ResearchMind

A multi-agent AI research system that searches, scrapes, writes, and critiques — delivering a polished research report on any topic.

Core Features

• 🔍 Search Agent — Fetches recent, relevant web results via Tavily API
• 📄 Reader Agent — Scrapes the most valuable source and extracts clean content
• ✍️ Writer Chain — Synthesises search + scraped data into a structured, professional report
• 🧐 Critic Chain — Reviews the report and provides a score, strengths, and areas to improve
• 🖥️ Streamlit UI — Dark-themed dashboard with live pipeline progress tracking
• ⏱️ Auto-retry — Handles API rate limits with exponential backoff
• 📥 Download — Export reports as Markdown files

Architecture

User Input (topic)
       │
       ▼
┌──────────────────┐
│  Search Agent    │  Tavily web search → top 5 results
│  (tool call)     │
└──────────────────┘
       │
       ▼
┌──────────────────┐
│  Reader Agent    │  Picks best URL → scrapes & extracts 3000 chars
│  (LLM + scrape)  │
└──────────────────┘
       │
       ▼
┌──────────────────┐
│  Writer Chain    │  LCEL: prompt → LLM → structured report
│  (LCEL pipeline) │
└──────────────────┘
       │
       ▼
┌──────────────────┐
│  Critic Chain    │  LCEL: prompt → LLM → score + feedback
│  (LCEL pipeline) │
└──────────────────┘
       │
       ▼
   Final Output

Tech Stack

Layer	Technology
Language	Python 3.12
LLM	Groq (openai/gpt-oss-20b)
Search	Tavily API
Scraping	Requests + BeautifulSoup
Orchestration	LangChain (agents + LCEL chains)
UI	Streamlit
Package manager	uv

Project Structure

research-mind/
├── agents/
│   ├── search_agent.py      # Tavily-powered web search agent
│   └── reader_agent.py      # URL scraping & extraction agent
├── tools/
│   ├── web_search.py        # Tavily search tool wrapper
│   └── scrape_url.py        # BeautifulSoup scraper tool
├── utils/
│   ├── chains.py            # Writer & Critic LCEL pipelines
│   ├── config.py            # Environment & config loader
│   └── prompts.py           # Prompt templates
├── architecture/
│   └── research-mind.png    # Architecture diagram
├── public/
│   └── Dashboard.png        # UI screenshot
├── main.py                  # CLI entry point
├── app.py                   # Streamlit dashboard
├── pyproject.toml           # Dependencies & metadata
├── .env.example             # API key template
├── .python-version          # Python 3.12
└── README.md

Local Setup

Prerequisites

• Python 3.12+
• uv installed
• API keys for Groq and Tavily

Installation

# Clone the repository
git clone <repo-url>
cd research-mind

# Create environment and install dependencies
uv sync

# Configure API keys
cp .env.example .env

Edit .env and add your keys:

TAVILY_API_KEY=tvly-xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
GROQ_API_KEY=gsk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Run

CLI mode:

uv run main.py

Streamlit dashboard:

uv run streamlit run app.py

Usage

1. Enter a research topic (e.g. "Recent AI news and fundings")
2. The pipeline runs 4 stages — watch progress in real time
3. Read the generated report and critic feedback
4. Download the report as a .md file

Live Demo

Launch ResearchMind →

License

MIT