A chatbot that allows users to interact with PDF documents using natural language queries. It leverages LangChain, Llama, and Pinecone for document retrieval and response generation. Gradio is used for the frontend interface, and FastAPI is used for the backend with Llama as the language model.
- Upload PDF files and interact with them using natural language.
- Utilizes LangChain for efficient document parsing and query processing.
- Uses Llama as the language model for generating responses.
- Stores and retrieves document embeddings using Pinecone.
- Provides a web-based interface powered by Gradio.
- Backend built using FastAPI to serve requests efficiently.
- Deployed using Kubernetes for scalability.
- Gradio - Provides a user-friendly web-based interface.
- FastAPI - Handles API requests and serves the chatbot.
- Llama - Generates responses for user queries.
- LangChain - Facilitates document parsing and query processing.
- Pinecone - Manages and retrieves document embeddings.
- Kubernetes - Orchestrates deployment and scaling.
Ensure you have the required API keys and credentials before running the chatbot. Add your secrets in the file kube/secrets-env.yaml with the following format:
PINECONE_API_KEY= <your pinecone api key>
PINECONE_API_ENV= <region of pinecone>
INDEX_NAME= <name of index in pinecone>
HUGGINGFACEHUB_API_TOKEN= <hugging face api key>To set up and deploy the chatbot, follow these steps:
-
Clone the repository:
git clone <repository-url> cd pdf_chatbot
-
Run the setup script to create docker images:
sh bash.sh
-
Deploy the application using Kubernetes:
kubectl apply -k kube/
The chatbot uses a Persistent Volume to store uploaded PDF files, ensuring accessibility across different pods. The configurations are defined in:
The frontend service is exposed using NodePort, making it accessible at:
http://0.0.0.0:<service-node-port>/
Replace <service-node-port> with the actual NodePort value assigned in the Kubernetes service configuration.
