Summary

Implements PydanticAI agent runner as requested in #272, following the existing LangChain integration pattern.

What is PydanticAI?

PydanticAI is a type-safe, model-agnostic agent framework that provides:

• Strong typing with Pydantic models
• Dependency injection for tools
• Streaming support
• Built-in tool/function calling
• Multiple model provider support

Changes

• New PydanticAIAgentRunner class in src/openagents/agents/pydantic_ai_agent.py
• Tool conversion utilities:
  • openagents_tool_to_pydantic() - convert OpenAgents tools to PydanticAI format
  • pydantic_tool_to_openagents() - convert PydanticAI tools to OpenAgents format
• Full feature parity with LangChain integration:
  • Event filtering (event_names, event_filter)
  • Custom response handlers
  • Network tool injection
• PydanticAI-specific: dependency injection support via deps parameter

Usage

from pydantic_ai import Agent
from openagents.agents import PydanticAIAgentRunner

# Create a type-safe PydanticAI agent
agent = Agent(
    'openai:gpt-4',
    system_prompt='You are a helpful assistant.'
)

# Connect to OpenAgents network
runner = PydanticAIAgentRunner(
    pydantic_agent=agent,
    agent_id='my-pydantic-agent'
)
runner.start(network_host='localhost', network_port=8600)
runner.wait_for_stop()

With Dependencies

from dataclasses import dataclass
from pydantic_ai import Agent, RunContext

@dataclass
class MyDeps:
    api_key: str
    user_id: str

agent = Agent('openai:gpt-4', deps_type=MyDeps)

@agent.tool
async def get_user_data(ctx: RunContext[MyDeps]) -> str:
    return f'Data for user {ctx.deps.user_id}'

runner = PydanticAIAgentRunner(
    pydantic_agent=agent,
    agent_id='my-agent',
    deps=MyDeps(api_key='...', user_id='123')
)

Reference

Follows the pattern established in src/openagents/agents/langchain_agent.py

Closes #272