🌌 Impactoverse AI Mentor

An AI-powered chat mentor that helps users interact with educational PDFs and documents. Built using DeepSeek API, FAISS, and Docling for contextual PDF understanding. Combines two separate backends: one for open-source AI models and another for MongoDB-based data management.

---

🔐 Environment Setup

Create a .env file in the Node.js backend and add the following:

DEEPSEEK_API_KEY=your_deepseek_api_key MONGO_URI=your_mongodb_connection_string

🚀 Tech Stack

Frontend:

• React

Backend:

• Node.js (API & MongoDB)
• FastAPI (Open-source models)
• Docling, FAISS, SentenceTransformers

AI & NLP:

• DeepSeek API
• Sentence Transformers
• FAISS (Facebook AI Similarity Search)
• Docling for parsing PDFs

---

🛠️ Installation

🔧 FastAPI Backend for AI Models

1. Navigate to the backend folder:

   cd backend

2. Create a virtual environment:

   python3.11 -m venv parse

3. Activate the virtual enviornment

   On Windows parse\Scripts\activate

   On macOS/Linux source parse/bin/activate

4. Install dependencies:

   pip install fastapi uvicorn faiss-cpu sentence-transformers docling python-multipart

5. Run the FastAPI server:

   uvicorn app.app:app --reload

   This will start the backend server on http://localhost:8000

🗃️ Node.js + MongoDB Backend

1. Navigate to the server directory:

   cd backend npm install

2. Run MongoDB server

   node server.js

⚠️ Ensure MongoDB is installed and running locally or provide a cloud URI.

💻 React Frontend

1. Navigate to the frontend directory: cd frontend

2. Install dependencies npm install

3. Start the frontend server: npm run dev

🌟 Features

• AI Mentor using DeepSeek API

• Intelligent PDF parsing with Docling

• Vector embeddings via Sentence Transformers

• Semantic search using FAISS

• MongoDB integration for chat and user history

• Modular backend architecture (FastAPI + Node.js)