feat: add PydanticAIAgentRunner for type-safe network integration

[Raghavan-04]

Description

This PR introduces the PydanticAIAgentRunner, enabling PydanticAI agents to seamlessly connect to the OpenAgents network.

Following the architectural pattern of the existing LangChainAgentRunner, this implementation provides a robust wrapper that allows PydanticAI's type-safe agents to function as first-class citizens within the network. By leveraging PydanticAI, we ensure that agent interactions remain strictly typed and validated, reducing runtime errors across the network.

Capabilities

• Event Processing: Allows agents to receive and process asynchronous network events.
• Network Tool Integration: Enables PydanticAI agents to utilize tools provided by the OpenAgents network dynamically.
• Type-Safe Responses: Guarantees that structured data returned to the network conforms to predefined Pydantic models.

Key Changes

• New Agent Runner: Created src/openagents/agents/pydantic_ai_agent.py to house the core runner logic.
• Message Mapping: Implemented high-fidelity mapping from OpenAgents internal message formats to PydanticAI ModelMessage structures.
• Dynamic Tooling: Added support for registering network-level tools within the PydanticAI run loop, ensuring agents can interact with the environment effectively.

Testing Performed

• Verified message mapping consistency.
• Tested basic tool calling via the OpenAgents network protocol.
• Validated output schema adherence using Pydantic models.

---

[vercel]

Someone is attempting to deploy a commit to the Raphael's projects Team on Vercel.

A member of the Team first needs to authorize it.