DocChatAI: RAG Chatbot & Report Generator

A Retrieval-Augmented Generation (RAG) application built with:

• Streamlit
• LangChain
• Gemini API, Model: Gemini-2.5-Flash
• FAISS
• Sentence Transformers

Features

• Upload PDFs and ask questions
• Context-aware answers using RAG
• Generate structured summary reports
• Download reports as text files

Tech Stack

• Python 3.10
• Streamlit
• LangChain
• FAISS
• HuggingFace Embeddings

Evaluation

Retrieval performance was evaluated using a curated set of question–document pairs on user-uploaded PDFs.

• Top-3 Retrieval Accuracy: 100%

• Embedding Model: sentence-transformers/all-MiniLM-L6-v2

• Vector Store: FAISS

• A query is considered correct if the expected document appears among the top-3 retrieved chunks.

• Gemini API

Setup

pip install -r requirements.txt
streamlit run app.py