Mistral Moderation Thresholds for Inline Guardrails

Configure Mistral's inline moderation_llm_v2 guardrails (block_on_error: true, ignore_other_categories: false, action: block) with custom category thresholds to enforce automated safety blocking at the API layer.

Config: model="mistral-small-latest", guardrails=[{"block_on_error": True, "moderation_llm_v2": {"custom_category_thresholds": {"hate_and_discrimination": 0.1}, "ignore_other_categories": False, "action": "block"}}]