@lssm-tech/lib.voice 1.7.29 → 3.0.2

package/dist/index.js

@@ -1,14 +0,0 @@
-// @bun
-class A{concatenate(q){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[J]=q;if(!J)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(q.length===1)return{...J};let{format:U,sampleRateHz:K}=J,X=J.channels??1;for(let V of q){if(V.format!==U)throw Error(`Format mismatch: expected ${U}, got ${V.format}`);if(V.sampleRateHz!==K)throw Error(`Sample rate mismatch: expected ${K}, got ${V.sampleRateHz}`)}let Q=q.reduce((V,L)=>V+L.data.length,0),Z=new Uint8Array(Q),$=0;for(let V of q)Z.set(V.data,$),$+=V.data.length;let Y=q.reduce((V,L)=>V+(L.durationMs??0),0);return{data:Z,format:U,sampleRateHz:K,durationMs:Y,channels:X}}}class G{static DEFAULT_WPM=150;estimateSeconds(q,J){let U=J??G.DEFAULT_WPM,K=q.split(/\s+/).filter(Boolean).length;return Math.ceil(K/U*60)}estimateMs(q,J){let U=J??G.DEFAULT_WPM,K=q.split(/\s+/).filter(Boolean).length;return Math.ceil(K/U*60*1000)}estimateWordCount(q,J){let U=J??G.DEFAULT_WPM;return Math.round(q/60*U)}}class P{convert(q,J){if(q.format===J)return q;return{...q,format:J}}isSupported(q,J){if(q===J)return!0;return{wav:["mp3","ogg","pcm","opus"],mp3:["wav"],ogg:["wav"],pcm:["wav"],opus:["wav"]}[q]?.includes(J)??!1}}class E{generate(q,J="wav",U=44100,K=1){let X=Math.ceil(U*q/1000),Q=2,Z=X*2*K;return{data:new Uint8Array(Z),format:J,sampleRateHz:U,durationMs:q,channels:K}}}class O{stt;llm;tts;modelSelector;selectionContext;conversationHistory=[];constructor(q,J,U){if("stt"in q)this.stt=q.stt,this.llm=q.llm,this.tts=q.tts,this.modelSelector=q.modelSelector,this.selectionContext=q.selectionContext;else{if(!J||!U)throw Error("ResponseOrchestrator requires llm and tts when constructed with positional arguments");this.stt=q,this.llm=J,this.tts=U}}async*processUserTurn(q,J){let K=(await this.stt.transcribe({audio:q,language:J.language,wordTimestamps:!1})).text;yield{type:"user_speech_ended",transcript:K},yield{type:"transcript",role:"user",text:K,timestamp:Date.now()},this.conversationHistory.push({role:"user",content:K});let X=await this.resolveModel(J.llmModel),Z=(await this.llm.chat([{role:"system",content:[{type:"text",text:J.systemPrompt}]},...this.conversationHistory.map((V)=>({role:V.role,content:[{type:"text",text:V.content}]}))],{model:X})).message.content.find((V)=>V.type==="text"),$=Z&&Z.type==="text"?Z.text:"I apologize, I could not generate a response.";this.conversationHistory.push({role:"assistant",content:$}),yield{type:"agent_speech_started",text:$},yield{type:"agent_audio",audio:(await this.tts.synthesize({text:$,voiceId:J.voiceId,language:J.language,format:J.outputFormat})).audio.data},yield{type:"agent_speech_ended"},yield{type:"transcript",role:"agent",text:$,timestamp:Date.now()}}async resolveModel(q){if(q)return q;if(this.modelSelector){let J=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(J)).modelId}return}reset(){this.conversationHistory.length=0}}class I{turns=[];currentTurn=null;sessionStartMs=Date.now();getTranscript(){return[...this.turns]}toText(){return this.turns.map((q)=>`[${q.role}] ${q.text}`).join(`
-`)}getTurnCount(){return this.turns.length}processEvent(q){switch(q.type){case"session_started":this.sessionStartMs=Date.now();break;case"user_speech_started":this.currentTurn={role:"user",startMs:Date.now()-this.sessionStartMs};break;case"user_speech_ended":if(this.currentTurn&&this.currentTurn.role==="user")this.currentTurn.text=q.transcript,this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"agent_speech_started":this.currentTurn={role:"agent",text:q.text,startMs:Date.now()-this.sessionStartMs};break;case"agent_speech_ended":if(this.currentTurn&&this.currentTurn.role==="agent")this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"transcript":break;default:break}}reset(){this.turns.length=0,this.currentTurn=null,this.sessionStartMs=Date.now()}}class k{silenceThresholdMs;energyThreshold;silenceStartMs=null;constructor(q=800,J=0.01){this.silenceThresholdMs=q,this.energyThreshold=J}processChunk(q,J){if(this.calculateEnergy(q)>this.energyThreshold)return this.silenceStartMs=null,!1;if(this.silenceStartMs===null)this.silenceStartMs=J;return J-this.silenceStartMs>=this.silenceThresholdMs}reset(){this.silenceStartMs=null}calculateEnergy(q){if(q.length<2)return 0;let J=0,U=Math.floor(q.length/2);for(let K=0;K<q.length-1;K+=2){let X=q[K]??0,Q=q[K+1]??0,$=((X|Q<<8)<<16>>16)/32768;J+=$*$}return Math.sqrt(J/U)}}class D{provider;constructor(q){this.provider=q.conversational}async startSession(q){let J=new I,U=await this.provider.startSession({voiceId:q.voiceId,language:q.language,systemPrompt:q.systemPrompt,llmModel:q.llmModel,inputFormat:q.inputFormat,outputFormat:q.outputFormat,turnDetection:q.turnDetection,silenceThresholdMs:q.silenceThresholdMs,maxDurationSeconds:q.maxDurationSeconds}),K={sessionId:"",status:"connecting",currentTurn:"idle",turnCount:0,durationMs:0,transcript:[]},X=this.wrapEvents(U.events,K,J);return{state:K,sendAudio:(Q)=>U.sendAudio(Q),sendText:(Q)=>U.sendText(Q),interrupt:()=>U.interrupt(),close:async()=>{let Q=await U.close();return K.status="ended",Q},events:X}}async*wrapEvents(q,J,U){for await(let K of q){switch(U.processEvent(K),K.type){case"session_started":J.sessionId=K.sessionId,J.status="active";break;case"user_speech_started":J.currentTurn="user";break;case"user_speech_ended":J.currentTurn="idle",J.turnCount+=1;break;case"agent_speech_started":J.currentTurn="agent";break;case"agent_speech_ended":J.currentTurn="idle",J.turnCount+=1;break;case"session_ended":J.status="ended",J.durationMs=K.durationMs;break}J.transcript=U.getTranscript(),yield K}}}class T{map(q,J="Speaker"){let U=[],K=new Map;for(let Z of q)if(Z.speakerId&&!U.includes(Z.speakerId))U.push(Z.speakerId),K.set(Z.speakerId,{segmentCount:0,totalSpeakingMs:0});let X=q.map((Z)=>{if(!Z.speakerId)return Z;let $=U.indexOf(Z.speakerId),Y=`${J} ${$+1}`,V=K.get(Z.speakerId);if(!V)return{...Z,speakerLabel:Y};return V.segmentCount+=1,V.totalSpeakingMs+=Z.endMs-Z.startMs,{...Z,speakerLabel:Y}}),Q=U.map((Z,$)=>{let Y=K.get(Z);return{id:Z,label:`${J} ${$+1}`,segmentCount:Y?.segmentCount??0,totalSpeakingMs:Y?.totalSpeakingMs??0}});return{segments:X,speakers:Q}}}class C{static DEFAULT_MAX_CHUNK_MS=300000;split(q,J=C.DEFAULT_MAX_CHUNK_MS){let U=q.durationMs??this.estimateDurationMs(q);if(U<=J)return[q];let K=[],X=q.data.length/Math.max(U,1),Q=0;while(Q<U){let Z=Math.min(J,U-Q),$=Math.floor(Q*X),Y=Math.floor((Q+Z)*X);K.push({data:q.data.slice($,Y),format:q.format,sampleRateHz:q.sampleRateHz,durationMs:Z,channels:q.channels}),Q+=Z}return K}estimateDurationMs(q){let U=q.channels??1,K=q.data.length/(2*U);return Math.ceil(K/q.sampleRateHz*1000)}}class N{toSRT(q){return q.map((J,U)=>{let K=this.formatTimeSRT(J.startMs),X=this.formatTimeSRT(J.endMs),Q=J.speakerLabel?`[${J.speakerLabel}] `:"";return`${U+1}
-${K} --> ${X}
-${Q}${J.text}`}).join(`
-
-`)}toVTT(q){return`WEBVTT
-
-`+q.map((K,X)=>{let Q=this.formatTimeVTT(K.startMs),Z=this.formatTimeVTT(K.endMs),$=K.speakerLabel?`<v ${K.speakerLabel}>`:"";return`${X+1}
-${Q} --> ${Z}
-${$}${K.text}`}).join(`
-
-`)}formatTimeSRT(q){let J=Math.floor(q/3600000),U=Math.floor(q%3600000/60000),K=Math.floor(q%60000/1000),X=q%1000;return`${this.pad(J,2)}:${this.pad(U,2)}:${this.pad(K,2)},${this.pad(X,3)}`}formatTimeVTT(q){let J=Math.floor(q/3600000),U=Math.floor(q%3600000/60000),K=Math.floor(q%60000/1000),X=q%1000;return`${this.pad(J,2)}:${this.pad(U,2)}:${this.pad(K,2)}.${this.pad(X,3)}`}pad(q,J){return q.toString().padStart(J,"0")}}class x{stt;segmentSplitter=new C;diarizationMapper=new T;subtitleFormatter=new N;constructor(q){this.stt=q.stt}async transcribe(q){let J=b(),U=this.segmentSplitter.split(q.audio),K=[],X="",Q=0,Z=0;for(let W of U){let R=await this.stt.transcribe({audio:W,language:q.language,diarize:q.diarize,speakerCount:q.speakerCount,wordTimestamps:!0,vocabularyHints:q.vocabularyHints}),h=R.segments.map((H)=>({text:H.text,startMs:H.startMs+Z,endMs:H.endMs+Z,speakerId:H.speakerId,speakerName:H.speakerName,confidence:H.confidence}));K.push(...h),X+=(X?" ":"")+R.text,Q+=R.durationMs,Z+=W.durationMs??0}let $=K,Y;if(q.diarize){let W=this.diarizationMapper.map(K);$=W.segments,Y=W.speakers}let V={text:X,segments:$,language:q.language??"en",durationMs:Q},L,_=q.subtitleFormat??"none";if(_==="srt")L=this.subtitleFormatter.toSRT($);else if(_==="vtt")L=this.subtitleFormatter.toVTT($);return{id:J,transcript:V,subtitles:L,speakers:Y}}async*transcribeStream(q,J){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let U=this.stt.transcribeStream(q,{language:J?.language,diarize:J?.diarize,speakerCount:J?.speakerCount,wordTimestamps:!0,vocabularyHints:J?.vocabularyHints});for await(let K of U)yield{text:K.text,startMs:K.startMs,endMs:K.endMs,speakerId:K.speakerId,speakerLabel:K.speakerName,confidence:K.confidence}}}function b(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`stt_${q}_${J}`}class B{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(q,J){let U=[],K=q.segments.map((X)=>{let Q=J.get(X.sceneId);if(Q===void 0)return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:X.recommendedSceneDurationInFrames,voiceDurationInFrames:X.durationInFrames,action:"no_change",finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X;let Z=X.durationInFrames/Q;if(Z>B.UPPER_THRESHOLD){let $=Math.min(Z,B.MAX_RATE);return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:Z>B.