contaigents/dist/tests/testToolFirstBehavior.js

Modular AI Content Ecosystem with Audio Generation

#!/usr/bin/env node
import { ChatService } from '../services/chatService.js';
async function testToolFirstBehavior() {
    console.log('🔧 Testing Tool-First Behavior with Enhanced Prompts...\n');
    try {
        // Initialize chat service
        const chatService = new ChatService(process.cwd());
        // Test the exact scenario the user mentioned
        console.log('📋 Testing: "learn project setup in detail and update the readme"');
        const session = await chatService.createSession();
        console.log('Sending message...\n');
        const response = await chatService.sendMessage(session.id, "learn project setup in detail and update the readme");
        console.log('🤖 Response Analysis:');
        console.log('Response length:', response.content.length);
        // Check if the response contains tool calls (XML format)
        const hasToolCalls = response.content.includes('<tool_call');
        console.log('Contains tool calls:', hasToolCalls ? '✅ YES' : '❌ NO');
        // Check if response starts with planning text vs tool calls
        const startsWithPlanning = response.content.trim().startsWith('To help you') ||
            response.content.trim().startsWith('I\'ll help you') ||
            response.content.trim().startsWith('Let me plan');
        console.log('Starts with planning:', startsWithPlanning ? '❌ YES (BAD)' : '✅ NO (GOOD)');
        // Show first 300 characters to analyze behavior
        console.log('\n📝 First 300 characters of response:');
        console.log('---');
        console.log(response.content.substring(0, 300));
        console.log('---\n');
        // Test another scenario
        console.log('📋 Testing: "analyze my project structure"');
        const response2 = await chatService.sendMessage(session.id, "analyze my project structure");
        const hasToolCalls2 = response2.content.includes('<tool_call');
        console.log('Contains tool calls:', hasToolCalls2 ? '✅ YES' : '❌ NO');
        console.log('\n📝 First 200 characters of response:');
        console.log('---');
        console.log(response2.content.substring(0, 200));
        console.log('---\n');
        console.log('✅ Tool-first behavior test completed!');
        if (hasToolCalls && hasToolCalls2) {
            console.log('🎉 SUCCESS: Agent is using tools immediately!');
        }
        else {
            console.log('⚠️  NEEDS IMPROVEMENT: Agent should use tools first for information gathering');
        }
    }
    catch (error) {
        console.error('❌ Error testing tool-first behavior:', error);
        process.exit(1);
    }
}
// Run the test
testToolFirstBehavior();