> ## Documentation Index
> Fetch the complete documentation index at: https://polos.dev/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Thinking Agent

> Chain-of-thought reasoning with structured output

Build agents that show their reasoning process step by step.

## Define a reasoning schema

<CodeGroup>
  ```python Python theme={null}
  from pydantic import BaseModel, Field

  class ReasoningOutput(BaseModel):
      problem: str = Field(description="The original problem")
      thinking_steps: list[str] = Field(description="Step-by-step reasoning")
      conclusion: str = Field(description="Final answer")
      confidence: str = Field(description="high, medium, or low")
  ```

  ```typescript TypeScript theme={null}
  import { z } from 'zod';

  const reasoningOutputSchema = z.object({
    problem: z.string().describe('The original problem statement'),
    thinking_steps: z.array(z.string()).describe('Step-by-step reasoning process'),
    conclusion: z.string().describe('The final answer or conclusion'),
    confidence: z.string().describe('Confidence level: high, medium, or low'),
  });

  type ReasoningOutput = z.infer<typeof reasoningOutputSchema>;
  ```
</CodeGroup>

## Create a thinking agent

<CodeGroup>
  ```python Python theme={null}
  from polos import Agent, max_steps, MaxStepsConfig

  thinking_agent = Agent(
      id="thinking_agent",
      provider="openai",
      model="gpt-4o-mini",
      system_prompt="""You are a logical reasoning expert. When given a problem:
  1. Restate the problem
  2. Break down your thinking into clear steps
  3. Consider potential pitfalls
  4. Arrive at a conclusion
  5. State your confidence level""",
      output_schema=ReasoningOutput,
      stop_conditions=[max_steps(MaxStepsConfig(limit=20))],
  )
  ```

  ```typescript TypeScript theme={null}
  import { defineAgent, maxSteps } from '@polos/sdk';
  import { openai } from '@ai-sdk/openai';

  const thinkingAgent = defineAgent({
    id: 'thinking_agent',
    model: openai('gpt-4o-mini'),
    systemPrompt:
      'You are a logical reasoning expert. When given a problem:\n' +
      '1. Restate the problem\n' +
      '2. Break down your thinking into clear steps\n' +
      '3. Consider potential pitfalls\n' +
      '4. Arrive at a conclusion\n' +
      '5. State your confidence level',
    outputSchema: reasoningOutputSchema,
    stopConditions: [maxSteps({ count: 20 })],
  });
  ```
</CodeGroup>

## Get structured reasoning

<CodeGroup>
  ```python Python theme={null}
  result = await thinking_agent.run(polos, "If all roses are flowers and some flowers fade quickly, can we conclude that some roses fade quickly?")

  print(f"Problem: {result.result.problem}")
  for i, step in enumerate(result.result.thinking_steps):
      print(f"Step {i+1}: {step}")
  print(f"Conclusion: {result.result.conclusion}")
  print(f"Confidence: {result.result.confidence}")
  ```

  ```typescript TypeScript theme={null}
  const result = await thinkingAgent.stream(polos, {
    input: 'If all roses are flowers and some flowers fade quickly, can we conclude that some roses fade quickly?',
  });

  for await (const chunk of result.textChunks) {
    process.stdout.write(chunk);
  }
  ```
</CodeGroup>

## Run it

<CodeGroup>
  ```bash Python theme={null}
  git clone https://github.com/polos-dev/polos.git
  cd polos/python-examples/05-thinking-agent
  cp .env.example .env  # Add your POLOS_PROJECT_ID and API key
  uv sync
  python main.py
  ```

  ```bash TypeScript theme={null}
  git clone https://github.com/polos-dev/polos.git
  cd polos/typescript-examples/05-thinking-agent
  cp .env.example .env  # Add your POLOS_PROJECT_ID and API key
  npm install
  npx tsx main.ts
  ```
</CodeGroup>

Open [http://localhost:5173](http://localhost:5173) to view your agents and workflows, run them from the UI, and see execution traces.

[Python example on GitHub](https://github.com/polos-dev/polos/tree/main/python-examples/05-thinking-agent) | [TypeScript example on GitHub](https://github.com/polos-dev/polos/tree/main/typescript-examples/05-thinking-agent)