🔬 ResearchMind — Multi-Agent AI Research System

An autonomous multi-agent pipeline that researches any topic — searches, reads, writes, and critiques its own work.

Live Demo · Report a Bug · Request a Feature

---

📖 About The Project

ResearchMind is a multi-agent AI system that automates the entire research workflow — from a single topic input to a polished, self-reviewed research report — with zero manual effort.

Instead of one large model trying to do everything, ResearchMind breaks the task into four specialized agents, each with a narrow, well-defined job, working together in a pipeline. This mirrors how a real research team operates: someone gathers sources, someone reads and digs deeper, someone writes it up, and someone reviews it critically before it ships.

This project was built to demonstrate practical, production-style agentic AI system design — tool-calling agents, LCEL chains, structured state passing between stages, and a polished UI layer — rather than a single prompt-in/prompt-out wrapper.

🎯 What It Does

Give it any topic — "Quantum computing breakthroughs in 2026", "CRISPR gene editing", "Fusion energy progress" — and it will:

1. 🔍 Search the live web for recent, reliable sources
2. 📄 Scrape and extract deep content from the most relevant source
3. ✍️ Draft a structured, multi-section research report
4. 🧐 Critically review and score its own report
5. ⬇️ Let you download the final report as Markdown

---

🔗 Live Demo

👉 researchmind-ai-garv.streamlit.app

No installation needed — open the link, type a topic, and watch the four agents work in real time.

---

🖼️ Screenshots

Landing Page	Pipeline in Action	Final Report

---

✨ Key Features

• 🤖 True multi-agent architecture — 4 independent agents/chains, not one monolithic prompt
• 🔎 Live web-grounded research — real-time search via Tavily, not hallucinated facts
• 🌐 Autonomous web scraping — reads full page content beyond search snippets
• 🧠 Self-critique loop — a dedicated Critic agent scores and reviews the final report
• 📊 Real-time pipeline visualization — watch each agent's status live (waiting → running → done)
• ⬇️ Exportable output — download the finished report as a .md file
• 🎨 Custom-designed UI — fully custom dark-themed Streamlit interface, not default styling

---

🏗️ How It Works — Architecture

                    ┌─────────────────┐
     User Topic ──▶ │   Search Agent   │  → Tavily web search (Titles, URLs, Snippets)
                    └────────┬─────────┘
                             ▼
                    ┌─────────────────┐
                    │   Reader Agent   │  → Scrapes most relevant URL for deep content
                    └────────┬─────────┘
                             ▼
                    ┌─────────────────┐
                    │   Writer Chain   │  → Synthesizes findings into structured report
                    └────────┬─────────┘
                             ▼
                    ┌─────────────────┐
                    │   Critic Chain   │  → Scores & reviews the report (1–10 + feedback)
                    └────────┬─────────┘
                             ▼
                Final Report + Feedback (downloadable)

Search Agent and Reader Agent are true tool-calling agents (built with LangChain's create_agent) that decide how to use their tools (web_search, scrape_url). Writer Chain and Critic Chain are LCEL (prompt | llm | parser) chains — deterministic, single-purpose steps that transform state into the next stage's input.

---

🛠️ Tech Stack

Layer	Technology
LLM	Mistral AI (mistral-medium-3-5) via langchain-mistralai
Agent Orchestration	LangChain create_agent + LangGraph
Web Search	Tavily API
Web Scraping	requests + BeautifulSoup4
UI / Frontend	Streamlit (custom CSS, no default theme)
Deployment	Streamlit Community Cloud
Language	Python 3.10+

---

📂 Project Structure

ResearchMind-Multi-Agent-AI/
├── app.py              # Streamlit UI — main entry point for deployment
├── agents.py           # Agent + chain definitions (search, reader, writer, critic)
├── pipeline.py          # CLI-based pipeline runner (for local/terminal use)
├── tools.py             # Custom tools: web_search (Tavily), scrape_url (BeautifulSoup)
├── requirements.txt     # Python dependencies
├── .gitignore            # Excludes .env, .venv, __pycache__
└── README.md

---

🚀 Getting Started — Run It Locally

Prerequisites

• Python 3.10 or higher
• A Mistral AI API key
• A Tavily API key (free tier available)

Installation

1. Clone the repository

git clone https://github.com/garvkumarsharma/ResearchMind-Multi-Agent-AI.git
cd ResearchMind-Multi-Agent-AI

2. Create and activate a virtual environment

python -m venv .venv

# Windows
.venv\Scripts\activate

# macOS/Linux
source .venv/bin/activate

3. Install dependencies

pip install -r requirements.txt

4. Set up environment variables

Create a .env file in the project root:

MISTRAL_API_KEY=your_mistral_api_key_here
TAVILY_API_KEY=your_tavily_api_key_here

⚠️ Never commit your .env file. It's already excluded via .gitignore.

5. Run the app

To launch the Streamlit UI:

streamlit run app.py

Or, to run the pipeline directly from the terminal (no UI):

python pipeline.py

The app will open at http://localhost:8501.

---

☁️ Deployment

This project is deployed on Streamlit Community Cloud, connected directly to this GitHub repository.

If you'd like to deploy your own fork:

1. Push your fork to GitHub
2. Go to share.streamlit.io → Create app
3. Point it at your repo, branch main, main file app.py
4. Under Advanced settings → Secrets, add:

   MISTRAL_API_KEY = "your_key_here"
   TAVILY_API_KEY = "your_key_here"

5. Deploy 🚀

---

🧭 Roadmap / Future Improvements

• Add support for multiple LLM providers (OpenAI, Gemini) via a model switcher
• PDF export in addition to Markdown
• Conversation memory for follow-up/refinement queries
• Unit tests for agents and tools

---

👤 Author

Garv Kumar Sharma

• GitHub: @garvkumarsharma
• LinkedIn: linkedin.com/in/garv-kumar-sharma
• Email: garvdhiman2004@gmail.com

---

📄 License

This project is open source and available under the MIT License.

---

If you found this project interesting, consider giving it a ⭐ on GitHub!