Handle None content in stop-sequence trimming

[chahn]

Prevent stop-sequence trimming from crashing on tool-only GPT-5 outputs

Linked Issue

Fixes Bug #1825

Summary

• Prevent remove_content_after_stop_sequences from crashing when a provider returns content=None (e.g. gpt-5 tool-only turns).
• Ignore empty stop strings and tolerate a None stop list to avoid accidental truncation.

Details

• Treat content=None and stop_sequences=None as no-op cases so we only trim when actual text is present.
• Skip zero-length stop strings to keep the full generation intact when upstream configuration leaves empty entries.
• This is enough to unblock the agent and any other flows relying on gpt-5 tool-calling without changing behavior for normal string outputs.

Tests

• tests/test_models.py::test_remove_content_after_stop_sequences_handles_none – confirms the helper returns None when the provider emits tool-only output, mirroring the bug scenario.
• tests/test_models.py::test_remove_content_after_stop_sequences_ignores_empty_stop_sequences – ensures harmless empty stop strings do not strip valid text.
• tests/test_models.py::test_remove_content_after_stop_sequences_with_none_stop_sequences – protects the default code path where stop_sequences is None, preserving existing behavior.

Thanks a lot for feedback and support!

[aymeric-roucher]

Why support having one of the stop sequences being None?
Apart from that I agree with the other introduced changes!

[chahn]

Thanks for spotting that @aymeric-roucher! You’re right, we don’t expect providers to return None inside stop_sequences.