[Qiming Bao (14H034160212)]

Motivation and Context

MistralAIChatPromptExecutionSettings rejects valid values for three fields (source):

• presence_penalty and frequency_penalty use Field(gt=0). That rejects 0, which means no penalty, and every negative value.
• n uses Field(gt=1), so n=1 is rejected.

This also breaks the generic settings path. A PromptExecutionSettings whose extension_data contains presence_penalty: 0 or frequency_penalty: 0 (common in prompt templates written for several providers) fails in MistralAIChatPromptExecutionSettings.from_prompt_execution_settings with a ValidationError:

from semantic_kernel.connectors.ai.mistral_ai import MistralAIChatPromptExecutionSettings
from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings

MistralAIChatPromptExecutionSettings(n=1)                   # ValidationError: greater than 1
MistralAIChatPromptExecutionSettings(presence_penalty=0.0)  # ValidationError: greater than 0
MistralAIChatPromptExecutionSettings(presence_penalty=-0.5) # ValidationError: greater than 0

generic = PromptExecutionSettings(extension_data={"max_tokens": 256, "presence_penalty": 0, "frequency_penalty": 0})
MistralAIChatPromptExecutionSettings.from_prompt_execution_settings(generic)  # ValidationError

The valid ranges, from primary sources:

• Mistral API spec: openapi.yaml, schema ChatCompletionRequest (used by POST /v1/chat/completions, see the API reference), defines presence_penalty and frequency_penalty as number | null with minimum: -2, maximum: 2, and n as integer | null with minimum: 1. AgentsCompletionRequest and CompletionArgs use the same penalty range.
• mistralai SDK (1.12.4, the version the mistralai extra resolves to): ChatCompletionRequest declares presence_penalty: Optional[float], frequency_penalty: Optional[float] and n: OptionalNullable[int] with no validators, so the SDK does not check the range. The service does.

Description

• presence_penalty and frequency_penalty: Field(gt=0) changed to Field(ge=-2.0, le=2.0), the range in the Mistral API spec.

• n: Field(gt=1) changed to Field(ge=1).

• Defaults stay None, so nothing extra is sent unless the user sets a value.

• Tests added to tests/unit/connectors/ai/mistral_ai/test_mistralai_request_settings.py:

  • both penalties accept -2.0, -0.5, 0.0, 0.5 and 2.0, and the value reaches prepare_settings_dict();
  • both penalties reject -2.1 and 2.1;
  • n accepts 1 and 2 and rejects 0;
  • from_prompt_execution_settings converts generic settings with presence_penalty=0, frequency_penalty=0 and n=1.

  10 of the new cases fail on main.

Verification (from python/):

uv run --no-sync pytest tests/unit/connectors/ai/mistral_ai     # 76 passed
uv run --no-sync ruff check <changed files> && uv run --no-sync ruff format --check <changed files>
uv run --no-sync mypy -p semantic_kernel.connectors.ai.mistral_ai --config-file mypy.ini   # Success: no issues found
uv run --no-sync pre-commit run -c python/.pre-commit-config.yaml --files <changed files>  # all hooks passed

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the SK Contribution Guidelines and the pre-submission formatting script raises no violations
• All unit tests pass, and I have added new tests where possible
• I didn't break anyone 😄

[Copilot]

🟢 Approval recommended

The review reports no active issues with the Mistral AI chat-settings validation changes.

No open issues 🎉

What changed in this PR

Corrects validation in Python’s Mistral AI chat settings to accept valid API values, including zero penalties and n=1, while preserving None defaults.

Changes:

• Accepts presence and frequency penalties in the inclusive range [-2, 2].
• Allows n >= 1.
• Adds regression tests for valid values, rejected values, serialization, and conversion from generic prompt settings.

File	Description
python/​tests/​unit/​connectors/​ai/​mistral_ai/​test_mistralai_request_settings.py	Adds validation, serialization, and generic-settings conversion tests.
python/​semantic_kernel/​connectors/​ai/​mistral_ai/​prompt_execution_settings/​mistral_ai_prompt_execution_settings.py	Corrects penalty bounds and the minimum value for n.

_{Review effort: Balanced}