Confirm LiteLLM provider registrations are cached per router

[Copilot]

LiteLLM provider setup must not be repeated for every request. The router now already satisfies this through its registration lifecycle, so no code changes are proposed.

• Cached registration

  • AgentRouter constructs one _ChatRegistration during initialization.
  • The registration retains resolved model, endpoint, API key, and provider defaults.

• Request dispatch

  • Requests retrieve the existing registration from _agent_registry.
  • Dispatch invokes LiteLLM directly without reconstructing provider configuration.

agent_instance = self.get_agent_instance(registration_key)
response = self._dispatch_via_litellm(
    registration_key=registration_key,
    agent_instance=agent_instance,
    provider_config=provider_config,
    request_data=request_data,
)

• Fixes Cache the LiteLLM provider in the router instead of resolving per-request #397

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[Copilot]

Copilot wasn't able to review any files in this pull request.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Marco Russo (marcorusso97)]

Ok

[Marco Russo (marcorusso97)]

Ok