Pydantic AI

Pydantic AI reaches this API through OpenAIChatModel with an OpenAICompatibleProvider: a name for the service, the base URL https://test-demo.pinerouter.ai/v1 and your key.

Install

pip install "pydantic-ai-slim[openai]"

uv add "pydantic-ai-slim[openai]"

Configure API keys

.env

API_KEY=sk-...

Load it into the environment before you run the code (for example with python-dotenv). Keep the key out of your code.

Run

Text

import os
from pydantic_ai import Agent
from pydantic_ai.models.openai import OpenAIChatModel
from pydantic_ai.providers.openai_compatible import OpenAICompatibleProvider

model = OpenAIChatModel(
    "gpt-6-luna",
    provider=OpenAICompatibleProvider("ai", base_url="https://test-demo.pinerouter.ai/v1", api_key=os.environ["API_KEY"]),
)
agent = Agent(model, instructions="You are concise.")

print(agent.run_sync("Name three rivers in Europe.").output)

Stream

import asyncio

async def main():
    async with agent.run_stream("Write a haiku about rivers.") as result:
        async for text in result.stream_text(delta=True):
            print(text, end="", flush=True)

asyncio.run(main())

Tool calling

@agent.tool_plain
def get_weather(city: str) -> str:
    """Current weather for a city."""
    return f"Sunny, 21 C in {city}"

print(agent.run_sync("What is the weather in Paris?").output)

Specifics

Use OpenAICompatibleProvider, not OpenAIProvider(base_url=...). OpenAIProvider assumes OpenAI's own models and features.
Keep the provider name ("ai" above) the same over time. Pydantic AI uses it to decide which data from earlier answers it sends back.
If structured output or thinking is ignored or refused by a model, set a model profile as the Pydantic AI docs describe.
OpenAIResponsesModel with the same provider uses /v1/responses instead of Chat Completions.
For images, video and other media call Jobs with the HTTP helper inside a tool.

Pydantic AI compatible APIs
Pydantic AI tools
Text: tools, JSON output and vision per model