fix: Include original error message in ImportError for LLM client dependencies

[karthik-idikuda]

Summary

Improved import error handling across autogen-ext to include the original error message in re-raised ImportError exceptions. This makes it significantly easier for users to diagnose dependency issues.

Problem

When an import error occurs while initializing LLM clients or other extension components, the code raises a generic ImportError with a hardcoded message suggesting to install specific packages. However, the actual import error might be caused by other issues (like missing transitive dependencies or version conflicts).

Before:

ImportError: Dependencies for Llama Cpp not found. Please install llama-cpp-python: pip install autogen-ext[llama-cpp]

After:

ImportError: Dependencies for Llama Cpp not found. Original error: No module named 'llama_cpp'
Please install llama-cpp-python: pip install autogen-ext[llama-cpp]

Changes

Updated 10 files across autogen-ext to include the original exception message:

• models/llama_cpp/__init__.py
• agents/azure/__init__.py
• agents/magentic_one/__init__.py
• runtimes/grpc/__init__.py and 3 worker runtime files
• memory/chromadb/_chromadb.py
• memory/redis/_redis_memory.py
• tools/azure/_ai_search.py

Testing

The change is minimal and only affects error message formatting. The from e chain is preserved in all cases, and the original exception object is still accessible via __cause__.

Closes #4605

[karthik-idikuda]

@microsoft-github-policy-service agree