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:$,finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X}if(Z<B.LOWER_THRESHOLD){let $=Math.max(Z,B.MIN_RATE);return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:"pad_silence",suggestedRate:$,finalSceneDurationInFrames:Q}),{...X,recommendedSceneDurationInFrames:Q}}return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:"no_change",finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X});return{timingMap:{...q,segments:K},adjustments:U}}}class S{durationEstimator=new G;adapt(q){let J=q.scenes.filter((Q)=>Q.narrationText&&Q.narrationText.trim().length>0),U=J.map((Q,Z)=>{let $=Q.narrationText??"";return{sceneId:Q.id,text:$,estimatedDurationSeconds:this.durationEstimator.estimateSeconds($),contentType:this.inferContentType(Z,J.length)}}),K=U.map((Q)=>Q.text).join(" "),X=U.reduce((Q,Z)=>Q+Z.estimatedDurationSeconds,0);return{fullText:K,segments:U,estimatedDurationSeconds:X}}inferContentType(q,J){if(q===0)return"intro";if(q===J-1)return"cta";if(q===1&&J>3)return"problem";if(q===J-2&&J>3)return"metric";return"solution"}}class w{calculate(q,J,U=1.15){let K=q.map((Q)=>{let Z=Math.ceil(Q.durationMs/1000*J),$=Math.ceil(Z*U),Y=Q.wordTimings?.map((V)=>({word:V.word,startMs:V.startMs,endMs:V.endMs}));return{sceneId:Q.sceneId,durationMs:Q.durationMs,durationInFrames:Z,recommendedSceneDurationInFrames:$,wordTimings:Y}});return{totalDurationMs:q.reduce((Q,Z)=>Q+Z.durationMs,0),segments:K,fps:J}}recalculateForFps(q,J){let U=q.segments.map((K)=>{let X=Math.ceil(K.durationMs/1000*J),Q=K.recommendedSceneDurationInFrames/Math.max(K.durationInFrames,1);return{...K,durationInFrames:X,recommendedSceneDurationInFrames:Math.ceil(X*Q)}});return{...q,segments:U,fps:J}}}class z{concatenator=new A;silenceGenerator=new E;assemble(q,J,U=500){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[K]=q;if(!K)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let X=new Map(J.map(($)=>[$.sceneId,$])),Q=K.audio,Z=[];for(let $=0;$<q.length;$++){let Y=q[$];if(!Y)continue;let V=X.get(Y.sceneId),L=V?.leadingSilenceMs??0;if(L>0)Z.push(this.silenceGenerator.generate(L,Q.format,Q.sampleRateHz,Q.channels??1));Z.push(Y.audio);let _=V?.trailingSilenceMs??($<q.length-1?U:0);if(_>0)Z.push(this.silenceGenerator.generate(_,Q.format,Q.sampleRateHz,Q.channels??1))}return this.concatenator.concatenate(Z)}}var v={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class j{analyze(q,J=1){return q.map((U)=>{let K=v[U.contentType];return{sceneId:U.sceneId,rate:K.rate*J,emphasis:K.emphasis,tone:K.tone,leadingSilenceMs:K.leadingSilenceMs,trailingSilenceMs:K.trailingSilenceMs}})}getDefaults(q){return{...v[q]}}}class F{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(q){this.llm=q?.llm,this.model=q?.model,this.modelSelector=q?.modelSelector,this.selectionContext=q?.selectionContext,this.paceAnalyzer=new j}async plan(q,J=1){if(!this.llm)return this.paceAnalyzer.analyze(q,J);try{return await this.planWithLlm(q,J)}catch{return this.paceAnalyzer.analyze(q,J)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let q=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(q)).modelId}return}async planWithLlm(q,J){if(!this.llm)return this.paceAnalyzer.analyze(q,J);let U=await this.resolveModel(),X=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(q.map((Z)=>({sceneId:Z.sceneId,text:Z.text,contentType:Z.contentType})))}]}],{model:U,temperature:0.3,responseFormat:"json"})).message.content.find((Z)=>Z.type==="text");if(!X||X.type!=="text")return this.paceAnalyzer.analyze(q,J);return JSON.parse(X.text).map((Z)=>({...Z,rate:Z.rate*J}))}}class y{tts;constructor(q){this.tts=q}async synthesizeAll(q,J,U){let K=new Map(U.map((Q)=>[Q.sceneId,Q]));return await Promise.all(q.map((Q)=>this.synthesizeOne(Q,J,K.get(Q.sceneId))))}async synthesizeOne(q,J,U){let K=await this.tts.synthesize({text:q.text,voiceId:J.voiceId,language:J.language,style:J.style,stability:J.stability,rate:U?.rate,emphasis:U?.emphasis});return{sceneId:q.sceneId,audio:K.audio,durationMs:K.audio.durationMs??0,wordTimings:K.wordTimings?.map((X)=>({word:X.word,startMs:X.startMs,endMs:X.endMs}))}}}class M{segmentSynthesizer;emphasisPlanner;audioAssembler=new z;durationEstimator=new G;paceAnalyzer=new j;options;constructor(q){this.options=q,this.segmentSynthesizer=new y(q.tts),this.emphasisPlanner=new F({llm:q.llm,model:q.model})}async synthesize(q){let J=this.buildScript(q);return this.executePipeline(J,q.voice,q.pacing)}async synthesizeForVideo(q){let J=this.buildScriptFromScenePlan(q);return this.executePipeline(J,q.voice,q.pacing,q.fps)}async executePipeline(q,J,U,K){let X=f(),Q=U?.baseRate??1,Z=await this.emphasisPlanner.plan(q.segments,Q),$=await this.segmentSynthesizer.synthesizeAll(q.segments,J,Z),Y=U?.segmentPauseMs??500,V=this.audioAssembler.assemble($,Z,Y),L=K??this.options.fps??30,_=U?.breathingRoomFactor??1.15,W=this.buildTimingMap($,L,_);return{id:X,script:q,pacingDirectives:Z,segments:$,assembledAudio:V,timingMap:W}}buildScript(q){let J=[],U=`${q.content.title}. ${q.content.summary}`;if(J.push({sceneId:"intro",text:U,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(U),contentType:"intro"}),q.content.problems.length>0){let Q=q.content.problems.join(". ");J.push({sceneId:"problems",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"problem"})}if(q.content.solutions.length>0){let Q=q.content.solutions.join(". ");J.push({sceneId:"solutions",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"solution"})}if(q.content.metrics&&q.content.metrics.length>0){let Q=q.content.metrics.join(". ");J.push({sceneId:"metrics",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"metric"})}if(q.content.callToAction)J.push({sceneId:"cta",text:q.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(q.content.callToAction),contentType:"cta"});let K=J.map((Q)=>Q.text).join(" "),X=J.reduce((Q,Z)=>Q+Z.estimatedDurationSeconds,0);return{fullText:K,segments:J,estimatedDurationSeconds:X}}buildScriptFromScenePlan(q){let J=q.scenePlan.scenes.filter((X)=>X.narrationText).map((X)=>{let Q=X.narrationText??"";return{sceneId:X.id,text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"intro"}}),U=J.map((X)=>X.text).join(" "),K=J.reduce((X,Q)=>X+Q.estimatedDurationSeconds,0);return{fullText:U,segments:J,estimatedDurationSeconds:K}}buildTimingMap(q,J,U){let K=q.map((Q)=>{let Z=Math.ceil(Q.durationMs/1000*J);return{sceneId:Q.sceneId,durationMs:Q.durationMs,durationInFrames:Z,recommendedSceneDurationInFrames:Math.ceil(Z*U),wordTimings:Q.wordTimings?.map(($)=>({word:$.word,startMs:$.startMs,endMs:$.endMs}))}});return{totalDurationMs:q.reduce((Q,Z)=>Q+Z.durationMs,0),segments:K,fps:J}}}function f(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`tts_${q}_${J}`}import{defineFeature as p}from"@contractspec/lib.contracts-spec/features";var gq=p({meta:{key:"libs.voice",version:"1.0.0",title:"Voice",description:"Voice capabilities: TTS, STT, and conversational AI",domain:"voice",owners:["@contractspec-core"],tags:["package","libs","voice"],stability:"experimental"}});export{M as VoiceSynthesizer,D as VoiceSessionManager,gq as VoiceFeature,k as TurnDetector,I as TranscriptBuilder,x as Transcriber,w as TimingCalculator,N as SubtitleFormatter,E as SilenceGenerator,y as SegmentSynthesizer,C as SegmentSplitter,S as SceneAdapter,O as ResponseOrchestrator,j as PaceAnalyzer,P as FormatConverter,F as EmphasisPlanner,B as DurationNegotiator,G as DurationEstimator,T as DiarizationMapper,A as AudioConcatenator,z as AudioAssembler};

