uneeq-js 3.16.0 → 3.16.2

package/dist/esm/chunks/deepgram-stt-QCVIQMXI.js

@@ -1 +0,0 @@
-import{a as b}from"./chunk-J5C63PLV.js";import{D as p,U as h,V as C,f as t,h as S,j as T,l as g,n as v,o as f,r as k,v as y,w as M}from"./chunk-OXR5BQBV.js";var n="[Deepgram STT]",E=1e4,U=250,R=48e3,$=1e3,N=3e4,D=2,d=5,l=2e3;var m=3,w=class{constructor(e){this.options=e;this.options.model=this.options.model||"nova-3",this.options.language=this.options.language||"en",this.options.smartFormat=this.options.smartFormat??!0,this.options.interimResults=this.options.interimResults??!0,this.options.utteranceEndMs=this.options.utteranceEndMs??1500,this.options.vadEvents=this.options.vadEvents??!0,this.options.fillerWords=this.options.fillerWords??!1,this.options.endpointing=this.options.endpointing??500,this.options.echoCancellation=this.options.echoCancellation??!0,this.options.noiseSuppression=this.options.noiseSuppression??!0,this.options.autoGainControl=this.options.autoGainControl??!0,this.options.interruptionWordThreshold=this.options.interruptionWordThreshold??m,this.options.noDelay=this.options.noDelay??!1,this.options.safetyNetTimeoutMs=this.options.safetyNetTimeoutMs??l,this.options.safetyNetTimeoutMs<=500?(t.warn(`${n} safetyNetTimeoutMs is set to ${this.options.safetyNetTimeoutMs}ms. This is very short and may cause premature transcript finalization. Ignoring and default the value to ${l}ms.`),this.options.safetyNetTimeoutMs=l):this.options.safetyNetTimeoutMs<=1e3?t.warn(`${n} safetyNetTimeoutMs is set to ${this.options.safetyNetTimeoutMs}ms. This is very short and may cause premature transcript finalization. Recommended value is ${l}ms.`):t.info(`${n} safetyNetTimeoutMs is set to ${this.options.safetyNetTimeoutMs}ms.`),this.handleAppMessages()}options;connection=null;state="Idle";shouldReconnect=!0;microphone=null;stream=null;reconnectAttempts=0;reconnectDelay=$;reconnectTimeoutId=null;digitalHumanSpeaking=!1;isUserCurrentlySpeaking=!1;isUiShowingSpeaking=!1;accumulatedTranscript="";accumulatedConfidenceSum=0;accumulatedWordCount=0;lastDeepgramEventTime=0;safetyNetTimeoutId=null;async startRecognition(){t.info(`${n} Starting speech recognition`),this.shouldReconnect=!0,this.resetReconnectionState(),await this.connect()}async stopRecognition(){t.info(`${n} Stopping speech recognition`),this.shouldReconnect=!1,this.clearReconnectTimeout(),await this.disconnect()}async pause(){return t.info(`${n} Pausing speech recognition`),this.state="Paused",this.safetyNetTimeoutId&&(clearTimeout(this.safetyNetTimeoutId),this.safetyNetTimeoutId=null),this.resetAccumulatedState(),this.resetSpeakingStates(),this.stream&&(this.stream.getTracks().forEach(e=>{e.enabled=!1}),t.debug(`${n} Audio tracks disabled`)),!0}async resume(){if(t.info(`${n} Resuming speech recognition`),this.state==="Paused"){if(this.stream)return this.state="Connected",this.stream.getTracks().forEach(e=>{e.enabled=!0}),t.debug(`${n} Audio tracks re-enabled`),!0;if(this.connection)return this.state="Connected",await this.startMicrophone(),!0;this.state="Disconnected"}return t.debug(`${n} Initiating connection`),await this.connect(),!0}setChatMetadata(e){this.options.promptMetadata=e}async