Dr Doc

Demo

About

--- 🚀 **Complete Project Description: Dr.Doc** 📋 **Project Overview** Dr.Doc is a groundbreaking hybrid intelligence platform that fuses neural RAG (Retrieval-Augmented Generation) with symbolic MeTTa reasoning, creating the world’s first developer-centric AI assistant that delivers both deep contextual understanding and precise technical accuracy. 🎯 **Mission Statement** "To transform developer productivity by merging the power of neural and symbolic AI, offering accurate, contextual, and verifiable support for API documentation and development workflows." 🎯 **Key Features & Innovations** 🧠 **Hybrid Intelligence (Core USP)** * **Neural RAG**: BGE embeddings + PostgreSQL PgVector for semantic search and understanding * **Symbolic MeTTa**: Structured reasoning for precise pattern matching and logical inference * **Unified Pipeline**: Neural and symbolic layers work together for robust, well-grounded answers 🤖 **ASI:One Integration** * **Fetch.ai ASI:One Mini**: Advanced LLM powering response generation * **Context-Aware Prompting**: Optimized prompts enriched by hybrid intelligence * **Real-Time Processing**: Lightning-fast, sub-second response times 💻 **Developer-First Experience** * **Modern Interface**: Built with Next.js and Tailwind CSS, offering a sleek real-time chat UI * **Rich Formatting**: Markdown support, syntax highlighting, and interactive citations * **Session Control**: Persistent conversation history and backend health monitoring 🔓 **Cost-Optimized Architecture** * **Free Embeddings**: Local BGE model eliminates external API costs * **Self-Contained Stack**: Minimal external dependencies beyond ASI:One * **Scalable Core**: PostgreSQL + PgVector built for production-grade performance ---

How it's made

# 🏗️ Project Architecture Overview ## 📋 **High-Level System Design** **ethNewDelhi2025** follows a **layered architecture** with **hybrid intelligence** combining neural RAG and symbolic MeTTa reasoning. ## 🎯 **Core Architecture Principles** ### **1. Hybrid Intelligence Design** - **Neural Layer**: Semantic understanding and contextual reasoning - **Symbolic Layer**: Precise pattern recognition and logical reasoning - **Unified Interface**: Combines both approaches for superior results ### **2. Microservices Architecture** - **Frontend Service**: Next.js user interface and chat management - **Backend Service**: Python uAgents with HTTP API endpoints - **Data Service**: PostgreSQL with PgVector for knowledge storage - **Intelligence Service**: BGE embeddings + MeTTa reasoning ## 🔧 **System Components** ### **Frontend Layer** - **Technology**: Next.js with React and TypeScript - **Purpose**: User interface, real-time chat, session management - **Communication**: HTTP API calls to backend ### **Backend Layer** - **Technology**: Python with uAgents framework - **Purpose**: Agent orchestration, request processing, response generation - **Features**: HTTP API endpoints, uAgent communication, hybrid intelligence coordination ### **Intelligence Layer** - **Neural Component**: BGE embeddings with vector similarity search - **Symbolic Component**: MeTTa knowledge base with pattern matching - **Integration**: Unified query processing combining both approaches ### **Data Layer** - **Vector Database**: PostgreSQL with PgVector extension - **Document Storage**: Structured text and metadata storage - **Knowledge Base**: MeTTa atoms and pattern definitions ## 🔄 **Data Flow Architecture** ### **Ingestion Pipeline** 1. **Document Processing**: Markdown files converted to structured text 2. **Fact Extraction**: MeTTa patterns extracted from documentation 3. **Embedding Generation**: BGE model creates vector representations 4. **Storage**: Data stored in PostgreSQL with vector indexing ### **Query Processing Pipeline** 1. **User Input**: Natural language questions from frontend 2. **Dual Processing**: Both neural and symbolic systems activated 3. **Context Assembly**: Retrieved documents and patterns combined 4. **Response Generation**: ASI:One processes enhanced context 5. **Output Formatting**: Structured response with citations ## 🧠 **Intelligence Architecture** ### **Neural Intelligence (RAG)** - **Embedding Model**: BGE for semantic understanding - **Vector Search**: Cosine similarity matching in 768-dimensional space - **Strengths**: Contextual understanding, fuzzy matching, semantic search ### **Symbolic Intelligence (MeTTa)** - **Knowledge Base**: Structured facts and patterns as MeTTa atoms - **Pattern Matching**: Logical reasoning and rule-based inference - **Strengths**: Exact matching, logical consistency, verifiable facts ### **Hybrid Integration** - **Parallel Processing**: Both systems query simultaneously - **Context Fusion**: Neural context combined with symbolic patterns - **Enhanced Prompting**: LLM receives both types of intelligence ## 🚀 **Agent Architecture** ### **ASI:One Agent** - **Role**: Primary agent for user interaction and response generation - **Integration**: Direct access to hybrid intelligence systems - **Communication**: HTTP API and uAgent messaging protocols ### **System Integration** - **RAG System**: BGE embeddings + PostgreSQL vector search - **MeTTa System**: Hyperon MeTTa engine for symbolic reasoning - **Unified Processing**: Combined neural and symbolic intelligence ## 🔧 **Infrastructure Architecture** ### **Database Design** - **Primary Database**: PostgreSQL with PgVector extension - **Schema**: Documents table with content, metadata, and vector embeddings - **Scalability**: Horizontal scaling with connection pooling ### **Deployment Architecture** - **Containerization**: Docker containers for consistent deployment - **Service Discovery**: Environment-based configuration - **Monitoring**: Health checks and status monitoring ## 🎯 **Scalability Design** ### **Horizontal Scaling** - **Stateless Services**: Backend services can scale independently - **Load Distribution**: Multiple agent instances for high availability - **Performance Optimization**: Lazy loading, batch processing, connection pooling ## 🔄 **Integration Architecture** ### **External Integrations** - **ASI:One API**: Fetch.ai's LLM service integration - **BGE Model**: Local embedding model for cost efficiency - **Hyperon MeTTa**: Symbolic reasoning engine integration ### **Internal Communication** - **HTTP APIs**: RESTful communication between services - **Agent Messaging**: uAgent protocol for agent-to-agent communication - **Error Handling**: Comprehensive error handling and reporting ## 🎪 **User Experience Architecture** ### **Real-Time Interaction** - **WebSocket Support**: Real-time chat interface - **Session Management**: Persistent user sessions - **Status Monitoring**: Live backend status and connection monitoring ### **Response Architecture** - **Rich Formatting**: Markdown rendering with syntax highlighting - **Citation System**: Clickable links to source documentation - **Accessibility**: Screen reader support and keyboard navigation This architecture enables **hybrid intelligence** through a **scalable, maintainable platform** that combines neural and symbolic AI approaches.