package/dist/node/audio/audio-concatenator.js

@@ -1 +0,0 @@
-class L{concatenate(k){if(k.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[q]=k;if(!q)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(k.length===1)return{...q};let{format:x,sampleRateHz:z}=q,H=q.channels??1;for(let j of k){if(j.format!==x)throw Error(`Format mismatch: expected ${x}, got ${j.format}`);if(j.sampleRateHz!==z)throw Error(`Sample rate mismatch: expected ${z}, got ${j.sampleRateHz}`)}let J=k.reduce((j,A)=>j+A.data.length,0),E=new Uint8Array(J),G=0;for(let j of k)E.set(j.data,G),G+=j.data.length;let K=k.reduce((j,A)=>j+(A.durationMs??0),0);return{data:E,format:x,sampleRateHz:z,durationMs:K,channels:H}}}export{L as AudioConcatenator};

package/dist/node/audio/duration-estimator.js

@@ -1 +0,0 @@
-class q{static DEFAULT_WPM=150;estimateSeconds(b,h){let j=h??q.DEFAULT_WPM,v=b.split(/\s+/).filter(Boolean).length;return Math.ceil(v/j*60)}estimateMs(b,h){let j=h??q.DEFAULT_WPM,v=b.split(/\s+/).filter(Boolean).length;return Math.ceil(v/j*60*1000)}estimateWordCount(b,h){let j=h??q.DEFAULT_WPM;return Math.round(b/60*j)}}export{q as DurationEstimator};