connect(){if(this.state==="Connected"){t.warn(`${n} Already connected`);return}if(this.state==="Connecting"){t.warn(`${n} Connection already in progress`);return}this.state="Connecting";try{let e=await this.getToken();t.info(`${n} Connecting to Deepgram`);let i=new b({accessToken:e.token,baseUrl:e.api_url}),a={model:this.options.model,language:this.options.language,smart_format:String(this.options.smartFormat),interim_results:String(this.options.interimResults),utterance_end_ms:this.options.utteranceEndMs,vad_events:String(this.options.vadEvents),filler_words:String(this.options.fillerWords),endpointing:this.options.endpointing,mip_opt_out:"true",...this.options.keyterms&&this.options.keyterms.length>0&&{keyterm:this.options.keyterms},...this.options.noDelay&&{queryParams:{no_delay:"true"}}},s=i.listen;if(this.connection=await s.v1.connect(a),this.connection.connect(),await Promise.race([this.connection.waitForOpen(),new Promise((o,r)=>setTimeout(()=>r(new Error("Connection timeout")),E))]),this.state!=="Paused"&&(this.state="Connected"),t.info(`${n} Connection opened`),this.setupEventHandlers(),this.state==="Paused"){t.info(`${n} Pause requested during connection \u2014 staying paused`),this.resetReconnectionState();return}await this.startMicrophone(),t.info(`${n} Connected successfully`),this.resetReconnectionState()}catch(e){this.state="Disconnected",t.error(`${n} Connection error`,t.serialiseError(e)),this.shouldReconnect&&(this.emitTransientError(e),this.scheduleReconnect())}}async disconnect(){if(!(this.state==="Idle"||this.state==="Disconnected"&&!this.connection)){t.info(`${n} Disconnecting`),this.safetyNetTimeoutId&&(clearTimeout(this.safetyNetTimeoutId),this.safetyNetTimeoutId=null);try{this.stopMicrophone(),this.connection&&(this.connection.close(),this.connection=null)}catch(e){t.error(`${n} Disconnect error`,t.serialiseError(e))}this.resetAccumulatedState(),this.resetSpeakingStates(),this.state="Disconnected",this.clientMsgSend(new p(!1))}}scheduleReconnect(){if(this.reconnectAttempts>=d){t.error(`${n} Max reconnection attempts (${d}) reached`),this.clientMsgSend(new y(`Unable to connect to speech recognition service after ${d} attempts`));return}this.reconnectAttempts++,t.info(`${n} Scheduling reconnection attempt ${this.reconnectAttempts}/${d} in ${this.reconnectDelay}ms`),this.reconnectTimeoutId=setTimeout(()=>{this.connect()},this.reconnectDelay),this.reconnectDelay=Math.min(this.reconnectDelay*D,N)}resetReconnectionState(){this.reconnectAttempts=0,this.reconnectDelay=$,this.clearReconnectTimeout()}clearReconnectTimeout(){this.reconnectTimeoutId&&(clearTimeout(this.reconnectTimeoutId),this.reconnectTimeoutId=null)}async getToken(){let e=this.options.model||"nova-3",i=`${this.options.connectionUrl}/speech-recognition-service/deepgram/token?model=${encodeURIComponent(e)}`,a=await fetch(i,{method:"GET",headers:{Authorization:`Bearer ${this.options.jwtToken}`,"Content-Type":"application/json"}});if(!a.ok)throw new Error(`Token fetch failed: ${a.status} ${a.statusText}`);return await a.json()}async startMicrophone(){try{if(t.info(`${n} Starting microphone`),this.stopMicrophone(),!navigator.mediaDevices?.getUserMedia)throw new Error("Microphone access is not available in this context");if(this.stream=await navigator.mediaDevices.getUserMedia({audio:{deviceId:this.options.microphoneDeviceId?