package capability // MultimodalOperation identifies one bounded local media skill. It is kept // separate from general cognition so health, routing and failure evidence do // not collapse into a generic model response. type MultimodalOperation string const ( MultimodalTranscribe MultimodalOperation = "TRANSCRIBE_AUDIO" MultimodalUnderstand MultimodalOperation = "UNDERSTAND_IMAGE" MultimodalRecognizeOCR MultimodalOperation = "RECOGNIZE_OCR" MultimodalUnderstandVideo MultimodalOperation = "UNDERSTAND_VIDEO" MultimodalGenerateImage MultimodalOperation = "GENERATE_IMAGE" MultimodalGenerateVideo MultimodalOperation = "GENERATE_VIDEO" ) type MultimodalResponse struct { Operation MultimodalOperation `json:"operation"` Text string `json:"text"` Provider string `json:"provider,omitempty"` ObservedAt int64 `json:"observed_at_unix_millis"` ConfidenceNote string `json:"confidence_note,omitempty"` ContentType string `json:"content_type,omitempty"` DataBase64 string `json:"data_base64,omitempty"` }