feat(xai): add streaming support for xAI provider

[jeongyoonm]

Summary

This PR resolves issue #1663. It adds streaming support for the xAI provider.

Description

• Implemented iterable and partial streaming features for sync and async xAI clients.
• Works for both XAI_JSON and XAI_TOOLS mode.
• Added test cases that do simple and union extraction for every case.

Note: XAI_TOOLS does not support streaming. The current implementation works, but effectively it is not streaming.
See https://docs.x.ai/docs/guides/function-calling .

PTAL @jxnl @ivanleomk

---

Important

Adds streaming support for xAI provider with iterable and partial streaming features, updates from_xai function, and introduces new tests.

• Behavior:
  • Adds streaming support for xAI provider in instructor/providers/xai/client.py.
  • Implements iterable and partial streaming for sync and async clients.
  • Supports XAI_JSON mode; XAI_TOOLS mode does not effectively stream.
• Functions:
  • Modifies acreate() and create() to process streaming responses in instructor/providers/xai/client.py.
• Tests:
  • Adds test_stream.py with tests for iterable and partial models, both sync and async.
  • Tests include simple and union extraction cases.
• Misc:
  • Changes default mode to XAI_JSON in from_provider() in auto_client.py.

^{This description was created by for 5652050. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to 5652050 in 1 minute and 16 seconds. Click for details.

• Reviewed 366 lines of code in 4 files
• Skipped 0 files when reviewing.
• Skipped posting 7 draft comments. View those below.
• Modify your settings and rules to customize what types of comments Ellipsis leaves. And don't forget to react with 👍 or 👎 to teach Ellipsis.

1. .ruff.toml:53

• Draft comment:
  Removed 'ignore-init-module-imports'. Confirm this removal is intentional and does not break existing projects.
• Reason this comment was not posted:
  Comment looked like it was already resolved.

2. instructor/auto_client.py:920

• Draft comment:
  Default mode updated to Mode.XAI_JSON instead of Mode.JSON. Verify this change for backward compatibility.
• Reason this comment was not posted:
  Comment did not seem useful. Confidence is useful = 0% <= threshold 85% The comment is asking the author to verify backward compatibility, which is against the rules. It doesn't provide a specific suggestion or point out a clear issue with the code change. Therefore, it should be removed.

3. instructor/providers/xai/client.py:83

• Draft comment:
  Extraction of the 'stream' flag via call_kwargs.pop('stream', False) is clear. Consider refactoring similar kwarg cleanup into a helper for DRY.
• Reason this comment was not posted:
  Confidence changes required: 50% <= threshold 85% None

4. instructor/providers/xai/client.py:95

• Draft comment:
  Streaming logic for JSON mode is duplicated in async and sync paths. Consider extracting the common logic to a helper function to reduce code repetition.
• Reason this comment was not posted:
  Confidence changes required: 80% <= threshold 85% None

5. instructor/providers/xai/client.py:125

• Draft comment:
  In the async tool mode branch, the async generator comprehension for 'args' may be empty if no tool_calls are returned. Consider handling an empty stream explicitly.
• Reason this comment was not posted:
  Confidence changes required: 80% <= threshold 85% None

6. instructor/providers/xai/client.py:202

• Draft comment:
  In the synchronous tool mode streaming branch, if no tool_calls are found during iteration the function may return None. Consider adding explicit handling (e.g. raise an error) for such cases.
• Reason this comment was not posted:
  Confidence changes required: 80% <= threshold 85% None

7. tests/llm/test_xai/test_stream.py:15

• Draft comment:
  Test cases cover sync and async streaming for iterable and partial models. Consider adding tests for failure cases or empty streaming responses.
• Reason this comment was not posted:
  Confidence changes required: 50% <= threshold 85% None

Workflow ID: wflow_fDjUNsMVzweKS8N9

^{You can customize by changing your verbosity settings, reacting with 👍 or 👎, replying to comments, or adding code review rules.}

[jxnl]

This is great. I'll take a look this weekend after I see the tests pass.

[jeongyoonm]

@jxnl Checks are failing because API keys are not provided.

Please let me know if there's anything I can do regarding this. Thanks.

[jeongyoonm]

@jxnl Just to follow up on this. Checks are failing because API keys are not provided. Please let me know if there's a way to fix this on my end.

[jxnl]

Sorry, just life got ahead of me. Taking a look now.