RAG QA Chatbot with Memory

A Question-Answering chatbot built using RAG (Retrieval-Augmented Generation) with conversation memory. This project uses LangChain, various LLM options, and vector stores to create an intelligent chatbot that can answer questions about Jessup Cellars winery.

Features

• RAG-based question answering
• Conversation memory to maintain context
• Support for multiple LLM options (Groq, OpenAI)
• Vector store options (Pinecone, FAISS)
• Environment variable configuration
• Flexible embedding models (HuggingFace, OpenAI)

Installation

1.Clone the repository:

git clone https://github.com/Piyush-sri11/QA-chatbot.git

2.Create and activate a virtual environment:

virtualenv env
env/Scripts/activate

3.Install required packages:

pip install -r requirements.txt

Configuration

1.Create a .env file in the project root and add your API keys:

LANGCHAIN_API_KEY="your_langchain_api_key"
LANGCHAIN_PROJECT="RAG QA Chatbot with Memory"
OPENAI_API_KEY="your_openai_api_key"
GOOGLE_API_KEY="your_google_api_key"
GROQ_API_KEY="your_groq_api_key"
HUGGING_FACE_TOKEN="your_huggingface_token"
PINECONE_API_KEY="your_pinecone_api_key"

For Google Colab, use secrets to store API keys:

from google.colab import userdata
os.environ['OPENAI_API_KEY'] = userdata.get('OPENAI_API_KEY')
os.environ['GROQ_API_KEY'] = userdata.get('GROQ_API_KEY')
# Add other API keys as needed

LLM Options

Default Configuration (Open Source)

The project uses Groq's Llama3-70b-8192 model by default, as it's a powerful open-source alternative:

llm = ChatGroq(
    groq_api_key=groq_api_key,
    model_name="Llama3-70b-8192",
    temperature=0
)

OpenAI Configuration (Commented Out)

To use OpenAI's models (requires API credits):

llm = ChatOpenAI(
    model="gpt-4",
    temperature=0,
    max_tokens=None,
    timeout=None,
    max_retries=4,
    api_key=openai_api_key,
)

Vector Store Options

Pinecone (Default)

The project uses Pinecone as the default vector store for production use.

FAISS (Local Alternative)

For local development or testing, uncomment the FAISS implementation in the code.

Running the Project

• 1.Ensure all configurations are set up properly
• 2.Run the main script:

python yard.py

• 3.Start asking questions about Jessup Cellars.
• 4.Type 'exit' to end the session.

Sample Usage

Enter your question: What makes Jessup Cellars wines special?
[Response will appear here]
Response time: [time in seconds]
##################################################
Enter your question:

Note on LLM Choice

• This project was developed using open-source LLMs (Groq's Llama3-70b-8192) due to OpenAI API credit limitations. The code includes commented sections for OpenAI integration if you have API credits available.

Corpus Information

• See corpus_info.md for detailed information about the Jessup Cellars knowledge base used in this project.