Foreign async RPC handler registration after connect panics (no reactor running)

[pham-tuan-binh]

Summary

Registering a foreign (Python) async RPC handler after connect() raises:

InternalError: there is no reactor running, must be called from the context of a Tokio 1.x runtime

The failure is at register_rpc_method itself, before any perform_rpc. The SDK's register_rpc_method runs on the asyncio thread, which has no tokio reactor.

Pre-existing

Found while building PR #58. Reproduces identically through the Python facade on unmodified main, so it is not related to the role-split bindings.

Reproduce

async def handler(data): return f"pong:{data.payload}"
await robot.connect(url, token)
robot.register_rpc_method("ping", handler)  # raises here

Workaround (supported path)

Register handlers before connect(). The core stores pre-connect handlers and applies them inside connect(), which runs on the tokio runtime. PR #58's test_rpc_roundtrip uses this path and passes.

Possible fix directions

• Marshal the SDK register_rpc_method call onto the tokio runtime instead of invoking it on the caller's asyncio thread.
• Or document pre-connect registration as the only supported path and reject post-connect registration with a clear error rather than a panic.