diff --git a/contracts/capability/multimodal.go b/contracts/capability/multimodal.go index c242192..c12946e 100644 --- a/contracts/capability/multimodal.go +++ b/contracts/capability/multimodal.go @@ -6,9 +6,12 @@ package capability type MultimodalOperation string const ( - MultimodalTranscribe MultimodalOperation = "TRANSCRIBE_AUDIO" - MultimodalUnderstand MultimodalOperation = "UNDERSTAND_IMAGE" - MultimodalRecognizeOCR MultimodalOperation = "RECOGNIZE_OCR" + 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 { @@ -17,4 +20,6 @@ type MultimodalResponse struct { 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"` } diff --git a/contracts/capability/types.go b/contracts/capability/types.go index 3e13aba..70ded9b 100644 --- a/contracts/capability/types.go +++ b/contracts/capability/types.go @@ -32,6 +32,9 @@ const ( CapabilityKindSpeechToText CapabilityKind = "SPEECH_TO_TEXT" CapabilityKindImageUnderstanding CapabilityKind = "IMAGE_UNDERSTANDING" CapabilityKindOCR CapabilityKind = "OCR" + CapabilityKindVideoUnderstanding CapabilityKind = "VIDEO_UNDERSTANDING" + CapabilityKindImageGeneration CapabilityKind = "IMAGE_GENERATION" + CapabilityKindVideoGeneration CapabilityKind = "VIDEO_GENERATION" ) // Binding is Capability-owned correctness state. SecretRef is an opaque