[Hardik K (hardikkaurani)]

Motivation and Context

Closes #14328

When optional connector dependencies (such as transformers, torch, onnxruntime-genai, or chromadb) are not installed in the environment, callers could encounter abrupt module-level import errors or generic ModuleNotFoundError/ImportError exceptions that did not clarify which Semantic Kernel extra needed to be installed.

Specifically:

1. hf_prompt_execution_settings.py attempted to load transformers unconditionally at the module level without try...except, crashing upon import when optional Hugging Face extras were omitted. Additionally, get_generation_config() imported GenerationConfig before verifying the optional dependency status.
2. hf_text_completion.py and hf_text_embedding.py eagerly imported torch and transformers at the top level, preventing callers from importing or handling Hugging Face settings or services gracefully without pre-installed packages.
3. OnnxGenAICompletionBase and ChromaMemoryStore raised generic error messages that did not identify the specific Semantic Kernel install extra (semantic-kernel[onnx] and semantic-kernel[chroma]).

Description

• Hugging Face Connector:
  • Guarded module-level transformers and torch imports in hf_prompt_execution_settings.py, hf_text_completion.py, and hf_text_embedding.py with try...except (ImportError, ModuleNotFoundError).
  • Added clear, actionable error messages directing users to install semantic-kernel[hugging_face] when attempting to instantiate uninstalled services or invoke get_generation_config().
• ONNX Connector:
  • Updated OnnxGenAICompletionBase lazy-import error message to specify pip install semantic-kernel[onnx].
  • Cleaned up generator comprehension in _generate_next_token.
• Chroma Connector:
  • Updated ChromaMemoryStore to explicitly reference pip install semantic-kernel[chroma].
• Unit Tests:
  • Added tests/unit/connectors/test_optional_dependencies.py covering:
    • Instantiation and config generation for HuggingFacePromptExecutionSettings both with and without transformers.
    • Instantiation of HuggingFaceTextCompletion and HuggingFaceTextEmbedding without required extras.
    • Instantiation of OnnxGenAICompletionBase without onnxruntime-genai.
    • Instantiation of ChromaMemoryStore without chromadb.
  • Tests remain fully isolated from local extra installations via mock patching.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines and the pre-submission formatting script raises no violations
• All unit tests pass, and I have added new tests where possible
• I didn't break anyone 😄

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[Hardik K (hardikkaurani)]

Hi Eduard van Valkenburg (@eavanvalkenburg),

Could you please take a look at this pull request when you have a moment? It addresses #14328 by adding actionable install-extra error messages for optional connector dependencies (semantic-kernel[hugging_face], semantic-kernel[onnx], and semantic-kernel[chroma]), fixing unguarded module-level imports in the Hugging Face connector, and adding isolated regression tests for missing optional dependencies.

All Ruff checks and unit tests are passing cleanly. Thank you!