Implements vad_events for SpeechStarted

examples/streaming/microphone/main.go

@@ -28,7 +28,7 @@ func (c MyCallback) Message(mr *api.MessageResponse) error {
 	if len(mr.Channel.Alternatives) == 0 || len(sentence) == 0 {
 		return nil
 	}
-	fmt.Printf("\n%s\n", sentence)
+	fmt.Printf("\nspeaker: %s\n", sentence)
 	return nil
 }
 
@@ -41,6 +41,11 @@ func (c MyCallback) Metadata(md *api.MetadataResponse) error {
 	return nil
 }
 
+func (c MyCallback) SpeechStarted(ssr *api.SpeechStartedResponse) error {
+	fmt.Printf("\n[SpeechStarted] Received\n")
+	return nil
+}
+
 func (c MyCallback) UtteranceEnd(ur *api.UtteranceEndResponse) error {
 	fmt.Printf("\n[UtteranceEnd] Received\n")
 	return nil
@@ -85,8 +90,9 @@ func main() {
 		SampleRate:  16000,
 		SmartFormat: true,
 		// To get UtteranceEnd, the following must be set:
-		// InterimResults: true,
-		// UtteranceEndMs: "1000",
+		InterimResults: true,
+		UtteranceEndMs: "1000",
+		VadEvents:      true,
 	}
 
 	// example on how to send a custom parameter

pkg/api/live/v1/default.go

@@ -95,6 +95,41 @@ func (dch DefaultCallbackHandler) Metadata(md *interfaces.MetadataResponse) erro
 	return nil
 }
 
+func (dch DefaultCallbackHandler) SpeechStarted(ssr *interfaces.SpeechStartedResponse) error {
+	var debugStr string
+	if v := os.Getenv("DEEPGRAM_DEBUG"); v != "" {
+		klog.V(4).Infof("DEEPGRAM_DEBUG found")
+		debugStr = v
+	}
+
+	if strings.Compare(strings.ToLower(debugStr), "true") == 0 {
+		data, err := json.Marshal(ssr)
+		if err != nil {
+			klog.V(1).Infof("SpeechStarted json.Marshal failed. Err: %v\n", err)
+			return err
+		}
+
+		prettyJson, err := prettyjson.Format(data)
+		if err != nil {
+			klog.V(1).Infof("prettyjson.Marshal failed. Err: %v\n", err)
+			return err
+		}
+		klog.V(2).Infof("\n\nSpeechStarted Object:\n%s\n\n", prettyJson)
+
+		return nil
+	}
+
+	// handle the message
+	fmt.Printf("\nSpeechStarted.Timestamp: %f\n", ssr.Timestamp)
+	fmt.Printf("SpeechStarted.Channels:\n")
+	for _, val := range ssr.Channel {
+		fmt.Printf("\tChannel: %d\n", val)
+	}
+	fmt.Printf("\n")
+
+	return nil
+}
+
 func (dch DefaultCallbackHandler) UtteranceEnd(ur *interfaces.UtteranceEndResponse) error {
 	fmt.Printf("\nUtteranceEnd\n")
 	return nil

pkg/api/live/v1/interfaces/constants.go

@@ -1,4 +1,4 @@
-// Copyright 2023 Deepgram SDK contributors. All Rights Reserved.
+// Copyright 2023-2024 Deepgram SDK contributors. All Rights Reserved.
 // Use of this source code is governed by a MIT license that can be found in the LICENSE file.
 // SPDX-License-Identifier: MIT
 
@@ -7,9 +7,10 @@ package interfaces
 // These are the message types that can be received from the live API
 const (
 	// message types
-	TypeMessageResponse      string = "Results"
-	TypeMetadataResponse     string = "Metadata"
-	TypeUtteranceEndResponse string = "UtteranceEnd"
+	TypeMessageResponse       string = "Results"
+	TypeMetadataResponse      string = "Metadata"
+	TypeUtteranceEndResponse  string = "UtteranceEnd"
+	TypeSpeechStartedResponse string = "SpeechStarted"
 
