Fix enum validation error in Google GenAI provider

[DaveOkpare]

Changes

• Fix response parsing: Change model_validate() to model_validate_json() in parse_genai_tools() to properly
  handle string-to-enum conversion
• Fixes: Gemini cannot deal with Enums in the JSON schema #1756

---

Important

Fixes enum validation error in Google GenAI provider by changing model_validate() to model_validate_json() in parse_genai_tools().

• Behavior:
  • Change model_validate() to model_validate_json() in parse_genai_tools() to handle string-to-enum conversion properly.
• Fixes:
  • Addresses issue Gemini cannot deal with Enums in the JSON schema #1756 related to enum validation error in Google GenAI provider.

^{This description was created by for d28e76d. You can customize this summary. It will automatically update as commits are pushed.}

[ellipsis-dev]

Important

Looks good to me! 👍

Reviewed everything up to d28e76d in 48 seconds. Click for details.

• Reviewed 16 lines of code in 1 files
• Skipped 0 files when reviewing.
• Skipped posting 1 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. instructor/processing/function_calls.py:286

• Draft comment:
  Using json.dumps to wrap function_call.args and then calling model_validate_json is a clear fix for string-to-enum conversion (#1756). Please ensure that function_call.args is always JSON serializable to prevent runtime issues. Additionally, consider adding a test case that verifies enum conversion behavior.
• Reason this comment was not posted:
  Comment looked like it was already resolved.

Workflow ID: wflow_WILmjoF4w4JcHmG7

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

[jxnl]

this does not feel right / feels like would slow things randomly @claude can you figure out if we can add some tests or fix this another way

[claude]

Claude encountered an error —— View job

Failed with exit code 128

---

I'll analyze this and get back to you.

[DaveOkpare]

@jxnl Gemini’s structured output docs state:

“Passing JSON Schema directly is not yet supported when using the SDK.”

The SDK turns enum fields into strings with "enum": [...] constraints, so while using strict=True with Pydantic we have these options:

• model_validate() rejects these strings and add a fallback to model_validate_json() to parse valid enum strings into Enum members while preserving strictness.
• use model_validate_json() directly.

Performance test

Case	n=5	n=500	Winner
Needs coercion (strings)	11 µs vs 17 µs	1,090 µs vs 1,674 µs	model_validate_json
Already valid (Enums)	5 µs vs 13 µs	603 µs vs 1,294 µs	model_validate

Test compared using model_validate() with a fallback to model_validate_json() against using model_validate_json() for all the Gemini cases.