How aymenfurter/smartrag works
Main System Components
a) Frontend Layer:
• React-based web application
• Key components for different interaction modes (Chat, Voice, Research, Upload)
• Configuration management
b) Backend Services:
• Flask API server
• Multiple specialized modules (compare, ingestion, query, integration)
• Queue processors for indexing and uploads
c) Azure Cloud Services:
• Azure OpenAI Service
• Azure AI Search
• Document Intelligence
• Container Apps
• Blob Storage
• Cognitive Services
Architectural Patterns to Highlight
a) Multi-Agent System:
• Interaction between Researcher and Reviewer agents
• AutoGen framework integration
• Collaborative querying flow
b) Document Processing Pipeline:
• Document ingestion flow
• Indexing queue processing
• PDF processing and enhancement
• Table and image post-processing
c) Query Processing:
• GraphRAG implementation
• Voice processing pipeline
• Research workflow
[User Interface Layer]
[API Gateway Layer]
[Service Layer]
• Ingestion Services
• Query Services
• Comparison Services
• Integration Services
[Agent Layer]
• Researcher Agents
• Reviewer Agents
[Azure Services Layer]
Important Flows to Illustrate
• Document ingestion flow (Upload → Processing → Indexing)
• Query processing flow (User Input → Agent Processing → Response)
• Voice interaction flow (Speech → Text → Processing → Speech)
• Research workflow (Query → Multi-Agent Processing → Synthesis)
Technical Details
Technical elements:
• Container deployment boundaries
• Azure service connections
• Queue processing flows
• Storage interactions
• API endpoints
Component Relationships
Relationships:
• Frontend to API connections
• Service interdependencies
• Agent communication patterns
• Azure service integration points
This system design should reflect SmartRAG's nature as a demonstration tool while showing its sophisticated implementation of RAG concepts, multi-agent systems, and cloud service integration.