How agora-protocol/paper-demo works
Main Components
A. Agent Layer
• User Agents (/agents/user/)
• Server Agents (/agents/server/)
• Protocol DB Agents (/agents/protocol_db/)
• Common Agent Components (/agents/common/)
B. Model Layer (/models/)
• OpenAI Integration
• Multiple LLM Support (GPT-4, Llama 3 405b, Gemini 1.5 Pro)
C. Toolformers Layer
• Base Toolformer (/toolformers/base.py)
• Specialized Toolformers:
• Negotiator
• Programmer
• Protocol Checker
• Querier
• Responder
D. Database Layer
• MongoDB Integration
• SQL Integration
Rectangle: Core Components
Hexagon: Agents
Rounded Rectangle: LLM Models
Diamond: Toolformers
Blue: Agent components
Green: Database systems
Purple: LLM models
Orange: Toolformers
Gray: Utility/Support components
Key Relationships to Highlight
A. Communication Flow
1. Natural language communication (initial)
2. Protocol negotiation
3. Routine implementation
4. Optimized communication path
B. Integration Points
• How agents interact with:
• Different LLM models
• Database systems
• Specialized toolformers
Architectural Patterns to Emphasize
A. Protocol Negotiation System
• Transition from natural language to structured protocols
• Protocol storage and retrieval mechanisms
B. Multi-Model Architecture
• Demonstrate how different LLM models are unified through the toolformers layer
• Abstraction layer that enables model interoperability
Important Implementation Details
A. Memory Management
• Memory components from server/memory.py and user/memory.py
• How state is maintained across interactions
B. Configuration System
• Config management between different components
• How environment variables and config files are utilized