LLM-Powered Chatbot

A hands-on project exploring LangChain's core capabilities through two interactive Jupyter notebooks: a conversational chatbot with session memory and a retrieval-augmented generation (RAG) pipeline using vector stores.

Project Structure

LLM-powered-chatbot/
├── chatbots.ipynb          # Conversational chatbot with message history
├── vectorretriever.ipynb   # Vector store, retrievers, and RAG pipeline
├── requirements.txt        # Python dependencies
├── .env                    # API keys (not tracked by git)
└── .gitignore

Notebooks

1. Chatbots (chatbots.ipynb)

Builds a conversational chatbot powered by Groq's Llama 3.1 model. Covers:

• Basic chat — Sending and receiving messages with ChatGroq.
• Message history — Using RunnableWithMessageHistory and ChatMessageHistory to maintain separate conversation sessions via session IDs.
• Prompt templates — Structuring system instructions and user messages with ChatPromptTemplate and MessagesPlaceholder.
• Multi-language support — Parameterizing the prompt to respond in a specified language.
• Conversation trimming — Managing context window limits with trim_messages to keep only the most recent messages.

2. Vector Store & Retrievers (vectorretriever.ipynb)

Demonstrates LangChain's retrieval abstractions for RAG workflows. Covers:

• Documents — Creating Document objects with content and metadata.
• Embeddings — Generating vector embeddings using HuggingFace's all-MiniLM-L6-v2 model.
• Vector store — Storing and querying documents with Chroma (similarity search, scored search, async search).
• Retrievers — Wrapping vector stores as LangChain Runnable objects for use in LCEL chains.
• RAG chain — Combining a retriever, prompt template, and LLM into a full retrieval-augmented generation pipeline.

Prerequisites

• Python 3.12+
• A Groq API key
• A HuggingFace token (for the embeddings model)

Setup

1. Clone the repository

   git clone https://github.com/sothulthorn/LLM-powered-chatbot.git
   cd LLM-powered-chatbot

2. Create and activate a virtual environment

   python -m venv .venv
   # Windows
   .venv\Scripts\activate
   # macOS / Linux
   source .venv/bin/activate

3. Install dependencies

   pip install -r requirements.txt

4. Configure environment variables

   Create a .env file in the project root:

   GROQ_API_KEY=your_groq_api_key
   HF_TOKEN=your_huggingface_token
   

5. Run the notebooks

   jupyter notebook

   Open either chatbots.ipynb or vectorretriever.ipynb to get started.

Key Technologies

License

This project is for educational purposes.