feat: AE-1018 Add support for new serverless runtime

[pandyamarut]

No description provided.

[Copilot]

Pull Request Overview

This PR adds support for a new serverless runtime called "Load Balancer SLS" that provides dual-capability execution - both traditional remote execution and HTTP endpoint exposure. The implementation introduces a FastAPI-based server that can handle both programmatic calls via RemoteExecutor and HTTP requests to decorated class methods.

• Implements Load Balancer SLS server with dual execution capabilities (remote + HTTP endpoints)
• Adds @endpoint decorator for marking class methods as HTTP endpoints
• Creates class registry system for managing deployed classes and their endpoint mappings

Reviewed Changes

Copilot reviewed 7 out of 8 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
tetra-rp	Updates subproject commit reference
src/load_balancer_sls_handler.py	Main server implementation with FastAPI integration and dual execution support
src/endpoint.py	HTTP endpoint decorator for marking class methods as web endpoints
src/class_registry.py	Class management system for registration, instantiation, and method execution
pyproject.toml	Adds FastAPI, uvicorn, and aiohttp dependencies
Makefile	Adds build and test targets for Load Balancer SLS image
Dockerfile-load-balancer-sls	Docker configuration for the new serverless runtime

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

[Copilot]

This line will raise an AttributeError if input_data.function_name doesn't exist when class_name is also missing. The getattr call should have a default value for the fallback case.

Suggested change

	f"Remote execution: {getattr(input_data, 'class_name', input_data.function_name)}"
	f"Remote execution: {getattr(input_data, 'class_name', None) or getattr(input_data, 'function_name', None) or '<unknown>'}"

[Copilot]

The logic in lines 298-301 is incorrect. It's comparing metadata.get('class_name') with instance_key, but instance_key is an instance identifier, not a class name. This will never match and the metadata updates will be duplicated or missed.

Suggested change

	# Also update by class name for default instances
	for key, metadata in self.instance_metadata.items():
	if metadata.get("class_name") == instance_key:
	metadata["method_calls"] += 1
	metadata["last_used"] = datetime.now().isoformat()

[Copilot]

Using a mutable list as a default argument is an anti-pattern. The list should be None with a fallback inside the function: methods: Optional[List[str]] = None and then methods = methods or ['POST'] in the function body.

[Copilot]

This line assumes request_data is always a valid dictionary for keyword expansion. If request_data contains keys that don't match the method's parameters, this will raise a TypeError. Consider adding parameter validation or using inspect.signature() to filter valid parameters.

Suggested change

	result = method(**request_data)
	sig = inspect.signature(method)
	params = sig.parameters
	accepts_kwargs = any(
	p.kind == inspect.Parameter.VAR_KEYWORD
	for p in params.values()
	)
	if accepts_kwargs:
	filtered_data = request_data
	else:
	filtered_data = {
	k: v
	for k, v in request_data.items()
	if k in params
	}
	result = method(**filtered_data)

[deanq]

Superceded by #45