feat(agenthub): add agenthub llm

[ionmincu]

Description

Add agenthub llm endpoint

Development Package

• Add this package as a dependency in your pyproject.toml:

[project]
dependencies = [
  # Exact version:
  "uipath==2.0.75.dev1004340540",

  # Any version from PR
  "uipath>=2.0.75.dev1004340000,<2.0.75.dev1004350000"
]

[[tool.uv.index]]
name = "testpypi"
url = "https://test.pypi.org/simple/"
publish-url = "https://test.pypi.org/legacy/"
explicit = true

[tool.uv.sources]
uipath = { index = "testpypi" }

[Copilot]

Pull Request Overview

This PR introduces an AgentHub-capable LLM endpoint by adding an EndpointManager that routes calls based on AgentHub availability, refactors existing LLM service methods to use the new manager, and removes legacy usage-only methods and their tests.

• Add EndpointManager to detect and cache AgentHub availability and choose the correct endpoint paths.
• Refactor UiPathOpenAIService and related methods to use EndpointManager for embeddings and chat completions.
• Remove old *_usage methods and corresponding integration tests.

Reviewed Changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
tests/sdk/services/test_llm_integration.py	Removed legacy embeddings_usage and chat_completions_usage tests.
src/uipath/_utils/_endpoints_manager.py	Added EndpointManager with AgentHub detection and endpoint getters.
src/uipath/_utils/init.py	Exported EndpointManager in module exports.
src/uipath/_services/llm_gateway_service.py	Refactored embeddings and chat completion methods to use EndpointManager; removed old usage methods.

Comments suppressed due to low confidence (2)

tests/sdk/services/test_llm_integration.py:77

• Removing the test_embeddings_usage_real method eliminates coverage for the embeddings usage endpoint. Consider adding or updating tests for the new AgentHub-based embeddings usage workflow.

    @pytest.mark.asyncio

tests/sdk/services/test_llm_integration.py:104

• The test_chat_completions_usage_real method was removed, leaving no validation for chat completions usage. Please add or revise tests to cover the updated usage metrics endpoint.

    @pytest.mark.asyncio

[Copilot]

The chat_completions_usage method is returning ChatCompletion instead of usage information. It should return a usage-specific model (e.g., UsageInfo.model_validate) or rename the method accordingly.

Suggested change

	return ChatCompletion.model_validate(response.json())
	return UsageInfo.model_validate(response.json())