Deep ReSearch

Deep ReSearch is a Gradio-based research assistant that takes a question, runs an async multi-agent research workflow, and streams the final report back to the UI.

Features

• Simple web UI built with Gradio
• Async research pipeline
• Live status updates while research runs
• Custom search tool built with Tavily
• Example prompts for quick testing
• Environment variable support via python-dotenv

Requirements

• Python 3.10+
• gradio
• python-dotenv
• Tavily API access

Install dependencies:

pip install -r requirements.txt

Setup

Create a .env file in the project root and add your API keys:

TAVILY_API_KEY=your_tavily_key_here
OPENAI_API_KEY=your_openai_key_here

Run the app

python app.py

Then open the local Gradio URL shown in the terminal.

Project Structure

deep_research/
├── app.py
├── research_manager.py
├── planner_agent.py
├── search_agent.py
├── writer_agent.py
├── email_agent.py
├── messenger.py
├── styles.py
├── simple.py
├── requirements.txt
└── README.md

Notes

• ResearchManager().run(query) streams updates to the UI.
• The search step uses a custom Tavily-based search tool.
• To avoid committing generated files, add this to .gitignore: