Course contentsShow
← AI Engineering
Lesson 965 of 1,88624. Building AI APIsFree lesson

Pydantic Models for LLM Requests

Define request schemas with Pydantic to validate prompt parameters, temperature, max_tokens, and other LLM settings.

Pydantic Models for LLM Requests

What you'll learn: How to use Pydantic models to validate and structure incoming LLM API requests with type safety and automatic error handling.

Why Pydantic for LLM Requests?

When building AI APIs (as introduced in your previous FastAPI lessons), you need to validate that incoming requests contain proper data before sending them to expensive LLM providers. Pydantic models act as strict contracts — think of them as blueprints that ensure every request has the right shape, valid values, and required fields.

Defining a Request Schema

Here's a typical LLM request model:

from pydantic import BaseModel, Field
from typing import Optional

class LLMRequest(BaseModel):
    prompt: str = Field(..., min_length=1, max_length=5000)
    temperature: float = Field(default=0.7, ge=0.0, le=2.0)
    max_tokens: int = Field(default=256, gt=0, le=4096)
    model: str = Field(default="gpt-3.5-turbo")
    system_message: Optional[str] = None

What's happening:

  • prompt is required and must be 1-5000 characters
  • temperature defaults to 0.7 but must be between 0.0-2.0
  • max_tokens has range validation (1-4096)
  • model has a sensible default
  • system_message is optional

FastAPI automatically validates requests against this model. Invalid data? Users get clear error messages before you waste an API call.

The Power of Validation

Without Pydantic, you'd write dozens of if statements checking bounds and types. Pydantic handles this declaratively — you describe what's valid, and it enforces it automatically. This prevents common errors like negative temperatures or empty prompts from reaching your LLM provider.

Key Takeaway: Pydantic models provide type-safe, declarative request validation for LLM APIs, catching invalid parameters before expensive API calls and giving users clear feedback automatically.