[alexcastgonz (pcbeingused333)]

ChatRequest.Seed was serialized for Mistral as seed, but Mistral's chat completions take the value as random_seed, so a seed set on the request did not reach the model as the deterministic-sampling seed. random_seed could only be set through ChatRequestVendorMistralExtensions.RandomSeed.

The Mistral request now maps Seed to random_seed and no longer sends seed. ChatRequestVendorMistralExtensions.RandomSeed still takes precedence when set (before, the extension unconditionally overwrote the field, so a null extension value is now ignored instead of clearing the mapped seed).

Added MistralChatSeedTests (no network): Seed becomes random_seed without seed; the extension value wins over Seed; an extensions object without RandomSeed keeps the seed; no seed sends neither field. Three of the four fail on master.

This change was made with the help of an AI assistant (Claude Code); the tests were run locally on net8.0.

[Copilot]

🟢 Approval recommended

No active review issues are reported for the Mistral seed-parameter mapping change.

No open issues 🎉

What changed in this PR

This PR fixes seed handling in the Python Mistral AI connector by sending the SDK-supported random_seed parameter instead of seed.

Changes:

• Maps seed to random_seed, preserving an explicitly supplied random_seed.
• Adds unit tests for seed mapping and explicit-value precedence.
• Updates streaming and non-streaming service tests to expect random_seed.

File	Description
python/​tests/​unit/​connectors/​ai/​mistral_ai/​test_mistralai_request_settings.py	Adds seed mapping and precedence tests.
python/​tests/​unit/​connectors/​ai/​mistral_ai/​services/​test_mistralai_chat_completion.py	Updates expected SDK arguments for both completion modes.
python/​semantic_kernel/​connectors/​ai/​mistral_ai/​prompt_execution_settings/​mistral_ai_prompt_execution_settings.py	Converts seed to random_seed when preparing request settings.

_{Review effort: Balanced}