and prompt chains live in agents.py; tools.py implements Tavily search and page extraction, while agents.py configures the OpenAI model. The interface runs these stages directly; pipeline.py provides a separate command-line workflow. The README is empty, so capabilities and relationships here are based on the sampled source; no persistence layer is shown.
Main entry points: app.py Research UI · pipeline.py CLI pipeline · agents.py Search agent · tools.py Web search tool
Components 12
Entry points
- Research UI
app.py - CLI pipeline
pipeline.py
Research stages
- Search agent
agents.py - Reader agent
agents.py - Report writer
agents.py - Report critic
agents.py
External integrations
- Web search tool
tools.py - Page scraper
tools.py - OpenAI model
- Tavily search
Other components
- Researcher
- Web pages
Connections 18
- Researcher → Research UI: enters topic
app.py - Researcher → CLI pipeline: enters topic
pipeline.py - Research UI → Search agent: invokes
app.py - Research UI → Reader agent: invokes
app.py - Research UI → Report writer: invokes
app.py - Research UI → Report critic: invokes
app.py - CLI pipeline → Search agent: invokes
pipeline.py - CLI pipeline → Reader agent: invokes
pipeline.py - CLI pipeline → Report writer: invokes
pipeline.py - CLI pipeline → Report critic: invokes
pipeline.py - Search agent → Web search tool: registers tool
agents.py - Reader agent → Page scraper: registers tool
agents.py - Web search tool → Tavily search: queries
tools.py - Page scraper → Web pages: fetches page
tools.py - Search agent → OpenAI model: uses model
agents.py - Reader agent → OpenAI model: uses model
agents.py - Report writer → OpenAI model: uses model
agents.py - Report critic → OpenAI model: uses model
agents.py