ML Agent
The ML Agent is an intelligent component that processes inputs and generates responses based on configured language models and tools. It provides a flexible interface for natural language interactions with customizable behaviors and capabilities.
ML Agent interface and configuration options
Configuration Note: Ensure proper configuration of all required fields including Template, Language Model, and Agent Instructions for optimal performance.
Component Configuration
- Template: Message template configuration
Format: {{sender_name}}: {{text}}
- Language Model: The AI model to use for processing
Select from available language models
- Agent Instructions: Behavioral guidelines for the agent
Detailed instructions for agent behavior and responses
- Tools: Available tools and capabilities
Configure tools the agent can use
Advanced Settings
- Handle Parse Errors: Error handling configuration
- Verbose: Detailed logging options
- Max Iterations: Maximum number of processing iterations (default: 15)
- External Memory: Configuration for persistent memory
- Sender Type: Type of sender (Machine and User)
- Number of Messages: Message history limit (default: 100)
- Session ID: Unique identifier for conversation sessions
- Order: Message ordering preference (Ascending)
Implementation Example
const mlAgent = new MLAgent({
template: "{sender_name}: {text}",
languageModel: "gpt-4",
agentInstructions: "You are a helpful assistant...",
tools: ["search", "calculate", "fetch"],
maxIterations: 15,
senderType: "Machine and User",
numberOfMessages: 100,
order: "Ascending"
});
const response = await mlAgent.process({
input: "Hello, can you help me?",
sessionId: "user-session-123"
});
// Output:
// {
// response: "Hello! Of course, I'd be happy to help...",
// metadata: {
// toolsUsed: [],
// processingTime: "0.5s",
// confidence: 0.95
// }
// }Additional Resources
Best Practices
- Provide clear and specific agent instructions
- Configure appropriate error handling
- Monitor and limit iterations
- Maintain session context appropriately
- Regular performance monitoring