ResearchMind 🔬

Multi-agent AI research system — four specialized agents collaborate to search, scrape, write, and critique a polished research report on any topic.

Built with LangChain, Mistral AI, and Streamlit.

---

Architecture

╔══════════════════════════════════════════╗
║         01  ·  SEARCH AGENT              ║
║     Tavily web search → URLs & snippets  ║
╚══════════════════════════════════════════╝
                    │
                    ▼
╔══════════════════════════════════════════╗
║         02  ·  READER AGENT              ║
║   BeautifulSoup deep scrape → content    ║
╚══════════════════════════════════════════╝
                    │
                    ▼
╔══════════════════════════════════════════╗
║         03  ·  WRITER CHAIN              ║
║  Mistral LLM → structured report draft   ║
╚══════════════════════════════════════════╝
                    │
                    ▼
╔══════════════════════════════════════════╗
║         04  ·  CRITIC CHAIN              ║
║   Mistral LLM → score /10 + feedback     ║
╚══════════════════════════════════════════╝

Step-by-step

Step	Agent / Chain	What it does
01 Search	build_search_agent()	Queries Tavily for recent web results (titles, URLs, snippets) on the research topic.
02 Reader	build_reader_agent()	Picks the most relevant URL from search results and scrapes the full page content via BeautifulSoup.
03 Writer	writer_chain	Drafts a structured report with introduction, key findings, conclusion, and sources.
04 Critic	critic_chain	Reviews the report, assigns a score out of 10, lists strengths, areas to improve, and a one-line verdict.

---

Project structure

.
├── app.py              # Streamlit web UI
├── pipeline.py         # CLI runner (rich terminal output)
├── agents.py           # Agent & chain definitions
├── tools.py            # web_search + scrape_url LangChain tools
├── pyproject.toml      # Project metadata & dependencies
├── requirements.txt    # Pip-compatible dependency list
├── .env                # API keys (gitignored)
└── .streamlit/
    ├── config.toml     # Streamlit theme/config
    └── secrets.toml    # Streamlit secrets

---

Setup

1. Clone & enter the project

git clone <repo-url> && cd deep-research-agent

2. Install dependencies

pip install -r requirements.txt

Or with uv:

uv sync

3. Set your API keys

Create a .env file in the project root:

MISTRAL_API_KEY=your-mistral-api-key
TAVILY_API_KEY=your-tavily-api-key

• Mistral API key
• Tavily API key

---

Usage

Web UI (Streamlit)

streamlit run app.py

Open http://localhost:8501, enter a research topic, and click Run Research Pipeline. Watch each agent step execute live and get the final report with critic feedback.

CLI

python pipeline.py

Enter a topic at the prompt — the pipeline runs all four steps and prints results to the terminal with rich formatting.

---

Configuration

Setting	Where	Default
LLM model	agents.py	mistral-small-2603
LLM temperature	agents.py	0
Search results	tools.py	max_results=5
Scrape char limit	tools.py	3000 chars
Streamlit theme	.streamlit/config.toml	Dark theme

To switch the LLM provider, replace ChatMistralAI in agents.py with any LangChain chat model (OpenAI, Anthropic, Groq, etc.).

---

Dependencies

• LLM: langchain, langchain-mistralai
• Search: langchain-tavily
• Scraping: beautifulsoup4, requests
• UI: streamlit
• Utils: python-dotenv, tiktoken, rich