fix(python): ignore empty tool messages

[pragnyanramtha]

Summary

• avoid indexing the first tool result while scanning memory for the last tool call/response pair
• use the last tool call from a multi-tool assistant message when selecting the matching tool response
• ignore empty or unrelated tool messages instead of raising IndexError
• add regression coverage for matching tool responses, empty tool messages, and multi-tool assistant messages

Validation

• uv run --with pytest --with syrupy pytest tests/memory/test_utils.py -q -> 3 passed; emitted local warnings for an unknown asyncio config option and PyCharm diff patching
• uv run --with ruff ruff check beeai_framework/memory/utils.py tests/memory/test_utils.py
• uv run --with ruff ruff format --check beeai_framework/memory/utils.py tests/memory/test_utils.py
• git diff --check

Notes

The commits include DCO signoffs as required by the Python contributing guide.

[gemini-code-assist]

Code Review

This pull request updates the extract_last_tool_call_pair utility to more robustly identify matching tool responses by checking all results within a ToolMessage using any(). It also introduces unit tests to verify matching logic and the handling of empty tool messages. Review feedback suggests using the last tool call ID from the AssistantMessage to ensure the most recent interaction is captured and recommends optimizing the search loop by iterating over msg.content directly to reduce overhead.

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.