RAG Chatbot using LangChain

Project description

An AI chatbot with Retrieval-Augmented Generation (RAG) capabilities, supporting document uploads (PDF/DOCX/HTML), with usage of LLaMA models, and chat history persistence.

Project is based on an example described in this article.

---

Features

• Document upload and indexing
• Content-aware chat with history
• Support for LLaMA (via Ollama)
• Document deletion and management
• SQLite logging for auditability

---

Tech Stack

Component	Technology
Backend	FastAPI
Frontend	Streamlit
Vector DB	ChromaDB
NLP	LangChain + HuggingFace/HFEmbeddings
Models	LLaMA 3, LLaMA 2
Database	SQLite3

---

Project structure

├── application_api/ # FastAPI backend
│ ├── utils/ # Chroma/LangChain utilities
│ ├── model/ # Pydantic validation models
│ └── api.py # API endpoints
├── ui/ # Streamlit frontend
│ ├── components/ # UI components
│ ├── utils/ # UI utils
│ └── streamlit_app.py # Streamlit app
├── requirements.txt # Dependencies
└── main.py # Streamlit app entry point

API Endpoints

Endpoint	Method	Description
/	POST	Submit chat queries
/upload-doc	POST	Upload/index documents
/list-docs	GET	List uploaded documents
/delete-doc	POST	Delete document by ID

---

Setup & Installation

1. Clone repository

git clone https://github.com/oden73/RAG_Chatbot_with_LangChain.git
cd RAG_Chatbot_with_LangChain

2. Create virtual environment

python -m venv venv
source venv/bin/activate  # For Linux/Mac
venv\Scripts\activate     # For Windows

3. Install dependencies

pip install -r requirements.txt

4. Required services

1. Install Ollama
2. Launch Ollama with ollama serve
3. Install LLaMA models with ollama pull <model_name>

Usage

Start Backend

uvicorn application_api.api:app --reload --port 8000

Start Frontend

streamlit run main.py