{exact:this.options.microphoneDeviceId}:void 0,echoCancellation:this.options.echoCancellation,noiseSuppression:this.options.noiseSuppression,autoGainControl:this.options.autoGainControl}}),this.state==="Paused"){t.info(`${n} Paused during getUserMedia \u2014 keeping stream but disabling tracks`),this.stream.getTracks().forEach(e=>{e.enabled=!1});return}this.microphone=new MediaRecorder(this.stream,{mimeType:"audio/webm;codecs=opus",audioBitsPerSecond:R}),this.microphone.ondataavailable=e=>{e.data.size>0&&this.connection&&this.state==="Connected"&&e.data.arrayBuffer().then(i=>{this.connection?.sendMedia(i)}).catch(i=>{t.error(`${n} Error converting audio data`,t.serialiseError(i))})},this.microphone.start(U),t.debug(`${n} Microphone started`),this.clientMsgSend(new p(!0))}catch(e){t.error(`${n} Microphone error`,t.serialiseError(e)),this.clientMsgSend(new T(new Error(JSON.stringify(e))))}}stopMicrophone(){this.microphone&&this.microphone.state==="recording"&&(this.microphone.stop(),this.microphone=null),this.stream&&(this.stream.getTracks().forEach(e=>{e.stop()}),this.stream=null),t.debug(`${n} Microphone stopped`)}setupEventHandlers(){this.connection&&(this.connection.on("open",()=>{this.handleConnectionOpen()}),this.connection.on("message",e=>{if(e!==null&&typeof e=="object"&&"type"in e){let i=e;i.type==="Results"?this.handleTranscript(e):i.type==="UtteranceEnd"?this.handleUtteranceEnd(e):i.type==="Metadata"&&t.debug(`${n} Metadata`,e)}}),this.connection.on("close",()=>{this.handleConnectionClose()}),this.connection.on("error",e=>{t.error(`${n} WebSocket error event`,t.serialiseError(e)),this.emitTransientError(e)}))}handleAppMessages(){this.options.messages.subscribe(e=>{switch(e.uneeqMessageType){case"AvatarStartedSpeaking":this.digitalHumanSpeaking=!0;break;case"PromptResult":{e.promptResult.success||this.handleSpeakingEnd();break}case"AvatarAnswer":{e.answerSpeech.replace(/<[^>]*>/g,"")===""&&this.handleSpeakingEnd();break}case"AvatarStoppedSpeaking":{this.handleSpeakingEnd();break}case"SessionEnded":{this.shouldReconnect=!1,this.stopRecognition();break}case"SessionReconnecting":{this.handleSpeakingEnd(),this.pause();break}case"CustomMetadataUpdated":{this.options.promptMetadata=e.chatMetadata;break}case"SessionBackendError":{this.handleSpeakingEnd();break}default:}})}handleTranscript(e){try{this.lastDeepgramEventTime=Date.now();let i=e.channel;if(!i?.alternatives||i.alternatives.length===0)return;let a=i.alternatives[0],s=String(a.transcript||""),o=e.is_final,r=e.speech_final;if(t.debug(`${n} Transcript event: is_final=${o}, speech_final=${r}, transcript="${s.substring(0,50)}${s.length>50?"...":""}", accumulated="${this.accumulatedTranscript.substring(0,30)}${this.accumulatedTranscript.length>30?"...":""}", uiSpeaking=${this.isUiShowingSpeaking}, userSpeaking=${this.isUserCurrentlySpeaking}`),s===""&&r){this.endsWithPunctuation(this.accumulatedTranscript)?