package/dist/node/audio/format-converter.js

@@ -1 +0,0 @@
-class k{convert(c,j){if(c.format===j)return c;return{...c,format:j}}isSupported(c,j){if(c===j)return!0;return{wav:["mp3","ogg","pcm","opus"],mp3:["wav"],ogg:["wav"],pcm:["wav"],opus:["wav"]}[c]?.includes(j)??!1}}export{k as FormatConverter};

package/dist/node/audio/index.js

@@ -1 +0,0 @@
-class Q{concatenate(j){if(j.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[q]=j;if(!q)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(j.length===1)return{...q};let{format:v,sampleRateHz:A}=q,K=q.channels??1;for(let x of j){if(x.format!==v)throw Error(`Format mismatch: expected ${v}, got ${x.format}`);if(x.sampleRateHz!==A)throw Error(`Sample rate mismatch: expected ${A}, got ${x.sampleRateHz}`)}let O=j.reduce((x,N)=>x+N.data.length,0),J=new Uint8Array(O),L=0;for(let x of j)J.set(x.data,L),L+=x.data.length;let X=j.reduce((x,N)=>x+(N.durationMs??0),0);return{data:J,format:v,sampleRateHz:A,durationMs:X,channels:K}}}class G{static DEFAULT_WPM=150;estimateSeconds(j,q){let v=q??G.DEFAULT_WPM,A=j.split(/\s+/).filter(Boolean).length;return Math.ceil(A/v*60)}estimateMs(j,q){let v=q??G.DEFAULT_WPM,A=j.split(/\s+/).filter(Boolean).length;return Math.ceil(A/v*60*1000)}estimateWordCount(j,q){let v=q??G.DEFAULT_WPM;return Math.round(j/60*v)}}class T{convert(j,q){if(j.format===q)return j;return{...j,format:q}}isSupported(j,q){if(j===q)return!0;return{wav:["mp3","ogg","pcm","opus"],mp3:["wav"],ogg:["wav"],pcm:["wav"],opus:["wav"]}[j]?.includes(q)??!1}}class U{generate(j,q="wav",v=44100,A=1){let K=Math.ceil(v*j/1000),O=2,J=K*2*A;return{data:new Uint8Array(J),format:q,sampleRateHz:v,durationMs:j,channels:A}}}export{U as SilenceGenerator,T as FormatConverter,G as DurationEstimator,Q as AudioConcatenator};

package/dist/node/audio/silence-generator.js

@@ -1 +0,0 @@
-class x{generate(b,k="wav",g=44100,j=1){let q=Math.ceil(g*b/1000),y=2,v=q*2*j;return{data:new Uint8Array(v),format:k,sampleRateHz:g,durationMs:b,channels:j}}}export{x as SilenceGenerator};

package/dist/node/conversational/index.js

@@ -1,2 +0,0 @@
-class I{stt;llm;tts;modelSelector;selectionContext;conversationHistory=[];constructor(C,S,V){if("stt"in C)this.stt=C.stt,this.llm=C.llm,this.tts=C.tts,this.modelSelector=C.modelSelector,this.selectionContext=C.selectionContext;else{if(!S||!V)throw Error("ResponseOrchestrator requires llm and tts when constructed with positional arguments");this.stt=C,this.llm=S,this.tts=V}}async*processUserTurn(C,S){let D=(await this.stt.transcribe({audio:C,language:S.language,wordTimestamps:!1})).text;yield{type:"user_speech_ended",transcript:D},yield{type:"transcript",role:"user",text:D,timestamp:Date.now()},this.conversationHistory.push({role:"user",content:D});let E=await this.resolveModel(S.llmModel),F=(await this.llm.chat([{role:"system",content:[{type:"text",text:S.systemPrompt}]},...this.conversationHistory.map((G)=>({role:G.role,content:[{type:"text",text:G.content}]}))],{model:E})).message.content.find((G)=>G.type==="text"),q=F&&F.type==="text"?F.text:"I apologize, I could not generate a response.";this.conversationHistory.push({role:"assistant",content:q}),yield{type:"agent_speech_started",text:q},yield{type:"agent_audio",audio:(await this.tts.synthesize({text:q,voiceId:S.voiceId,language:S.language,format:S.outputFormat})).audio.data},yield{type:"agent_speech_ended"},yield{type:"transcript",role:"agent",text:q,timestamp:Date.now()}}async resolveModel(C){if(C)return C;if(this.modelSelector){let S=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(S)).modelId}return}reset(){this.conversationHistory.length=0}}class H{turns=[];currentTurn=null;sessionStartMs=Date.now();getTranscript(){return[...this.turns]}toText(){return this.turns.map((C)=>`[${C.role}] ${C.text}`).join(`
-`)}getTurnCount(){return this.turns.length}processEvent(C){switch(C.type){case"session_started":this.sessionStartMs=Date.now();break;case"user_speech_started":this.currentTurn={role:"user",startMs:Date.now()-this.sessionStartMs};break;case"user_speech_ended":if(this.currentTurn&&this.currentTurn.role==="user")this.currentTurn.text=C.transcript,this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"agent_speech_started":this.currentTurn={role:"agent",text:C.text,startMs:Date.now()-this.sessionStartMs};break;case"agent_speech_ended":if(this.currentTurn&&this.currentTurn.role==="agent")this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"transcript":break;default:break}}reset(){this.turns.length=0,this.currentTurn=null,this.sessionStartMs=Date.now()}}class J{silenceThresholdMs;energyThreshold;silenceStartMs=null;constructor(C=800,S=0.01){this.silenceThresholdMs=C,this.energyThreshold=S}processChunk(C,S){if(this.calculateEnergy(C)>this.energyThreshold)return this.silenceStartMs=null,!1;if(this.silenceStartMs===null)this.silenceStartMs=S;return S-this.silenceStartMs>=this.silenceThresholdMs}reset(){this.silenceStartMs=null}calculateEnergy(C){if(C.length<2)return 0;let S=0,V=Math.floor(C.length/2);for(let D=0;D<C.length-1;D+=2){let E=C[D]??0,j=C[D+1]??0,q=((E|j<<8)<<16>>16)/32768;S+=q*q}return Math.sqrt(S/V)}}class K{provider;constructor(C){this.provider=C.conversational}async startSession(C){let S=new H,V=await this.provider.startSession({voiceId:C.voiceId,language:C.language,systemPrompt:C.systemPrompt,llmModel:C.llmModel,inputFormat:C.inputFormat,outputFormat:C.outputFormat,turnDetection:C.turnDetection,silenceThresholdMs:C.silenceThresholdMs,maxDurationSeconds:C.maxDurationSeconds}),D={sessionId:"",status:"connecting",currentTurn:"idle",turnCount:0,durationMs:0,transcript:[]},E=this.wrapEvents(V.events,D,S);return{state:D,sendAudio:(j)=>V.sendAudio(j),sendText:(j)=>V.sendText(j),interrupt:()=>V.interrupt(),close:async()=>{let j=await V.close();return D.status="ended",j},events:E}}async*wrapEvents(C,S,V){for await(let D of C){switch(V.processEvent(D),D.type){case"session_started":S.sessionId=D.sessionId,S.status="active";break;case"user_speech_started":S.currentTurn="user";break;case"user_speech_ended":S.currentTurn="idle",S.turnCount+=1;break;case"agent_speech_started":S.currentTurn="agent";break;case"agent_speech_ended":S.currentTurn="idle",S.turnCount+=1;break;case"session_ended":S.status="ended",S.durationMs=D.durationMs;break}S.transcript=V.getTranscript(),yield D}}}export{K as VoiceSessionManager,J as TurnDetector,H as TranscriptBuilder,I as ResponseOrchestrator};

