Overview
Avocado Health is an AI copilot that provides accurate, personalized health information by leveraging PubMed articles and advanced AI techniques. It combines RAG (Retrieval Augmented Generation) with strict guardrails to ensure reliable, scientifically-backed responses while maintaining accessibility for different target audiences.
Key Features
- Scientific Accuracy: All responses are based on peer-reviewed PubMed articles, with direct citations and references
- Personalized Explanations: Content is automatically adapted based on the target audience's age and background
- Custom Knowledge Base: Users can expand the system's knowledge by adding their own PubMed articles
- Anti-Hallucination Measures: Strict guardrails ensure responses stick to verified medical information
- Source Transparency: Every response includes links to the original PubMed articles used
Technical Implementation
RAG Architecture
- PubMed articles are vectorized and stored in a Pinecone database
- User queries trigger semantic search to find relevant article sections
- Retrieved context is used by the LLM to generate accurate responses
- Citations are automatically included with each response
Guardrails Implementation
- Strict context adherence - responses must be based on provided PubMed articles
- Citation requirements - every claim must reference a specific source
- Medical terminology validation
- Confidence thresholds for retrieved content
Technology Stack
- Next.js 14 with App Router
- LangChain for RAG implementation
- Pinecone for vector database storage
- OpenAI API for content generation
- Guardrails for hallucination prevention
- TypeScript for type safety