fix(snowflake): transform tool_choice string to object format

[stevejaker]

Snowflake's Cortex API documentation is pretty inconsistent, but it requires tool_choice to be an object, not a string. For example, {"type": "auto"} instead of "auto".

Relevant issues

Fixes #23284

Pre-Submission checklist

Please complete all items before asking a LiteLLM maintainer to review your PR

• I have Added testing in the tests/test_litellm/ directory, Adding at least 1 test is a hard requirement - see details
• My PR passes all unit tests on make test-unit
• My PR's scope is as isolated as possible, it only solves 1 specific problem
• I have requested a Greptile review by commenting @greptileai and received a Confidence Score of at least 4/5 before requesting a maintainer review

CI (LiteLLM team)

CI status guideline:

• 50-55 passing tests: main is stable with minor issues.
• 45-49 passing tests: acceptable but needs attention
• <= 40 passing tests: unstable; be careful with your merges and assess the risk.

• Branch creation CI run
  Link:

• CI run for the last commit
  Link:

• Merge / cherry-pick CI run
  Links:

Type

🐛 Bug Fix

Changes

• Updated litellm/llms/snowflake/chat/transformation.py to transform string values for tool_choice to object format

Test plan

• Unit tests updated and passing
• Tested against real Snowflake Cortex API with tool calls

Ref: https://docs.snowflake.com/en/developer-guide/snowflake-rest-api/reference/cortex-inference#post--api-v2-cortex-inference-complete-req-body-schema

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Error		Mar 10, 2026 3:03pm

[stevejaker]

@greptileai

[greptile-apps]

Greptile Summary

This PR fixes a bug in the Snowflake Cortex integration where tool_choice string values ("auto", "required", "none") were being forwarded to the Snowflake API as raw strings, which the API does not accept. The fix wraps them as {"type": value} objects in _transform_tool_choice. The return type annotation is also correctly narrowed from Union[str, Dict[str, Any]] to Dict[str, Any].

Key changes:

• _transform_tool_choice in transformation.py now converts string tool choices to their object equivalent (e.g. "auto" → {"type": "auto"}) instead of passing them through unchanged.
• The updated test correctly asserts the new object format for all three string values.
• All tests remain pure unit tests with no real network calls, which aligns with the repository's test policy.

Implementation details:
The fix handles three input cases:

1. String values are wrapped as {"type": value}
2. Dict with type: "function" are converted to Snowflake's {"type": "tool", "name": [function_name]} format
3. Already-formatted Snowflake objects pass through unchanged

This is a minimal, targeted bug fix with no side effects on unrelated functionality.

Confidence Score: 5/5

• This PR is safe to merge — it's a targeted, well-tested bug fix with no side effects on unrelated functionality.
• The change is minimal and isolated to a single transformation method. The logic is straightforward and aligns with Snowflake's documented API requirement. The existing test was updated to match the new behavior, and no network calls are made in tests. No backwards-incompatible breakage is introduced since the previous string passthrough was already causing API errors. The return type annotation is correctly narrowed.
• No files require special attention.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[tool_choice input] --> B{isinstance str?}
    B -- Yes --> C["Return {type: tool_choice}\ne.g. 'auto' → {type: 'auto'}"]
    B -- No --> D{isinstance dict?}
    D -- No --> E[Return tool_choice as-is]
    D -- Yes --> F{type == 'function'?}
    F -- No --> E
    F -- Yes --> G{function_name present?}
    G -- No --> E
    G -- Yes --> H["Return {type: 'tool', name: [function_name]}"]

Loading

_{Last reviewed commit: 68617bf}