Ashfaqbs opened a new pull request, #1113: URL: https://github.com/apache/flink-agents/pull/1113
Linked issue: #1106 (Watsonx portion; the Anthropic portion is covered separately by #1107) ### Purpose of change Watsonx's chat response already carries a `finish_reason` in the shared OpenAI-compatible vocabulary (`stop`, `tool_calls`, etc.), but the connection only used it to emit a log warning and never surfaced it to callers. This preserves it verbatim in `extra_args["finish_reason"]`, mirroring the existing OpenAI/Azure OpenAI connection behavior, for both the Java and Python connections. - A present, non-null reason is copied verbatim into `extra_args["finish_reason"]`, including values outside `stop`/`tool_calls`. - A missing or JSON-null reason adds no key (matches the OpenAI connection's documented contract). - The existing "response may be truncated" log warning is unchanged. - Capture is independent of the token-usage metadata branch (verified with a case that has a finish reason but no usage block). ### Tests - Java: `WatsonxChatModelConnectionTest` — added 4 tests covering a documented reason, an unrecognized/vendor-specific reason stored verbatim, absent `finish_reason` member, JSON-null `finish_reason`, and a reason captured without usage metadata present. `mvn test -Dtest=WatsonxChatModelConnectionTest` → 24 passed. - Python: `test_watsonx_chat_model.py` — added 2 tests (unknown reason verbatim, no key when `None`) and asserted `finish_reason` in the existing mocked-chat test. `pytest -m "not integration"` → 14 passed. - `mvn spotless:check` and `ruff check`/`ruff format --diff` both clean on the touched files. ### API No public API signature changes; adds a new, optional key to the existing `extra_args` metadata map returned by both connections. ### Documentation - [ ] `doc-needed` - [x] `doc-not-needed` - [ ] `doc-included` ### Was this patch authored or co-authored using generative AI tooling? - [x] Yes Generated-by: Claude Code 2.1.226 (Claude Sonnet 5) -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
