🧠 ResearchMind — Multi-Agent AI Research System

Four specialized AI agents work in concert — searching, scraping, writing, and critiquing — to deliver polished, in-depth research reports on any topic.

ResearchMind is a premium, state-of-the-art multi-agent research assistant built using Streamlit, LangChain, and Google's Gemini 2.5 Flash. By combining the power of web search tools, scrapers, and structured LLM chains, it automates the laborious research lifecycle, transforming a simple prompt into an academic-grade report with critical evaluation scores.

---

🗺️ System Architecture & Workflow

ResearchMind orchestrates four distinct intelligence units to perform deep investigation:

graph TD
    User([User Prompt]) --> SearchAgent[🔎 1. Search Agent<br/>Tavily Search API]
    SearchAgent --> ReaderAgent[📄 2. Reader Agent<br/>BeautifulSoup Web Scraper]
    ReaderAgent --> WriterChain[✍️ 3. Writer Chain<br/>Gemini 2.5 Flash]
    WriterChain --> CriticChain[⚖️ 4. Critic Chain<br/>Gemini 2.5 Evaluation]
    CriticChain --> UI[[🧠 ResearchMind UI]]

Loading

1. 🔎 Search Agent: Generates tailored search queries, scans the web using the Tavily Search API, and retrieves reliable metadata (Titles, Snippets, URLs). It includes automatic query sanitization and fallback fallback-logic.
2. 📄 Reader Agent: Selects the most relevant resource and extracts up to 3,000 characters of clean text (excluding headers, footers, and scripts) using BeautifulSoup.
3. ✍️ Writer Chain: Consolidates all gathered web snippets and scraped text to draft a comprehensive, structured markdown report (Introduction, Key Findings, Conclusion, Sources).
4. ⚖️ Critic Chain: Acts as an objective reviewer, evaluating the report for factual consistency and depth, scoring it out of 10, outlining strengths and areas for improvement, and providing a final verdict.

---

✨ Key Features

• 💎 Premium Glassmorphic UI: Beautiful dark theme styled with custom Google Fonts (Space Grotesk, Inter, JetBrains Mono), ambient glowing gradient backdrops, animated noise overlays, and floating background orbs.
• 📡 Real-Time Agent Progress Tracker: Visual progress bar showing which agent is currently active, complete with step logs, status indicators, and execution time tracking.
• 💻 Dual Execution Modes:
  • Web App: Complete interactive experience via Streamlit (app.py).
  • CLI Pipeline: Rapid command-line research execution (pipline.py).
• ⚙️ Smart Search Sanitizer: Automatically cleans queries containing unsupported search operators (like site:) and retries searches automatically to ensure high reliability.
• 🔒 Secure Secret Fallbacks: Seamlessly switches between local .env variables and production Streamlit Cloud Secrets (st.secrets).

---

🛠️ Technology Stack

• Front-End: Streamlit with custom CSS styling.
• Orchestration: LangChain & langchain-core for agent tools and chain piping.
• LLM Service: Google Gemini 2.5 Flash via langchain-google-genai (optimized for speed and free-tier compatibility).
• Search: Tavily Search API for developer-friendly search results.
• Scraping: BeautifulSoup4 + requests + lxml.

---

📁 Project Structure

├── .streamlit/
│   └── config.toml       # Streamlit dark theme & layout configuration
├── agents.py             # Agent definitions & LangChain setup
├── app.py                # Premium Streamlit web UI & orchestration
├── pipline.py            # Command Line Interface (CLI) version
├── tools.py              # Custom tools (web_search, scrape_url)
├── requirements.txt      # Project dependencies
└── README.md             # Project documentation (this file)

---

🚀 Getting Started

Prerequisites

• Python 3.9 or higher
• A Google Gemini API Key (Get one from Google AI Studio)
• A Tavily Search API Key (Get one from Tavily)

Installation

1. Clone the Repository:

   git clone https://github.com/prathamesh-6099/Multi-Agent-AI-Research-System.git
   cd Multi-Agent-AI-Research-System

2. Create and Activate Virtual Environment:

   • Windows:

     python -m venv .venv
     .venv\Scripts\activate

   • macOS / Linux:

     python3 -m venv .venv
     source .venv/bin/activate

3. Install Dependencies:

   pip install -r requirements.txt

4. Set Up Environment Secrets: Create a file named .env in the root of the project and add your API keys:

   GEMINI_API_KEY=your_gemini_api_key_here
   TAVILY_API_KEY=your_tavily_api_key_here

---

🖥️ How to Run

Option 1: Streamlit Web UI (Recommended)

Run the Streamlit application to experience the high-end dashboard:

streamlit run app.py

This will spin up a local server, usually opening automatically at http://localhost:8501.

Option 2: Command Line (CLI)

For a text-based, lightning-fast execution:

python pipline.py

You will be prompted to enter a topic, and the research stages will print directly in your terminal.

---

☁️ Streamlit Cloud Deployment

To host this application for free on Streamlit Community Cloud:

1. Push the project to your GitHub repository.
2. Log in to Streamlit Share and click New app.
3. Select your repository, branch (main), and main file path (app.py).
4. Expand the Advanced settings section.
5. In the Secrets text area, paste your production API keys in TOML format:

   GEMINI_API_KEY = "your-actual-gemini-api-key"
   TAVILY_API_KEY = "your-actual-tavily-api-key"

6. Click Deploy! The app will be built and launched online.

---

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.