StoicStocks

StoicStocks is a modern chatbot application that combines rule-based generation with a Retrieval-Augmented Generation (RAG) pipeline. It uses natural language processing to understand user intent and provides responses either from a predefined set or by querying a knowledge base.

Features

• Rule-based intent recognition for common queries
• RAG pipeline for complex queries using Weaviate as a vector database
• Llama3:8b language model running on Ollama for generating responses
• Nomic-embed-text model for text embeddings
• Sleek, responsive UI with real-time streaming of bot responses
• Toggle for context-aware responses

Prerequisites

• Python 3.7+
• Flask
• Docker (for running Weaviate)
• Ollama (for running Llama3:8b and nomic-embed-text)

Installation

1. Clone the repository:

   git clone https://github.com/Hamhunter23/CDSAML-LLM-RAG.git
   cd stoicstocks
   

2. Install the required Python packages:

   pip install flask weaviate-client ollama langchain
   

3. Install and set up Docker: Follow the instructions at Docker Installation Guide

4. Install and set up Weaviate using Docker:

   docker run -d -p 8080:8080 --name weaviate \
     -e AUTHENTICATION_ANONYMOUS_ACCESS_ENABLED=true \
     -e PERSISTENCE_DATA_PATH=/var/lib/weaviate \
     semitechnologies/weaviate:latest
   

5. Install Ollama: Follow the instructions at Ollama Installation Guide

6. Download the required models for Ollama:

   ollama pull llama3:8b
   ollama pull nomic-embed-text
   

Usage

1. Ensure the Weaviate Docker container is running:

   docker start weaviate
   

2. Run the Flask application:

   python app.py
   

3. Open a web browser and navigate to http://localhost:5000.

4. Start chatting with StoicStocks!

Data Ingestion

To populate the Weaviate database with your custom data:

1. Ensure the Weaviate Docker container is running.

2. Run the data ingestion script:

   python charcterSplitter.py
   

This script will chunk the text, generate embeddings using the nomic-embed-text model, and upload the data to Weaviate.

Customization

• Modify the intents and responses dictionaries in app.py to add or change rule-based responses.
• Adjust the chunking parameters in chunk_and_upload.py to optimize for your specific use case.
• Customize the UI by editing templates/index.html.

Troubleshooting

• If you encounter issues with Weaviate, ensure the Docker container is running and accessible on port 8080.
• For Ollama-related problems, check that the required models (llama3:8b and nomic-embed-text) are correctly installed.
• If you face memory issues when running the Llama3:8b model, consider using a smaller model or adjusting Ollama's resource allocation.