Our Result
GIF
Our Result

📄 AI Document Search (RAG Chatbot)

Chat with your PDF documents using an AI-powered chatbot.
This project uses Retrieval-Augmented Generation (RAG) so the answers are based on your uploaded files, not just the model’s memory.

✨ Features

📂 Upload PDF documents for ingestion
🔎 Semantic search with embeddings (finds meaning, not just keywords)
💬 Ask natural questions and get accurate, context-aware answers
📑 Source citations from your original documents
⚡ Lightweight Frontend (HTML, CSS, JS) + FastAPI backend
🧠 Powered by Ollama LLM + LangChain
📦 Vector database with FAISS (local)

🛠️ Tech Stack

Frontend: HTML, CSS, JavaScript
Backend: FastAPI (Python)
AI Model: Ollama (LLM) + LangChain (retrieval & QA chain)
Vector DB: FAISS (default)
Deployment: Docker (backend), Vercel/Static hosting (frontend)

⚙️ How It Works

Upload PDF → Extract text with LangChain loaders
Chunk text → Split into smaller sections for better retrieval
Embed chunks → Convert into vectors using Ollama embeddings
Store vectors → Save in FAISS database
Ask a question → Query is embedded and compared to stored vectors
Retrieve top matches → Most relevant document chunks are selected
Generate answer → Ollama LLM forms a response using retrieved chunks
Return results → Answer is displayed in the chatbot UI

📂 Project Structure

Built With

Submitted to

Maximally AI Shipathon

Created by

Backend Development – AI Document Search (RAG Chatbot)

Designed and implemented the FastAPI backend to handle PDF uploads, ingestion, and retrieval of information.

Integrated FAISS vector database to store and query document embeddings for semantic search.

Developed document parsing and embedding pipelines to convert uploaded PDFs into searchable vectors.

Built retrieval-augmented generation (RAG) workflows using LangChain to fetch relevant information from documents and query the Ollama LLM for context-aware responses.

Implemented endpoints for file upload, question querying, and JSON responses with accurate source citations.

Ensured robust error handling for file uploads and query processing to maintain backend stability.

Mihirkant Pradhan
Full-Stack Contribution – AI Document Search (RAG Chatbot)

Backend (FastAPI + Python + FAISS + LangChain):

Built the core FastAPI backend to handle PDF uploads, ingestion, and query processing.

Developed document parsing pipelines to extract text from PDFs and generate embeddings for semantic search.

Integrated FAISS vector database to enable efficient retrieval of relevant content from large document sets.

Implemented RAG workflows using LangChain and Ollama LLM to provide context-aware answers with source citations.

Designed robust API endpoints for file uploads, question queries, and structured JSON responses.

Ensured backend stability, error handling, and scalability for multiple concurrent users.

Frontend (HTML + CSS + JavaScript):

Created a lightweight, responsive, and intuitive UI for uploading PDFs and interacting with the AI chatbot.

Developed features for question input, real-time responses, and displaying source references.

Implemented dynamic content rendering and user notifications for smooth interaction.

Focused on user experience, ensuring semantic search results are displayed clearly and accurately.

Integration (Frontend ↔ Backend):

Connected the frontend with backend APIs for real-time document ingestion and query response.

Managed API requests and responses, including error handling, loading states, and formatting outputs.

Tested and debugged end-to-end workflows to ensure seamless communication between frontend, backend, and AI model.

Optimized the system to provide fast, reliable, and accurate semantic search for end-users.

Abhinav kumar
This project involved the development of a Retrieval-Augmented Generation (RAG) chatbot for intelligent document search, with contributions spanning both backend and frontend. On the backend, contributed to the design and development of a FastAPI service that managed PDF ingestion, text extraction, and embedding generation to support semantic search. Supported the integration of a FAISS vector database for efficient retrieval and contributed to implementing RAG workflows using LangChain and the Ollama language model to deliver context-aware responses with reliable source citations. Additionally, managed the creation of API endpoints for file uploads and query handling, focusing on stability, scalability, and error resilience.

On the frontend, contributed to building a responsive and user-friendly interface that allowed seamless document uploads, query inputs, and display of AI-generated responses with proper references. Managed aspects of frontend–backend integration by ensuring smooth API communication, error handling, and real-time responsiveness. These efforts collectively enhanced both the technical robustness and the overall user experience of the system.

Ankur Saini

Updates

Ankur Saini started this project — Aug 31, 2025 02:29 PM EDT

Leave feedback in the comments!

Log in or sign up for Devpost to join the conversation.