(this.sendAccumulatedTranscript("Final transcript (from accumulated)"),this.resetSpeakingStates()):t.debug(`${n} speech_final with empty transcript but no punctuation, waiting for more speech or safety net`);return}if(s==="")return;this.processTranscriptChunk(s,a.confidence,o,r),this.updateSpeakingAndInterruptionState(s,r),this.emitTranscriptionResult(s,a.confidence,r),this.resetSafetyNetTimeout()}catch(i){t.error(`${n} Error processing transcript`,t.serialiseError(i))}}processTranscriptChunk(e,i,a,s){if(a&&!s){this.accumulatedTranscript!==""&&(this.accumulatedTranscript+=" "),this.accumulatedTranscript+=e;let o=e.trim().split(/\s+/).length,r=i??1;this.accumulatedConfidenceSum+=r*o,this.accumulatedWordCount+=o,t.debug(`${n} Accumulated transcript: "${this.accumulatedTranscript}"`)}}updateSpeakingAndInterruptionState(e,i){let s=(this.accumulatedTranscript!==""?this.accumulatedTranscript:e).trim().split(/\s+/).length,o=this.options.interruptionWordThreshold??m;this.isUiShowingSpeaking||(this.isUiShowingSpeaking=!0,this.clientMsgSend(new v));let r=!this.digitalHumanSpeaking||s>=o,c=this.accumulatedTranscript!==""?this.accumulatedTranscript+" "+e:e,u=i&&this.endsWithPunctuation(c);!this.isUserCurrentlySpeaking&&r&&(this.isUserCurrentlySpeaking=!0,this.dataChannelMsgSend(new h("start"))),this.digitalHumanSpeaking&&s>=o&&(t.debug(`${n} Interrupting digital human`),this.dataChannelMsgSend(new C),this.clientMsgSend(new k),this.digitalHumanSpeaking=!1),this.isUserCurrentlySpeaking&&u&&(this.isUserCurrentlySpeaking=!1,this.dataChannelMsgSend(new h("stop"))),this.isUiShowingSpeaking&&u&&(this.isUiShowingSpeaking=!1,this.clientMsgSend(new f))}emitTranscriptionResult(e,i,a){let s=this.accumulatedTranscript!==""?this.accumulatedTranscript+" "+e:e,o=a&&this.endsWithPunctuation(s),r={transcript:a?s:e,final:o,confidence:i??1,language_code:this.options.language||""};if(this.clientMsgSend(new g(r)),o){let c=s.trim().split(/\s+/).length,u=this.options.interruptionWordThreshold??m;this.digitalHumanSpeaking&&c<u?t.debug(`${n} Discarding utterance during speaking (${c} words < ${u} threshold): "${s}"`):(t.info(`${n} Final transcript: "${s}"`),this.sendChatPrompt(s)),this.resetAccumulatedState()}else if(a){this.accumulatedTranscript!==""&&(this.accumulatedTranscript+=" "),this.accumulatedTranscript+=e;let c=e.trim().split(/\s+/).length;this.accumulatedConfidenceSum+=(i??1)*c,this.accumulatedWordCount+=c,t.debug(`${n} speech_final without punctuation, accumulated for safety net: "${this.accumulatedTranscript}"`)}}handleUtteranceEnd(e){this.lastDeepgramEventTime=Date.now(),t.debug(`${n} UtteranceEnd event received: last_word_end=${e?.last_word_end}, accumulated="${this.accumulatedTranscript.substring(0,50)}...", uiSpeaking=${this.isUiShowingSpeaking}, userSpeaking=${this.isUserCurrentlySpeaking}`),this.accumulatedTranscript.trim()!==""?this.endsWithPunctuation(this.accumulatedTranscript)?(this.safetyNetTimeoutId&&(clearTimeout(this.safetyNetTimeoutId),this.safetyNetTimeoutId=null),t.debug(`${n} UtteranceEnd fallback triggered with transcript: "${this.accumulatedTranscript}"`),this.sendAccumulatedTranscript("UtteranceEnd fallback"),this.resetSpeakingStates(),t.debug(`${n} UtteranceEnd: reset speaking states`)):t.debug(`${n} UtteranceEnd: no punctuation, waiting for more speech or safety net`):(this.safetyNetTimeoutId&&(clearTimeout(this.safetyNetTimeoutId),this.safetyNetTimeoutId=null),this.resetSpeakingStates(),t.debug(`${n} UtteranceEnd: no transcript, reset speaking states`))}resetSafetyNetTimeout(){this.safetyNetTimeoutId&&(clearTimeout(this.safetyNetTimeoutId),this.safetyNetTimeoutId=null),(this.isUiShowingSpeaking||this.isUserCurrentlySpeaking)&&this.accumulatedTranscript.trim()!