package/dist/node/conversational/response-orchestrator.js

@@ -1 +0,0 @@
-class D{stt;llm;tts;modelSelector;selectionContext;conversationHistory=[];constructor(b,j,k){if("stt"in b)this.stt=b.stt,this.llm=b.llm,this.tts=b.tts,this.modelSelector=b.modelSelector,this.selectionContext=b.selectionContext;else{if(!j||!k)throw Error("ResponseOrchestrator requires llm and tts when constructed with positional arguments");this.stt=b,this.llm=j,this.tts=k}}async*processUserTurn(b,j){let z=(await this.stt.transcribe({audio:b,language:j.language,wordTimestamps:!1})).text;yield{type:"user_speech_ended",transcript:z},yield{type:"transcript",role:"user",text:z,timestamp:Date.now()},this.conversationHistory.push({role:"user",content:z});let C=await this.resolveModel(j.llmModel),B=(await this.llm.chat([{role:"system",content:[{type:"text",text:j.systemPrompt}]},...this.conversationHistory.map((w)=>({role:w.role,content:[{type:"text",text:w.content}]}))],{model:C})).message.content.find((w)=>w.type==="text"),q=B&&B.type==="text"?B.text:"I apologize, I could not generate a response.";this.conversationHistory.push({role:"assistant",content:q}),yield{type:"agent_speech_started",text:q},yield{type:"agent_audio",audio:(await this.tts.synthesize({text:q,voiceId:j.voiceId,language:j.language,format:j.outputFormat})).audio.data},yield{type:"agent_speech_ended"},yield{type:"transcript",role:"agent",text:q,timestamp:Date.now()}}async resolveModel(b){if(b)return b;if(this.modelSelector){let j=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(j)).modelId}return}reset(){this.conversationHistory.length=0}}export{D as ResponseOrchestrator};

package/dist/node/conversational/transcript-builder.js

@@ -1,2 +0,0 @@
-class d{turns=[];currentTurn=null;sessionStartMs=Date.now();getTranscript(){return[...this.turns]}toText(){return this.turns.map((b)=>`[${b.role}] ${b.text}`).join(`
-`)}getTurnCount(){return this.turns.length}processEvent(b){switch(b.type){case"session_started":this.sessionStartMs=Date.now();break;case"user_speech_started":this.currentTurn={role:"user",startMs:Date.now()-this.sessionStartMs};break;case"user_speech_ended":if(this.currentTurn&&this.currentTurn.role==="user")this.currentTurn.text=b.transcript,this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"agent_speech_started":this.currentTurn={role:"agent",text:b.text,startMs:Date.now()-this.sessionStartMs};break;case"agent_speech_ended":if(this.currentTurn&&this.currentTurn.role==="agent")this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"transcript":break;default:break}}reset(){this.turns.length=0,this.currentTurn=null,this.sessionStartMs=Date.now()}}export{d as TranscriptBuilder};

package/dist/node/conversational/turn-detector.js

@@ -1 +0,0 @@
-class E{silenceThresholdMs;energyThreshold;silenceStartMs=null;constructor(f=800,j=0.01){this.silenceThresholdMs=f,this.energyThreshold=j}processChunk(f,j){if(this.calculateEnergy(f)>this.energyThreshold)return this.silenceStartMs=null,!1;if(this.silenceStartMs===null)this.silenceStartMs=j;return j-this.silenceStartMs>=this.silenceThresholdMs}reset(){this.silenceStartMs=null}calculateEnergy(f){if(f.length<2)return 0;let j=0,x=Math.floor(f.length/2);for(let q=0;q<f.length-1;q+=2){let A=f[q]??0,D=f[q+1]??0,B=((A|D<<8)<<16>>16)/32768;j+=B*B}return Math.sqrt(j/x)}}export{E as TurnDetector};

package/dist/node/conversational/voice-session-manager.js

@@ -1,2 +0,0 @@
-class D{turns=[];currentTurn=null;sessionStartMs=Date.now();getTranscript(){return[...this.turns]}toText(){return this.turns.map((j)=>`[${j.role}] ${j.text}`).join(`
-`)}getTurnCount(){return this.turns.length}processEvent(j){switch(j.type){case"session_started":this.sessionStartMs=Date.now();break;case"user_speech_started":this.currentTurn={role:"user",startMs:Date.now()-this.sessionStartMs};break;case"user_speech_ended":if(this.currentTurn&&this.currentTurn.role==="user")this.currentTurn.text=j.transcript,this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"agent_speech_started":this.currentTurn={role:"agent",text:j.text,startMs:Date.now()-this.sessionStartMs};break;case"agent_speech_ended":if(this.currentTurn&&this.currentTurn.role==="agent")this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"transcript":break;default:break}}reset(){this.turns.length=0,this.currentTurn=null,this.sessionStartMs=Date.now()}}class G{provider;constructor(j){this.provider=j.conversational}async startSession(j){let q=new D,A=await this.provider.startSession({voiceId:j.voiceId,language:j.language,systemPrompt:j.systemPrompt,llmModel:j.llmModel,inputFormat:j.inputFormat,outputFormat:j.outputFormat,turnDetection:j.turnDetection,silenceThresholdMs:j.silenceThresholdMs,maxDurationSeconds:j.maxDurationSeconds}),z={sessionId:"",status:"connecting",currentTurn:"idle",turnCount:0,durationMs:0,transcript:[]},F=this.wrapEvents(A.events,z,q);return{state:z,sendAudio:(C)=>A.sendAudio(C),sendText:(C)=>A.sendText(C),interrupt:()=>A.interrupt(),close:async()=>{let C=await A.close();return z.status="ended",C},events:F}}async*wrapEvents(j,q,A){for await(let z of j){switch(A.processEvent(z),z.type){case"session_started":q.sessionId=z.sessionId,q.status="active";break;case"user_speech_started":q.currentTurn="user";break;case"user_speech_ended":q.currentTurn="idle",q.turnCount+=1;break;case"agent_speech_started":q.currentTurn="agent";break;case"agent_speech_ended":q.currentTurn="idle",q.turnCount+=1;break;case"session_ended":q.status="ended",q.durationMs=z.durationMs;break}q.transcript=A.getTranscript(),yield z}}}export{G as VoiceSessionManager};

package/dist/node/docs/conversational.docblock.js

@@ -1 +0,0 @@
-var e=!0;export{e as conversationalDocblock};

package/dist/node/docs/stt.docblock.js

@@ -1 +0,0 @@
-var t=!0;export{t as sttDocblock};

package/dist/node/docs/sync.docblock.js

@@ -1 +0,0 @@
-var e=!0;export{e as syncDocblock};

package/dist/node/docs/tts.docblock.js

@@ -1 +0,0 @@
-var e=!0;export{e as ttsDocblock};

package/dist/node/docs/voice.docblock.js

@@ -1 +0,0 @@
-var i=!0;export{i as voiceDocblock};

