Overview
ADAS Architecture Overview
System Architecture
Core Components
1. Coordinator Agent
The Coordinator Agent is the central orchestrator of the ADAS system, managing communication and coordination between all other agents. It is implemented using n8n, a powerful workflow automation platform.
Key Responsibilities:
Route user requests to appropriate specialized agents
Manage conversation context and history
Coordinate responses from multiple agents
Handle error recovery and fallback mechanisms
Provide monitoring and logging
Implementation:
n8n workflows define the coordination logic
Webhook endpoints for receiving user requests
HTTP Request nodes for communicating with ElizaOS agents
Function nodes for processing and transforming data
Telegram integration for user interaction
2. Analytics Agent
The Analytics Agent is responsible for gathering and analyzing data from multiple sources to provide insights about the Aptos ecosystem.
Key Responsibilities:
Fetch real-time price data from CoinGecko
Retrieve TVL data from DefiLlama
Track protocol performance metrics
Analyze market trends
Generate portfolio analytics
Implementation:
ElizaOS agent with custom character definition
Data connector interfaces for external APIs
Processing utilities for data transformation
REST API endpoints for serving analytics data
RAG capabilities for contextual information retrieval
3. DeFi Agent
The DeFi Agent handles all on-chain operations on the Aptos blockchain, leveraging the Move Agent Kit for interacting with various protocols.
Key Responsibilities:
Execute token transfers
Perform swaps on DEXs (Liquidswap, Thala)
Manage lending positions on Joule Finance (deposit, borrow, repay, withdraw, claim rewards)
View detailed position information and pool statistics
Handle staking operations on Amnis Finance (stake and unstake APT)
Execute trading operations on Merkle Trade (limit orders, market orders, position management)
Manage lending and borrowing on Aries Protocol (profile creation, lending, borrowing, repayment)
Monitor transaction status
Display formatted portfolio information with proper token recognition
Implementation:
ElizaOS agent with custom character definition
Enhanced Aptos Plugin for blockchain interaction
Move Agent Kit integration for protocol-specific operations
Transaction utilities for building and signing transactions
REST API endpoints for executing operations
4. Aptos Expert Agent
The Aptos Expert Agent provides specialized knowledge and guidance about the Aptos ecosystem, helping users navigate protocols and make informed decisions.
Key Responsibilities:
Provide information about Aptos protocols
Explain DeFi concepts and strategies
Offer recommendations based on market conditions
Answer technical questions about Aptos
Educate users about best practices
Implementation:
ElizaOS agent with custom character definition
Comprehensive knowledge base of Aptos documentation
RAG system for accurate information retrieval
REST API endpoints for knowledge queries
Recommendation system based on user context
Integration Points
ElizaOS
ElizaOS serves as the foundation for all specialized agents, providing:
Character-based agent definitions
Plugin system for extending functionality
Multi-client support (web, Telegram, etc.)
LLM provider integration (Anthropic Claude, OpenAI)
Conversation management
Aptos Plugin v2
The Aptos Plugin v2 is the core integration for blockchain interactions, offering:
Production-ready implementation for multiple Aptos protocols
Token operations (transfer, mint, burn)
DEX operations (swaps, liquidity)
Lending operations (deposit, borrow, repay, withdraw, position viewing)
Staking operations
Trading operations (orders, position management)
Comprehensive protocol-specific implementations
n8n
n8n provides the workflow automation for agent orchestration:
Visual workflow builder
Webhook endpoints for API access
HTTP requests for inter-agent communication
Function nodes for data processing
Error handling and retries
Web Client
The web client offers a user interface for interacting with the agent swarm:
Chat interface for natural language interaction with Markdown support
Portfolio dashboard for tracking assets with formatted displays
Analytics visualizations
Agent directory for discovering capabilities
Settings management
Persistent chat history
Example prompts for user guidance
Data Flow
User sends a request via web interface or Telegram
Coordinator Agent receives the request and analyzes intent
Coordinator routes the request to appropriate specialized agent(s)
Specialized agent processes the request and returns a response
Coordinator aggregates responses and formats the final output
User receives the response with requested information or confirmation of action
Security Considerations
Private keys are stored securely and never exposed
Environment variables are used for sensitive configuration
API keys are managed with appropriate access controls
Transactions require explicit user confirmation
Rate limiting is implemented to prevent abuse
Scalability
The multi-agent architecture allows for horizontal scaling:
New specialized agents can be added for additional functionality
Existing agents can be replicated for handling increased load
Stateless design enables distributed deployment
Modular components can be scaled independently
User Experience Improvements
The ADAS interface has been enhanced with several user experience improvements:
Markdown support for formatted responses
Clean presentation of portfolio data with proper token recognition
Persistent chat history that survives page reloads
Example prompts to guide new users
Responsive design for various screen sizes
Future Extensions
The architecture is designed to accommodate future extensions:
Additional specialized agents for specific protocols
Cross-chain capabilities for interoperability
Advanced AI strategies for portfolio optimization
Mobile client integration
Community-contributed agents and plugins
Last updated