ZeroTang05/cyber-doctor

What it solves

Cyber-doctor addresses the imbalance of medical resources by providing an AI-powered health assistant. It helps users perform basic disease diagnosis, analyze medical records, and answer professional health questions, reducing the need for people in underdeveloped areas to travel to major cities for basic medical guidance.

How it works

The project integrates multiple multimodal large language models (LLMs) and uses an AI agent to determine the task type and coordinate different models. It employs several enhancement techniques to improve accuracy:

• RAG (Retrieval-Augmented Generation): Uses local knowledge bases (files) and web crawling to provide up-to-date and specific information.
• Knowledge Graph Integration: Connects to Neo4j graph databases to incorporate structured professional domain knowledge.
• Multimodal Processing: Integrates STT (Whisper) and TTS (edge-tts) for voice interaction, and uses multimodal LLMs for image and text recognition (e.g., analyzing medical reports).
• Tool Integration: Includes modules for generating formatted Word and PPT documents based on AI output.

Who it’s for

• Individuals seeking basic health management and medical advice.
• Users who prefer voice-based interaction over typing.
• Developers looking for a template to build domain-specific expert systems using RAG and Knowledge Graphs.

Highlights

• Multimodal Interaction: Supports text, image, audio, and video inputs/outputs.
• Hybrid Knowledge Retrieval: Combines traditional RAG with Neo4j knowledge graphs and real-time internet search.
• Voice-First Mode: Dedicated voice dialogue module for lower accessibility barriers.
• Document Automation: Automatically converts AI-generated health content into professional PPT and Word formats.