package/dist/node/i18n/catalogs/en.js

@@ -1,3 +0,0 @@
-import{defineTranslation as g}from"@contractspec/lib.contracts-spec/translations";var j=g({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"All user-facing, LLM-facing, and developer-facing strings for the voice package",owners:["platform"],stability:"experimental"},locale:"en",fallback:"en",messages:{"prompt.tts.system":{value:`You are a voice narration script writer.
-Analyze the content and produce a narration script with pacing directives.
-Return JSON with segments, each having sceneId, text, and contentType.`,description:"TTS system prompt for LLM-enhanced script generation"},"prompt.pace.sceneMatched":{value:"Match voice pacing to scene durations. Adjust rate and emphasis per segment to fit the video timeline.",description:"Prompt for scene-matched pacing strategy"},"prompt.emphasis.system":{value:"You are a voice director. For each segment, determine the optimal emphasis, tone, and speaking rate.",description:"Emphasis planner LLM system prompt"},"pace.intro.description":{value:"Authoritative opening at a measured pace",description:"Description for intro pacing"},"pace.problem.description":{value:"Urgent emphasis on the challenge",description:"Description for problem pacing"},"pace.solution.description":{value:"Calm, clear delivery of the solution",description:"Description for solution pacing"},"pace.metric.description":{value:"Excited emphasis on key results",description:"Description for metric pacing"},"pace.cta.description":{value:"Authoritative call to action",description:"Description for CTA pacing"},"pace.transition.description":{value:"Quick neutral transition",description:"Description for transition pacing"},"stt.transcribing":{value:"Transcribing audio...",description:"Status message during transcription"},"stt.diarization.speaker":{value:"Speaker {index}",description:"Default speaker label",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Subtitle timestamp format",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Voice session started",description:"Session start notification"},"conv.turn.user":{value:"User is speaking",description:"User turn indicator"},"conv.turn.agent":{value:"Agent is responding",description:"Agent turn indicator"},"conv.session.ended":{value:"Voice session ended. Duration: {durationMs}ms",description:"Session end notification",placeholders:[{name:"durationMs",type:"number"}]}}});export{j as enMessages};

package/dist/node/i18n/catalogs/es.js

@@ -1,3 +0,0 @@
-import{defineTranslation as g}from"@contractspec/lib.contracts-spec/translations";var j=g({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"Spanish translations for the voice package",owners:["platform"],stability:"experimental"},locale:"es",fallback:"en",messages:{"prompt.tts.system":{value:`Eres un escritor de guiones de narracion vocal.
-Analiza el contenido y produce un guion de narracion con directivas de ritmo.
-Devuelve JSON con segmentos, cada uno con sceneId, text y contentType.`,description:"Prompt del sistema TTS para generacion de guiones mejorada con LLM"},"prompt.pace.sceneMatched":{value:"Ajusta el ritmo vocal a la duracion de las escenas. Modifica la velocidad y el enfasis por segmento para coincidir con la linea de tiempo del video.",description:"Prompt para la estrategia de ritmo sincronizado con escenas"},"prompt.emphasis.system":{value:"Eres un director vocal. Para cada segmento, determina el enfasis, tono y velocidad optimos.",description:"Prompt del sistema LLM para el planificador de enfasis"},"pace.intro.description":{value:"Apertura autoritaria a un ritmo medido",description:"Descripcion del ritmo de introduccion"},"pace.problem.description":{value:"Enfasis urgente en el desafio",description:"Descripcion del ritmo de problema"},"pace.solution.description":{value:"Entrega calmada y clara de la solucion",description:"Descripcion del ritmo de solucion"},"pace.metric.description":{value:"Enfasis entusiasta en los resultados clave",description:"Descripcion del ritmo de metrica"},"pace.cta.description":{value:"Llamada a la accion autoritaria",description:"Descripcion del ritmo CTA"},"pace.transition.description":{value:"Transicion rapida y neutral",description:"Descripcion del ritmo de transicion"},"stt.transcribing":{value:"Transcribiendo audio...",description:"Mensaje de estado durante la transcripcion"},"stt.diarization.speaker":{value:"Hablante {index}",description:"Etiqueta de hablante predeterminada",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Formato de marca de tiempo para subtitulos",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Sesion de voz iniciada",description:"Notificacion de inicio de sesion"},"conv.turn.user":{value:"El usuario esta hablando",description:"Indicador de turno del usuario"},"conv.turn.agent":{value:"El agente esta respondiendo",description:"Indicador de turno del agente"},"conv.session.ended":{value:"Sesion de voz finalizada. Duracion: {durationMs}ms",description:"Notificacion de fin de sesion",placeholders:[{name:"durationMs",type:"number"}]}}});export{j as esMessages};

package/dist/node/i18n/catalogs/fr.js

@@ -1,3 +0,0 @@
-import{defineTranslation as g}from"@contractspec/lib.contracts-spec/translations";var j=g({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"French translations for the voice package",owners:["platform"],stability:"experimental"},locale:"fr",fallback:"en",messages:{"prompt.tts.system":{value:`Vous etes un redacteur de scripts de narration vocale.
-Analysez le contenu et produisez un script de narration avec des directives de rythme.
-Retournez du JSON avec des segments, chacun ayant sceneId, text et contentType.`,description:"Prompt systeme TTS pour la generation de scripts amelioree par LLM"},"prompt.pace.sceneMatched":{value:"Adaptez le rythme vocal aux durees des scenes. Ajustez le debit et l'emphase par segment pour correspondre a la timeline video.",description:"Prompt pour la strategie de rythme alignee sur les scenes"},"prompt.emphasis.system":{value:"Vous etes un directeur vocal. Pour chaque segment, determinez l'emphase, le ton et le debit optimaux.",description:"Prompt systeme LLM pour le planificateur d'emphase"},"pace.intro.description":{value:"Ouverture autoritaire a un rythme mesure",description:"Description du rythme d'introduction"},"pace.problem.description":{value:"Emphase urgente sur le defi",description:"Description du rythme de probleme"},"pace.solution.description":{value:"Livraison calme et claire de la solution",description:"Description du rythme de solution"},"pace.metric.description":{value:"Emphase enthousiaste sur les resultats cles",description:"Description du rythme de metrique"},"pace.cta.description":{value:"Appel a l'action autoritaire",description:"Description du rythme CTA"},"pace.transition.description":{value:"Transition rapide et neutre",description:"Description du rythme de transition"},"stt.transcribing":{value:"Transcription en cours...",description:"Message de statut pendant la transcription"},"stt.diarization.speaker":{value:"Locuteur {index}",description:"Label de locuteur par defaut",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Format de timestamp pour les sous-titres",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Session vocale demarree",description:"Notification de debut de session"},"conv.turn.user":{value:"L'utilisateur parle",description:"Indicateur de tour de l'utilisateur"},"conv.turn.agent":{value:"L'agent repond",description:"Indicateur de tour de l'agent"},"conv.session.ended":{value:"Session vocale terminee. Duree : {durationMs}ms",description:"Notification de fin de session",placeholders:[{name:"durationMs",type:"number"}]}}});export{j as frMessages};

