📄 Document RAG Chatbot

A production-style Retrieval-Augmented Generation (RAG) application built with FastAPI, LangChain, Groq, HuggingFace Embeddings, and ChromaDB.

The application allows users to upload PDF documents, generate vector embeddings, store them in a persistent Chroma Vector Database, and chat with the uploaded documents using semantic search.

---

🚀 Features

• 📄 Upload PDF documents
• 📖 Extract text from PDFs
• ✂️ Intelligent document chunking
• 🤗 HuggingFace Embeddings
• 🗄️ Persistent Chroma Vector Database
• 🔍 Semantic Similarity Search
• ⚡ Groq LLM Integration
• 💬 Context-Aware Chat
• 🧠 Conversation Memory
• 📚 Swagger API Documentation
• 📝 Structured Logging
• ⚙️ Environment-based Configuration
• 🚀 FastAPI Backend
• 📦 uv Package Management

---

🏗️ Project Architecture

                Upload PDF
                     │
                     ▼
              PDF Upload API
                     │
                     ▼
                Save PDF File
                     │
                     ▼
               PDF Loader
                     │
                     ▼
          Recursive Text Splitter
                     │
                     ▼
       HuggingFace Embeddings
                     │
                     ▼
         Chroma Vector Database
                     │
                     ▼
              Similarity Search
                     │
                     ▼
              Retrieved Chunks
                     │
                     ▼
             Prompt Template
                     │
                     ▼
                 Groq LLM
                     │
                     ▼
              Generated Answer
                     │
                     ▼
          Conversation Memory

---

📂 Project Structure

DOCU-RAG/

│

├── app/

│   ├── config/
│   │     settings.py
│   │
│   ├── memory/
│   │     chat_memory.py
│   │
│   ├── models/
│   │     request_models.py
│   │     response_models.py
│   │
│   ├── prompts/
│   │     prompt.py
│   │
│   ├── routers/
│   │     chat.py
│   │     health.py
│   │     upload.py
│   │
│   ├── services/
│   │     embedding_service.py
│   │     groq_service.py
│   │     pdf_loader.py
│   │     pdf_service.py
│   │     rag_chain.py
│   │     retriever.py
│   │     text_splitter.py
│   │     vector_store.py
│   │
│   ├── utils/
│   │     logger.py
│   │
│   └── main.py
│
├── chroma_db/
├── uploads/
├── logs/
│
├── tests/
│
├── .env
├── .gitignore
├── pyproject.toml
├── uv.lock
└── README.md

---

⚙️ Tech Stack

Category	Technology
Backend	FastAPI
LLM	Groq
Framework	LangChain
Embedding Model	HuggingFace Embeddings
Vector Database	ChromaDB
Document Loader	PyPDFLoader
Text Splitter	RecursiveCharacterTextSplitter
Package Manager	uv
API Documentation	Swagger UI
Configuration	Pydantic Settings
Logging	Python Logging

---

🔄 Workflow

Upload PDF
      │
      ▼
Save PDF
      │
      ▼
Read PDF
      │
      ▼
Split into Chunks
      │
      ▼
Generate Embeddings
      │
      ▼
Store in ChromaDB
      │
      ▼
User Question
      │
      ▼
Generate Query Embedding
      │
      ▼
Similarity Search
      │
      ▼
Retrieve Relevant Chunks
      │
      ▼
Prompt + Context
      │
      ▼
Groq LLM
      │
      ▼
Response

---

📦 Installation

Clone the repository

git clone <repository-url>

cd DOCU-RAG

Create virtual environment

uv venv

Activate virtual environment

Windows

.venv\Scripts\activate

Linux / macOS

source .venv/bin/activate

Install dependencies

uv sync

---

🔑 Environment Variables

Create a .env file in the project root.

GROQ_API_KEY=your_groq_api_key

EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2

CHROMA_DB_PATH=./chroma_db

UPLOAD_FOLDER=./uploads

CHUNK_SIZE=1000

CHUNK_OVERLAP=200

TOP_K_RESULTS=5

LLM_MODEL=llama-3.1-8b-instant

---

▶️ Run the Application

uv run uvicorn app.main:app --reload

Server

http://127.0.0.1:8000

Swagger UI

http://127.0.0.1:8000/docs

---

📌 API Endpoints

Health Check

GET /health

---

Upload PDF

POST /upload

Uploads a PDF document for processing.

---

Chat

POST /chat

Example Request

{
  "question": "What is Artificial Intelligence?"
}

Example Response

{
  "answer": "Artificial Intelligence is..."
}

---

🧠 RAG Pipeline

PDF
 │
 ▼
Upload
 │
 ▼
PyPDFLoader
 │
 ▼
RecursiveCharacterTextSplitter
 │
 ▼
HuggingFace Embeddings
 │
 ▼
ChromaDB
 │
 ▼
Retriever
 │
 ▼
Prompt Template
 │
 ▼
Groq LLM
 │
 ▼
Generated Answer

---

📚 Libraries Used

• FastAPI
• LangChain
• LangChain Community
• LangChain Groq
• LangChain HuggingFace
• LangChain Chroma
• ChromaDB
• Sentence Transformers
• PyPDF
• Pydantic Settings
• Python Dotenv
• Uvicorn

---

🚀 Future Improvements

• Multiple PDF Support
• Source Citation with Page Numbers
• Streaming Responses
• Hybrid Search (BM25 + Vector Search)
• User Authentication
• Conversation Persistence
• Document Deletion API
• Docker Support
• AWS Deployment
• Frontend Integration (React/Next.js)

---

👨‍💻 Author

Kallappa Kabbur

---

⭐ If you found this project useful, consider giving it a star!