ManthanAI - Research Assistant

ManthanAI is a lightweight research assistant designed to simplify web research using state-of-the-art AI models like LLaMA (structured output) and Google Gemini (research report generation). Built on the LangGraph framework, it provides real-time streaming updates, allowing you to watch your research unfold live.

Inspired by GPT Researcher, ManthanAI focuses on generating search queries, gathering and summarizing web data, and providing real-time, streaming updates as the research progresses.

Getting Started 🚀

Give it a go 💡

Graph 🦜

Here’s a look at the research flow:

• Safety Check: Ensures input topic safety before proceeding.
• Generate Queries: Generates search queries based on the topic.
• Conduct Research: For each query,:
  • Search Web: Gathers relevant data from the web.
  • Generate Research Summary: Compiles findings into a concise summaries.
• Write Report: Final research report is prepared.

Key Highlights ✨

• Backend: FastAPI backend manages research workflows and streaming.
• Frontend: Next.js and shadcn/ui offers a clean and interactive experience.
• AI-Powered Research: Uses LangGraph with LLaMA 70B and Google Gemini for web research.
• Safety First: Implemented content safety checks to ensure safe research topics.
• Real-Time Progress: Follow research updates and outcomes as they happen through the streaming API.
• Asynchronous Design: Handles multiple research tasks concurrently using asyncio for better performance.
• User Feedback: Allows users to upvote or downvote research results, stored with LangSmith.

Setup ⚙️

1. Clone the Repository:

git clone https://github.com/git-vish/manthan.git
cd manthan

Backend 🧩

2. Set Up Virtual Environment & Install Dependencies:

   • pip:

     cd backend
     python -m venv .venv
     source .venv/bin/activate
     pip install -r requirements.txt 

   • uv:

     pip install uv
     cd backend
     uv sync --frozen
     source .venv/bin/activate

3. Set Up Environment Variables:
   Copy .env.example to .env and fill in the required fields.

cp .env.example .env

4. Run the Backend:

uvicorn src.api:app --reload

Frontend 🖥️

5. Install Frontend Dependencies:

cd ../frontend
npm install

6. Set Up Environment Variables:
   Copy .env.local.example to .env.local and configure API details:

cp .env.local.example .env.local  

7. Run the Frontend:

npm run dev

License 📜

This project is licensed under the MIT License. Check the LICENSE file for more details.

Feedback & Contributions 🙌

Your feedback and contributions are welcome! Please feel free to open an issue or submit a pull request.