Rebuild with Pydantic AI

Port your scratch agent to a production-ready framework. Learn typed tools and structured outputs.

🎯 The Mission

You've built an agent from scratch—now you understand the fundamentals. But raw API calls don't scale. Tenzai uses frameworks that provide type safety, validation, and better developer experience.

Pydantic AI is a lightweight, type-first framework that's closer to the metal than LangChain but more structured than raw calls. Rebuild your agent using it.

What You'll Learn

Pydantic AI agents: How the framework structures tool definitions
Typed tool inputs: Using Pydantic models for tool parameters
Structured outputs: Getting typed responses from the LLM
Dependency injection: Passing runtime context to tools
Framework tradeoffs: What you gain and lose vs raw API calls

Comparing Approaches

Aspect	Raw API	Pydantic AI
Tool definitions	Manual JSON schemas	Derived from Pydantic models
Type safety	None (dict/str)	Full type checking
Validation	Manual	Automatic via Pydantic
Boilerplate	Higher	Lower
Flexibility	Maximum	Framework conventions

Implementation Guide

Step 1: Install Pydantic AI

pip install pydantic-ai

Step 2: Define Your Agent

from pydantic_ai import Agent
from pydantic import BaseModel

# Define the agent with a system prompt
agent = Agent(
    'openai:gpt-4',
    system_prompt='You are a helpful assistant that uses tools to solve problems.'
)

Step 3: Add Typed Tools

from pydantic import Field

class CalculatorInput(BaseModel):
    expression: str = Field(description="Math expression to evaluate")

@agent.tool
def calculator(input: CalculatorInput) -> str:
    """Perform arithmetic calculations."""
    # Safe evaluation (use a proper math parser in production!)
    try:
        result = eval(input.expression)
        return f"Result: {result}"
    except Exception as e:
        return f"Error: {e}"

class ReadFileInput(BaseModel):
    path: str = Field(description="Path to file to read")

@agent.tool
def read_file(input: ReadFileInput) -> str:
    """Read contents of a file."""
    try:
        with open(input.path, 'r') as f:
            return f.read()
    except FileNotFoundError:
        return f"File not found: {input.path}"

Step 4: Run the Agent

async def main():
    result = await agent.run("What is 25 * 17 + 89?")
    print(result.data)  # The final response
    print(result.all_messages())  # Full conversation history

import asyncio
asyncio.run(main())

Step 5: Add Structured Output

from pydantic import BaseModel

class MathResult(BaseModel):
    expression: str
    result: float
    explanation: str

# Create agent that returns structured output
math_agent = Agent(
    'openai:gpt-4',
    result_type=MathResult,
    system_prompt='Solve math problems and explain your work.'
)

# The agent will return a MathResult object, not a string
result = await math_agent.run("What is 25 * 17 + 89?")
print(result.data.result)  # 514.0
print(result.data.explanation)  # "First multiply 25 by 17..."