Sunflower Simple

Simple Sunflower inference endpoint for single instruction/response. This is a simplified interface for users who want to send a single instruction rather than managing conversation history. Uses form-based input for easier integration with simple clients. Args: request: The FastAPI request object (required for rate limiting). background_tasks: FastAPI background tasks for async operations. instruction: The question or instruction for the AI. model_type: Either 'qwen' (default) or 'gemma'. temperature: Controls randomness (0.0 = deterministic, 1.0 = creative). system_message: Optional custom system message. db: Database session. current_user: The authenticated user. Returns: Dictionary containing response, model_type, processing_time, usage, and success. Raises: BadRequestError: For validation errors. ValidationError: For invalid model type. ServiceUnavailableError: If the model is loading or request times out. ExternalServiceError: For unexpected errors. Example: Form data: - instruction: "Translate 'hello' to Luganda" - model_type: "qwen" - temperature: 0.3 Response: { "response": "In Luganda, 'hello' is 'Gyebaleko' or 'Wasuze otya' (Good morning).", "model_type": "qwen", "processing_time": 1.5, "usage": {"completion_tokens": 20, "prompt_tokens": 10, "total_tokens": 30}, "success": true }