package/dist/node/i18n/catalogs/index.js

@@ -1,7 +0,0 @@
-import{defineTranslation as g}from"@contractspec/lib.contracts-spec/translations";var m=g({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"All user-facing, LLM-facing, and developer-facing strings for the voice package",owners:["platform"],stability:"experimental"},locale:"en",fallback:"en",messages:{"prompt.tts.system":{value:`You are a voice narration script writer.
-Analyze the content and produce a narration script with pacing directives.
-Return JSON with segments, each having sceneId, text, and contentType.`,description:"TTS system prompt for LLM-enhanced script generation"},"prompt.pace.sceneMatched":{value:"Match voice pacing to scene durations. Adjust rate and emphasis per segment to fit the video timeline.",description:"Prompt for scene-matched pacing strategy"},"prompt.emphasis.system":{value:"You are a voice director. For each segment, determine the optimal emphasis, tone, and speaking rate.",description:"Emphasis planner LLM system prompt"},"pace.intro.description":{value:"Authoritative opening at a measured pace",description:"Description for intro pacing"},"pace.problem.description":{value:"Urgent emphasis on the challenge",description:"Description for problem pacing"},"pace.solution.description":{value:"Calm, clear delivery of the solution",description:"Description for solution pacing"},"pace.metric.description":{value:"Excited emphasis on key results",description:"Description for metric pacing"},"pace.cta.description":{value:"Authoritative call to action",description:"Description for CTA pacing"},"pace.transition.description":{value:"Quick neutral transition",description:"Description for transition pacing"},"stt.transcribing":{value:"Transcribing audio...",description:"Status message during transcription"},"stt.diarization.speaker":{value:"Speaker {index}",description:"Default speaker label",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Subtitle timestamp format",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Voice session started",description:"Session start notification"},"conv.turn.user":{value:"User is speaking",description:"User turn indicator"},"conv.turn.agent":{value:"Agent is responding",description:"Agent turn indicator"},"conv.session.ended":{value:"Voice session ended. Duration: {durationMs}ms",description:"Session end notification",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as x}from"@contractspec/lib.contracts-spec/translations";var M=x({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"Spanish translations for the voice package",owners:["platform"],stability:"experimental"},locale:"es",fallback:"en",messages:{"prompt.tts.system":{value:`Eres un escritor de guiones de narracion vocal.
-Analiza el contenido y produce un guion de narracion con directivas de ritmo.
-Devuelve JSON con segmentos, cada uno con sceneId, text y contentType.`,description:"Prompt del sistema TTS para generacion de guiones mejorada con LLM"},"prompt.pace.sceneMatched":{value:"Ajusta el ritmo vocal a la duracion de las escenas. Modifica la velocidad y el enfasis por segmento para coincidir con la linea de tiempo del video.",description:"Prompt para la estrategia de ritmo sincronizado con escenas"},"prompt.emphasis.system":{value:"Eres un director vocal. Para cada segmento, determina el enfasis, tono y velocidad optimos.",description:"Prompt del sistema LLM para el planificador de enfasis"},"pace.intro.description":{value:"Apertura autoritaria a un ritmo medido",description:"Descripcion del ritmo de introduccion"},"pace.problem.description":{value:"Enfasis urgente en el desafio",description:"Descripcion del ritmo de problema"},"pace.solution.description":{value:"Entrega calmada y clara de la solucion",description:"Descripcion del ritmo de solucion"},"pace.metric.description":{value:"Enfasis entusiasta en los resultados clave",description:"Descripcion del ritmo de metrica"},"pace.cta.description":{value:"Llamada a la accion autoritaria",description:"Descripcion del ritmo CTA"},"pace.transition.description":{value:"Transicion rapida y neutral",description:"Descripcion del ritmo de transicion"},"stt.transcribing":{value:"Transcribiendo audio...",description:"Mensaje de estado durante la transcripcion"},"stt.diarization.speaker":{value:"Hablante {index}",description:"Etiqueta de hablante predeterminada",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Formato de marca de tiempo para subtitulos",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Sesion de voz iniciada",description:"Notificacion de inicio de sesion"},"conv.turn.user":{value:"El usuario esta hablando",description:"Indicador de turno del usuario"},"conv.turn.agent":{value:"El agente esta respondiendo",description:"Indicador de turno del agente"},"conv.session.ended":{value:"Sesion de voz finalizada. Duracion: {durationMs}ms",description:"Notificacion de fin de sesion",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as f}from"@contractspec/lib.contracts-spec/translations";var h=f({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"French translations for the voice package",owners:["platform"],stability:"experimental"},locale:"fr",fallback:"en",messages:{"prompt.tts.system":{value:`Vous etes un redacteur de scripts de narration vocale.
-Analysez le contenu et produisez un script de narration avec des directives de rythme.
-Retournez du JSON avec des segments, chacun ayant sceneId, text et contentType.`,description:"Prompt systeme TTS pour la generation de scripts amelioree par LLM"},"prompt.pace.sceneMatched":{value:"Adaptez le rythme vocal aux durees des scenes. Ajustez le debit et l'emphase par segment pour correspondre a la timeline video.",description:"Prompt pour la strategie de rythme alignee sur les scenes"},"prompt.emphasis.system":{value:"Vous etes un directeur vocal. Pour chaque segment, determinez l'emphase, le ton et le debit optimaux.",description:"Prompt systeme LLM pour le planificateur d'emphase"},"pace.intro.description":{value:"Ouverture autoritaire a un rythme mesure",description:"Description du rythme d'introduction"},"pace.problem.description":{value:"Emphase urgente sur le defi",description:"Description du rythme de probleme"},"pace.solution.description":{value:"Livraison calme et claire de la solution",description:"Description du rythme de solution"},"pace.metric.description":{value:"Emphase enthousiaste sur les resultats cles",description:"Description du rythme de metrique"},"pace.cta.description":{value:"Appel a l'action autoritaire",description:"Description du rythme CTA"},"pace.transition.description":{value:"Transition rapide et neutre",description:"Description du rythme de transition"},"stt.transcribing":{value:"Transcription en cours...",description:"Message de statut pendant la transcription"},"stt.diarization.speaker":{value:"Locuteur {index}",description:"Label de locuteur par defaut",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Format de timestamp pour les sous-titres",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Session vocale demarree",description:"Notification de debut de session"},"conv.turn.user":{value:"L'utilisateur parle",description:"Indicateur de tour de l'utilisateur"},"conv.turn.agent":{value:"L'agent repond",description:"Indicateur de tour de l'agent"},"conv.session.ended":{value:"Session vocale terminee. Duree : {durationMs}ms",description:"Notification de fin de session",placeholders:[{name:"durationMs",type:"number"}]}}});export{h as frMessages,M as esMessages,m as enMessages};

