Custom Chatbot with LangChain and Flask

This project is a custom chatbot that leverages LangChain, HuggingFace embeddings, and FAISS vector storage to provide conversational responses based on data extracted from a website. The chatbot is exposed as a RESTful API using Flask, making it easy to integrate into other applications.

---

Features

• Web Scraping: Extracts data from a specified URL using LangChain's WebBaseLoader.
• Embeddings: Uses HuggingFace embeddings to convert text into vector representations.
• Vector Storage: Stores embeddings in a FAISS vector store for efficient similarity search.
• Conversational Memory: Maintains conversation history using LangChain's ConversationBufferMemory.
• RESTful API: Exposes the chatbot as a Flask API for easy integration.
• Customizable: Supports different HuggingFace models and embedding configurations.

---

Technologies Used

• LangChain: For building the conversational chain and handling embeddings.
• HuggingFace Transformers: For generating embeddings and powering the chatbot's language model.
• FAISS: For efficient vector storage and similarity search.
• Flask: For creating the RESTful API.
• Python: The core programming language.

---

LLM Used

The chatbot uses HuggingFace's google/flan-t5-large as its language model. This model is fine-tuned for text generation tasks and provides high-quality responses for conversational AI.

---

Installation

Step 1: Clone the Repository

git clone https://github.com/your-username/your-repo-name.git
cd your-repo-name

Step 2: Install Dependencies

Install the required Python packages:

pip install -r requirements.txt

Step 3: Set HuggingFace API Token

1. Go to HuggingFace Hub and generate an API token.
2. Set the token as an environment variable:
   • Windows:

     set HUGGINGFACEHUB_API_TOKEN=your_api_token_here

   • Linux/macOS:

     export HUGGINGFACEHUB_API_TOKEN=your_api_token_here

---

Usage

Step 1: Run the Flask Server

Start the Flask server:

python app.py

The server will start at http://127.0.0.1:5000.

Step 2: Test the API

You can interact with the chatbot using curl, Postman, or any HTTP client.

Example Request

Send a POST request to the /chat endpoint with a JSON payload:

curl -X POST http://127.0.0.1:5000/chat \
-H "Content-Type: application/json" \
-d '{"message": "What technical courses are available?"}'

Example Response

{
  "response": "The available technical courses include Python programming, Machine Learning, and Data Science."
}

---

How It Works

Step 1: Data Extraction

The chatbot extracts data from a specified URL using LangChain's WebBaseLoader.

Step 2: Embeddings

The extracted data is converted into embeddings using HuggingFace's sentence-transformers/all-mpnet-base-v2.

Step 3: Vector Storage

The embeddings are stored in a FAISS vector store for efficient retrieval.

Step 4: Conversation Chain

A conversational chain is initialized using HuggingFace's google/flan-t5-large model and the FAISS vector store.

Step 5: API Endpoint

The chatbot is exposed as a RESTful API using Flask. Users can send messages to the /chat endpoint and receive responses.

---

Customization

Change the Embedding Model

Modify the model_name parameter in the create_vector_store function:

embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")

Change the Language Model

Replace the HuggingFace model (google/flan-t5-large) with any other supported model:

llm = HuggingFaceEndpoint(
    repo_id="another-model-name",
    huggingfacehub_api_token=huggingfacehub_api_token,
    temperature=0.7,
    max_length=512
)

Add More Endpoints

Extend the Flask API by adding more routes in app.py.

---

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

---

License

This project is licensed under the MIT License. See the LICENSE file for details.

---

Screenshots

---