Super-Accurate-RAG-based-Chatbot-for-Personal-Devices

Inspiration

In a world where documents can be overwhelmingly long, the dream is to have your computer sift through the content for you. That's the vision behind the Super-Accurate-RAG-based-Chatbot-for-Personal-Devices - a tool designed to make life easier by digesting extensive information on your behalf.

Unique Advantages

Why choose this over others?

• Minimal Hardware: Operates smoothly on just 8GB of RAM - no hefty hardware needed.
• Contextual Clarity: Unlike other chatbots that may struggle with context and produce irrelevant responses, this chatbot is engineered to provide precise and relevant answers.

How It Works

1. Topic Detection: Identifies key topics within documents (currently in development).
2. Intelligent Decision Making: Determines whether to answer queries based on the document content or a web search.
3. Data Retrieval: Searches for relevance information within the document or on the web.
4. Relevance Check: Evaluates the relevance of the information retrieved.
5. Answer Generation: Crafts responses while self-checking for hallucinations and relevance.
6. Answer Validation: Delivers the answer if relevant; otherwise, it reinitiates the search from step 3.

Frameworks Utilized

• Langchain's Langraph
• llama.cpp served using Ollama
• ChromaDB
• tavily (for web search)
• GPT4all (for document embeddings)

Model

This chatbot is powered by Meta's LLama 3, 4-bit quantized to GGUF.

See It in Action

TODO

• Add Topic Detection Capabilities
• Implement Graph RAG
• Implement MultiModal Capabilities
• Implement Conversation History
• Complete this README.md 😁