Skip to main content

Pydantic2

Pydantic2

A powerful AI framework with structured Pydantic response handling, LLM integration, and advanced agent capabilities.

Documentation PyPI version License: MIT

Installation​

pip install pydantic2

Set up your API key:

export OPENROUTER_API_KEY=your_api_key_here

Quick Start​

from pydantic import BaseModel, Field
from typing import List
from pydantic2 import LiteLLMClient, Request

class MovieReview(BaseModel):
    title: str = Field(description="The title of the movie")
    rating: float = Field(description="The rating of the movie")
    pros: List[str] = Field(description="The pros of the movie")
    cons: List[str] = Field(description="The cons of the movie")

client = LiteLLMClient(Request(
    model="openrouter/openai/gpt-4o-mini-2024-07-18",
    answer_model=MovieReview
))

response = client.generate_response(
    prompt="Review the movie 'Inception'"
)

print(f"Title: {response.title}")
print(f"Rating: {response.rating}/5")
print("Pros:", ", ".join(response.pros))
print("Cons:", ", ".join(response.cons))

Key Features​

Structured Responses​

  • Type-safe responses using Pydantic models
  • Automatic validation and parsing
  • IDE support with autocompletion
  • Custom response models with field descriptions
  • Nested model support

LLM Integration​

  • Support for multiple LLM providers
  • Unified API for all models
  • Easy provider switching
  • Automatic retries and fallbacks
  • Streaming support

Budget Control​

  • Built-in cost tracking
  • Budget limits per request/user
  • Usage statistics and analytics
  • Cost estimation before requests
  • Detailed usage reports

Message Handling​

  • System and user messages
  • Conversation history
  • Structured data support
  • Support for code blocks
  • Support for JSON and DataFrame inputs

Agent System​

  • Custom tools and functions
  • Gradio UI integration
  • Extensible framework
  • Tool decorators
  • Memory management

Configuration​

from pydantic2 import Request

config = Request(
    # Model settings
    model="openrouter/openai/gpt-4o-mini-2024-07-18",
    answer_model=YourModel,
    temperature=0.7,
    max_tokens=500,

    # Performance features
    online=True,
    cache_prompt=True,
    max_budget=0.05,

    # User tracking
    user_id="user123",
    client_id="my_app"
)

Framework Integration​

Django Example​

from rest_framework.views import APIView
from rest_framework.response import Response
from pydantic import BaseModel, Field
from typing import List
from pydantic2 import Request, LiteLLMClient

class FeedbackAnalysis(BaseModel):
    summary: str = Field(..., description="Summary of the feedback")
    sentiment: str = Field(..., description="Detected sentiment")
    key_points: List[str] = Field(..., description="Key points from the feedback")

class FeedbackView(APIView):
    def post(self, request):
        feedback = request.data.get('feedback', '')

        client = LiteLLMClient(Request(
            model="openrouter/openai/gpt-4o-mini-2024-07-18",
            answer_model=FeedbackAnalysis,
            max_budget=0.01,
            user_id=request.user.id
        ))

        response = client.generate_response(prompt=feedback)
        return Response(response.model_dump())

Why Pydantic2?​

  • Type Safety: Get structured responses with proper type hints and validation
  • Efficiency: Reduce boilerplate code and focus on your application logic
  • Reliability: Production-tested with comprehensive error handling
  • Flexibility: Support for multiple LLM providers and frameworks
  • Scalability: Built for both small projects and enterprise applications
  • Cost Control: Built-in budget management and usage tracking
  • Framework Support: Seamless integration with Django, FastAPI, and more
  • Developer Experience: Great IDE support and documentation

Documentation & Support​