==""&&(this.safetyNetTimeoutId=setTimeout(()=>{this.triggerSafetyNet()},this.options.safetyNetTimeoutMs))}triggerSafetyNet(){t.warn(`${n} Safety net triggered: no Deepgram events for ${this.options.safetyNetTimeoutMs}ms while speaking`),t.debug(`${n} Safety net triggered: accumulated="${this.accumulatedTranscript}", uiSpeaking=${this.isUiShowingSpeaking}, userSpeaking=${this.isUserCurrentlySpeaking}, timeSinceLastEvent=${Date.now()-this.lastDeepgramEventTime}ms`),this.sendAccumulatedTranscript("Safety net"),this.resetSpeakingStates(),this.safetyNetTimeoutId=null}sendAccumulatedTranscript(e){if(this.accumulatedTranscript.trim()==="")return;if(e!=="Safety net"&&!this.endsWithPunctuation(this.accumulatedTranscript)){t.debug(`${n} sendAccumulatedTranscript: not sending as source is ${e} and transcript doesn't end in punctuation: "${this.accumulatedTranscript}"`);return}let i=this.options.interruptionWordThreshold??m,a=this.accumulatedWordCount>0?this.accumulatedConfidenceSum/this.accumulatedWordCount:1,s={transcript:this.accumulatedTranscript,final:!0,confidence:a,language_code:this.options.language||""};this.clientMsgSend(new g(s));let o=this.accumulatedTranscript.trim().split(/\s+/).length;this.digitalHumanSpeaking&&o<i?t.debug(`${n} Discarding utterance during speaking (${o} words < ${i} threshold): "${this.accumulatedTranscript}"`):(t.info(`${n} ${e}: "${this.accumulatedTranscript}"`),this.sendChatPrompt(this.accumulatedTranscript)),this.resetAccumulatedState()}endsWithPunctuation(e){let i=e.trim();return i.length===0?!1:/[.!?;:]$/.test(i)}resetAccumulatedState(){this.accumulatedTranscript="",this.accumulatedConfidenceSum=0,this.accumulatedWordCount=0}resetSpeakingStates(){this.isUserCurrentlySpeaking&&(this.isUserCurrentlySpeaking=!1,this.dataChannelMsgSend(new h("stop"))),this.isUiShowingSpeaking&&(this.isUiShowingSpeaking=!1,this.clientMsgSend(new f))}handleConnectionOpen(){this.state!=="Paused"&&(this.state="Connected")}handleConnectionClose(){if(t.info(`${n} Connection closed`),this.state==="Paused"){t.info(`${n} Connection closed while paused \u2014 will reconnect on resume`),this.connection=null,this.stopMicrophone();return}this.state="Disconnected",this.clientMsgSend(new p(!1)),this.shouldReconnect&&(t.info(`${n} Unexpected disconnect, attempting reconnection...`),this.scheduleReconnect())}sendChatPrompt(e){e&&e.trim()!==""&&(this.options.language&&(this.options.promptMetadata.userSpokenLocale=this.options.language),this.dataChannelMsgSend(new S(e,this.options.promptMetadata)))}handleSpeakingEnd(){this.digitalHumanSpeaking=!1}emitTransientError(e){let i=e instanceof Error?e.message:String(e);this.clientMsgSend(new M(i))}dataChannelMsgSend(e){this.options.sendMessage(e)}clientMsgSend(e){this.options.messages.next(e)}};export{w as DeepgramSTT};