AutoMate is a context-aware chatbot designed to integrate with car manuals, providing drivers with real-time guidance and support. By leveraging Large Language Models (LLMs) and Retrieval Augmented Generation (RAG), AutoMate can effectively answer queries related to vehicle warning messages, their meanings, and recommended actions. The chatbot aims to enhance the driving experience by delivering information in a user-friendly manner, with potential integration with text-to-speech software to read responses aloud.
As a proof of concept, we utilize pages from the car manual for the MG ZS. This manual contains crucial information about various car warning messages and their associated actions.
AutoMate employs OpenAI GPT-3.5 for generating responses. The model has been fine-tuned to understand and provide context-aware answers based on the specific content of the car manual.
- Context-Aware Responses: AutoMate retrieves relevant information from the car manual using RAG and generates concise answers based on user queries.
- Integration with Text-to-Speech: Future enhancements will allow the chatbot's responses to be read aloud, improving accessibility for drivers.
- User-Friendly Interface: The chatbot provides a simple interaction model for users to query warning messages and receive guidance.
The architecture of Retrieval Augmented Generation (RAG) combines information retrieval with language generation. Below is a diagram that illustrates this architecture:
- LangChain: A framework for building applications with LLMs, facilitating easy retrieval and generation of context-aware responses.
- OpenAI API: Used for accessing the LLM to generate responses based on the retrieved information from the car manual.
- Chroma Vectorstore: A vector store for storing and retrieving documents related to the car manual.
- Streamlit: A framework for creating interactive web applications to facilitate user interaction with the chatbot.