feat: WatsonXChatGenerator - support Tools and use the StreamingChunk fields

[maxdswain]

Related Issues

• fixes feat: Update WatsonXChatGenerator to use the StreamingChunk fields #2217 & feat: Update our ChatGenerators in integrations to use the new StreamingChunk fields #2070

Proposed Changes:

feat: Update WatsonXChatGenerator to use the StreamingChunk index and finish_reason fields.

How did you test it?

Updated existing tests.

Notes for the reviewer

Checklist

• I have read the contributors guidelines and the code of conduct
• I have updated the related issue with new insights and changes
• I added unit tests and updated the docstrings
• I've used one of the conventional commit types for my PR title: fix:, feat:, build:, chore:, ci:, docs:, style:, refactor:, perf:, test:.

[anakin87]

Thanks for the contribution!

I think that a little more adaptation work will be needed, especially in handling tool calls.

I suggest taking inspiration from #2078 and #2108 (but keeping in mind that each provider has its own particularities and code can be very different).

Also, I'd like to see a test with real chunks coming from WatsonX (similar test for Google GenAI).

[maxdswain]

Thanks for the contribution!

I think that a little more adaptation work will be needed, especially in handling tool calls.

I suggest taking inspiration from #2078 and #2108 (but keeping in mind that each provider has its own particularities and code can be very different).

Also, I'd like to see a test with real chunks coming from WatsonX (similar test for Google GenAI).

Okay, I'll convert it to a draft for now and work on it a bit more to add support for tool calls as well as more tests. I'll let you know when it's ready.

[maxdswain]

Hi @anakin87, this PR should now be ready to be reviewed. I have added a bunch of tests for tool calling (mirroring the Google Gen AI integration) as well as a test that converts real chunks I got from making requests to Watsonx.ai.

[anakin87]

Thank you!

Left two initial comments

[anakin87]

Tests are not running on PRs from forks but when I run hatch run test:integration locally, I get 4 tests failing:

FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_toolset - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_tools_streaming - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_mixed_tools - AssertionError: Expected 2 tool calls, got 1
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_async_with_tools - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}

Could you please investigate?

[maxdswain]

Tests are not running on PRs from forks but when I run hatch run test:integration locally, I get 4 tests failing:

FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_toolset - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_tools_streaming - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_with_mixed_tools - AssertionError: Expected 2 tool calls, got 1
FAILED tests/test_chat_generator.py::TestWatsonxChatGeneratorIntegration::test_live_run_async_with_tools - assert '{\n  "city": "Paris"\n}' == {'city': 'Paris'}

Could you please investigate?

Thanks for spotting this, I was running my local tests using an IBM EU cloud account so had changed the endpoint as well as the model as less models are supported on the EU cloud when running the tests. I have now updated the tool call parsing to better handle output from the default model.

[anakin87]

Looks good!