Fix TypeError in MLE-STAR get_text_from_response when handling None text values

[0xnavarro]

Problem

The MLE-STAR agent crashes with a TypeError: can only concatenate str (not "NoneType") to str when processing LLM responses that contain parts with None text values.

Solution

Added a null check in get_text_from_response() to verify that response.content.parts[i].text is not None before concatenating it to the final text string.

Changes

• Modified machine_learning_engineering/shared_libraries/common_util.py
• Added condition: and response.content.parts[i].text is not None

Impact

• Prevents agent crashes during task execution
• Improves robustness when handling malformed or incomplete LLM responses
• Maintains backward compatibility

Fixes a critical bug that could cause the MLE-STAR system to crash unexpectedly.

[0xnavarro]

This change is necessary because, without the explicit check for response.content.parts[i].text is not None, the code may attempt to concatenate a None value, which causes a crash. In real usage, I have encountered cases where response.content.parts[i].text exists as an attribute but is actually None, leading to runtime errors. With this improvement, the code works reliably and avoids unexpected failures. Without it, the current implementation is not safe for production use.

[ivanmkc]

@raj-sinha can you review?

[0xnavarro]

@raj-sinha Could you please review it? Thank you.

[0xnavarro]

If there’s no maintenance for this repo, please consider archiving it or explicitly stating that it’s not suitable for production use.