Reflection Agent Example

A Reflection agent that critiques and improves its own outputs through self-reflection.

Overview

The Reflection pattern:

1. Generate - Create initial response
2. Reflect - Critique the response
3. Improve - Generate improved version
4. Repeat - Continue until quality threshold met

Complete Example

import { createReflectionAgent } from '@agentforge/patterns';
import { ChatOpenAI } from '@langchain/openai';
import { webScraper, calculator } from '@agentforge/tools';

const agent = createReflectionAgent({
  model: new ChatOpenAI({ model: 'gpt-4' }),

  tools: [webScraper, calculator],

  maxReflections: 3,

  reflectionPrompt: `Critique the previous response:
- Is it accurate and complete?
- Is it well-structured and clear?
- Are there any errors or omissions?
- How can it be improved?`,

  improvementThreshold: 0.8 // Quality score 0-1
});

// Use the agent
const result = await agent.invoke({
  messages: [{
    role: 'user',
    content: 'Write a comprehensive blog post about the future of AI'
  }]
});

console.log('Final Output:', result.messages[result.messages.length - 1].content);

Output Example

📝 Initial Draft:
AI is transforming industries...

🤔 Reflection 1:
The draft is too generic. It needs:
- Specific examples
- Recent developments
- Expert opinions
- Future predictions

✍️ Improved Version 1:
AI is transforming industries with recent breakthroughs like GPT-4...

🤔 Reflection 2:
Better, but still needs:
- More concrete data
- Balanced perspective
- Actionable insights

✍️ Final Version:
[Comprehensive, well-researched blog post with examples, data, and insights]

✅ Quality threshold met (0.85)

When to Use Reflection

Best for:

• Content creation
• Code generation
• Report writing
• Quality-critical tasks
• Creative work

Key Features

• ✅ Self-Critique - Identify weaknesses
• ✅ Iterative Improvement - Multiple refinement cycles
• ✅ Quality Control - Meet quality thresholds
• ✅ Learning - Improve over iterations

Next Steps

• Multi-Agent System - Combine with other patterns