Description

SummarizeExtractor would crash the entire indexing workflow for certain rejected LLM calls, like Azure content filtering issues (see #870). This is a major issue when indexing large datasets, if your input includes some explicit or violent content etc.

This PR implements the same try/catch approach used by the GraphExtractor to return an empty summarization, so indexing can continue.

Related Issues

#870

Proposed Changes

Implement a try/catch on SummarizeExtractor, so indexing doesn't stop in case of a rejected LLM call.

Checklist

• I have tested these changes locally.
• I have reviewed the code changes.
• I have updated the documentation (if necessary).
• I have added appropriate unit tests (if applicable).

Additional Notes

none