Overview

Learn how Handit.AI automatically optimizes your AI agents through continuous learning and adaptation.

Automatic Prompt Optimization

Handit.AI analyzes performance metrics to automatically improve prompts:

Node.js

import { fetchOptimizedPrompt, traceAgentNode } from '@handit.ai/node';

async function optimizedAgent(input) {
  // Get the optimized prompt
  const optimizedPrompt = await fetchOptimizedPrompt({
    modelId: agentsTrackingConfig.mariaManager.documentCompressor
  });

  // Use optimized prompt in your node
  const optimizedNode = traceAgentNode({
    agentNodeId: agentsTrackingConfig.mariaManager.documentCompressor,
    callback: async (data) => {
      return await llmCall({
        ...data,
        prompt: optimizedPrompt
      });
    }
  });

  return await optimizedNode(input);
}

Python

class OptimizedAgent:
    @start_agent_tracing
    async def process(self, input_data):
        # Get the optimized prompt
        optimized_prompt = await self.tracker.fetch_optimized_prompt(
            model_id=agent_config["mar_i_a_manager"]["document_compressor"]
        )

        # Use optimized prompt in your node
        optimized_node = trace_agent_node({
            "agent_node_id": agent_config["mar_i_a_manager"]["document_compressor"],
            "callback": lambda data: self.llm_call(data, optimized_prompt)
        })

        return await optimized_node(input_data)

A/B Testing

Handit.AI automatically tests different configurations to find the best performing version:

• Prompt variations

• Model parameters

• Processing steps

Continuous Learning

Your agent improves automatically through:

• Performance analysis

• Error pattern detection

• Dynamic configuration updates