 	// Error type
 	TypeErrorResponse string = "Error"

pkg/api/live/v1/interfaces/interfaces.go

@@ -9,6 +9,7 @@ package interfaces
 type LiveMessageCallback interface {
 	Message(mr *MessageResponse) error
 	Metadata(md *MetadataResponse) error
+	SpeechStarted(ssr *SpeechStartedResponse) error
 	UtteranceEnd(ur *UtteranceEndResponse) error
 	Error(er *ErrorResponse) error
 	// TODO: implement other conversation insights

pkg/api/live/v1/interfaces/types.go

@@ -78,6 +78,12 @@ type UtteranceEndResponse struct {
 	LastWordEnd float64 `json:"last_word_end,omitempty"`
 }
 
+type SpeechStartedResponse struct {
+	Type      string  `json:"type,omitempty"`
+	Channel   []int   `json:"channel,omitempty"`
+	Timestamp float64 `json:"timestamp,omitempty"`
+}
+
 // ErrorResponse is the response from a live transcription
 type ErrorResponse struct {
 	Description string `json:"description"`

pkg/api/live/v1/router.go

@@ -65,6 +65,8 @@ func (r *MessageRouter) Message(byMsg []byte) error {
 		err = r.MessageResponse(byMsg)
 	case interfaces.TypeMetadataResponse:
 		err = r.MetadataResponse(byMsg)
+	case interfaces.TypeSpeechStartedResponse:
+		err = r.SpeechStartedResponse(byMsg)
 	case interfaces.TypeUtteranceEndResponse:
 		err = r.UtteranceEndResponse(byMsg)
 	case interfaces.TypeErrorResponse:
@@ -150,6 +152,37 @@ func (r *MessageRouter) MetadataResponse(byMsg []byte) error {
 	return nil
 }
 
+func (r *MessageRouter) SpeechStartedResponse(byMsg []byte) error {
+	klog.V(6).Infof("router.SpeechStartedResponse ENTER\n")
+
+	// trace debugging
+	r.printDebugMessages(5, "SpeechStartedResponse", byMsg)
+
+	var ssr interfaces.SpeechStartedResponse
+	err := json.Unmarshal(byMsg, &ssr)
+	if err != nil {
+		klog.V(1).Infof("SpeechStartedResponse json.Unmarshal failed. Err: %v\n", err)
+		klog.V(6).Infof("router.SpeechStartedResponse LEAVE\n")
+		return err
+	}
+
+	if r.callback != nil {
+		err := r.callback.SpeechStarted(&ssr)
+		if err != nil {
+			klog.V(1).Infof("callback.SpeechStartedResponse failed. Err: %v\n", err)
+		} else {
+			klog.V(5).Infof("callback.SpeechStartedResponse succeeded\n")
+		}
+		klog.V(6).Infof("router.SpeechStartedResponse LEAVE\n")
+		return err
+	}
+
+	klog.V(1).Infof("User callback is undefined\n")
+	klog.V(6).Infof("router.SpeechStartedResponse ENTER\n")
+
+	return nil
+}
+
 func (r *MessageRouter) UtteranceEndResponse(byMsg []byte) error {
 	klog.V(6).Infof("router.UtteranceEndResponse ENTER\n")
 

pkg/client/interfaces/types-stream.go

@@ -36,5 +36,6 @@ type LiveTranscriptionOptions struct {
 	SmartFormat     bool     `json:"smart_format,omitempty" url:"smart_format,omitempty"`
 	Tag             []string `json:"tag,omitempty" url:"tag,omitempty"`
 	UtteranceEndMs  string   `json:"utterance_end_ms,omitempty" url:"utterance_end_ms,omitempty"`
+	VadEvents       bool     `json:"vad_events,omitempty" url:"vad_events,omitempty"`
 	Version         string   `json:"version,omitempty" url:"version,omitempty"`
 }