package/dist/node/i18n/index.js

@@ -1,7 +0,0 @@
-import{defineTranslation as S}from"@contractspec/lib.contracts-spec/translations";var e=S({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"All user-facing, LLM-facing, and developer-facing strings for the voice package",owners:["platform"],stability:"experimental"},locale:"en",fallback:"en",messages:{"prompt.tts.system":{value:`You are a voice narration script writer.
-Analyze the content and produce a narration script with pacing directives.
-Return JSON with segments, each having sceneId, text, and contentType.`,description:"TTS system prompt for LLM-enhanced script generation"},"prompt.pace.sceneMatched":{value:"Match voice pacing to scene durations. Adjust rate and emphasis per segment to fit the video timeline.",description:"Prompt for scene-matched pacing strategy"},"prompt.emphasis.system":{value:"You are a voice director. For each segment, determine the optimal emphasis, tone, and speaking rate.",description:"Emphasis planner LLM system prompt"},"pace.intro.description":{value:"Authoritative opening at a measured pace",description:"Description for intro pacing"},"pace.problem.description":{value:"Urgent emphasis on the challenge",description:"Description for problem pacing"},"pace.solution.description":{value:"Calm, clear delivery of the solution",description:"Description for solution pacing"},"pace.metric.description":{value:"Excited emphasis on key results",description:"Description for metric pacing"},"pace.cta.description":{value:"Authoritative call to action",description:"Description for CTA pacing"},"pace.transition.description":{value:"Quick neutral transition",description:"Description for transition pacing"},"stt.transcribing":{value:"Transcribing audio...",description:"Status message during transcription"},"stt.diarization.speaker":{value:"Speaker {index}",description:"Default speaker label",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Subtitle timestamp format",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Voice session started",description:"Session start notification"},"conv.turn.user":{value:"User is speaking",description:"User turn indicator"},"conv.turn.agent":{value:"Agent is responding",description:"Agent turn indicator"},"conv.session.ended":{value:"Voice session ended. Duration: {durationMs}ms",description:"Session end notification",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as f}from"@contractspec/lib.contracts-spec/translations";var o=f({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"Spanish translations for the voice package",owners:["platform"],stability:"experimental"},locale:"es",fallback:"en",messages:{"prompt.tts.system":{value:`Eres un escritor de guiones de narracion vocal.
-Analiza el contenido y produce un guion de narracion con directivas de ritmo.
-Devuelve JSON con segmentos, cada uno con sceneId, text y contentType.`,description:"Prompt del sistema TTS para generacion de guiones mejorada con LLM"},"prompt.pace.sceneMatched":{value:"Ajusta el ritmo vocal a la duracion de las escenas. Modifica la velocidad y el enfasis por segmento para coincidir con la linea de tiempo del video.",description:"Prompt para la estrategia de ritmo sincronizado con escenas"},"prompt.emphasis.system":{value:"Eres un director vocal. Para cada segmento, determina el enfasis, tono y velocidad optimos.",description:"Prompt del sistema LLM para el planificador de enfasis"},"pace.intro.description":{value:"Apertura autoritaria a un ritmo medido",description:"Descripcion del ritmo de introduccion"},"pace.problem.description":{value:"Enfasis urgente en el desafio",description:"Descripcion del ritmo de problema"},"pace.solution.description":{value:"Entrega calmada y clara de la solucion",description:"Descripcion del ritmo de solucion"},"pace.metric.description":{value:"Enfasis entusiasta en los resultados clave",description:"Descripcion del ritmo de metrica"},"pace.cta.description":{value:"Llamada a la accion autoritaria",description:"Descripcion del ritmo CTA"},"pace.transition.description":{value:"Transicion rapida y neutral",description:"Descripcion del ritmo de transicion"},"stt.transcribing":{value:"Transcribiendo audio...",description:"Mensaje de estado durante la transcripcion"},"stt.diarization.speaker":{value:"Hablante {index}",description:"Etiqueta de hablante predeterminada",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Formato de marca de tiempo para subtitulos",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Sesion de voz iniciada",description:"Notificacion de inicio de sesion"},"conv.turn.user":{value:"El usuario esta hablando",description:"Indicador de turno del usuario"},"conv.turn.agent":{value:"El agente esta respondiendo",description:"Indicador de turno del agente"},"conv.session.ended":{value:"Sesion de voz finalizada. Duracion: {durationMs}ms",description:"Notificacion de fin de sesion",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as E}from"@contractspec/lib.contracts-spec/translations";var m=E({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"French translations for the voice package",owners:["platform"],stability:"experimental"},locale:"fr",fallback:"en",messages:{"prompt.tts.system":{value:`Vous etes un redacteur de scripts de narration vocale.
-Analysez le contenu et produisez un script de narration avec des directives de rythme.
-Retournez du JSON avec des segments, chacun ayant sceneId, text et contentType.`,description:"Prompt systeme TTS pour la generation de scripts amelioree par LLM"},"prompt.pace.sceneMatched":{value:"Adaptez le rythme vocal aux durees des scenes. Ajustez le debit et l'emphase par segment pour correspondre a la timeline video.",description:"Prompt pour la strategie de rythme alignee sur les scenes"},"prompt.emphasis.system":{value:"Vous etes un directeur vocal. Pour chaque segment, determinez l'emphase, le ton et le debit optimaux.",description:"Prompt systeme LLM pour le planificateur d'emphase"},"pace.intro.description":{value:"Ouverture autoritaire a un rythme mesure",description:"Description du rythme d'introduction"},"pace.problem.description":{value:"Emphase urgente sur le defi",description:"Description du rythme de probleme"},"pace.solution.description":{value:"Livraison calme et claire de la solution",description:"Description du rythme de solution"},"pace.metric.description":{value:"Emphase enthousiaste sur les resultats cles",description:"Description du rythme de metrique"},"pace.cta.description":{value:"Appel a l'action autoritaire",description:"Description du rythme CTA"},"pace.transition.description":{value:"Transition rapide et neutre",description:"Description du rythme de transition"},"stt.transcribing":{value:"Transcription en cours...",description:"Message de statut pendant la transcription"},"stt.diarization.speaker":{value:"Locuteur {index}",description:"Label de locuteur par defaut",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Format de timestamp pour les sous-titres",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Session vocale demarree",description:"Notification de debut de session"},"conv.turn.user":{value:"L'utilisateur parle",description:"Indicateur de tour de l'utilisateur"},"conv.turn.agent":{value:"L'agent repond",description:"Indicateur de tour de l'agent"},"conv.session.ended":{value:"Session vocale terminee. Duree : {durationMs}ms",description:"Notification de fin de session",placeholders:[{name:"durationMs",type:"number"}]}}});var x={"prompt.tts.system":"prompt.tts.system","prompt.pace.sceneMatched":"prompt.pace.sceneMatched","prompt.emphasis.system":"prompt.emphasis.system"},p={"pace.intro.description":"pace.intro.description","pace.problem.description":"pace.problem.description","pace.solution.description":"pace.solution.description","pace.metric.description":"pace.metric.description","pace.cta.description":"pace.cta.description","pace.transition.description":"pace.transition.description"},L={"stt.transcribing":"stt.transcribing","stt.diarization.speaker":"stt.diarization.speaker","stt.subtitle.timestamp":"stt.subtitle.timestamp"},I={"conv.session.started":"conv.session.started","conv.turn.user":"conv.turn.user","conv.turn.agent":"conv.turn.agent","conv.session.ended":"conv.session.ended"},V={...x,...p,...L,...I};import{DEFAULT_LOCALE as u,isSupportedLocale as v,resolveLocale as A,SUPPORTED_LOCALES as C}from"@contractspec/lib.contracts-spec/translations";import{createI18nFactory as D}from"@contractspec/lib.contracts-spec/translations";var r=D({specKey:"voice.messages",catalogs:[e,m,o]}),O=r.create,_=r.getDefault,g=r.resetRegistry;export{A as resolveLocale,g as resetI18nRegistry,v as isSupportedLocale,_ as getDefaultI18n,m as frMessages,o as esMessages,e as enMessages,O as createVoiceI18n,x as TTS_PROMPT_KEYS,p as TTS_PACE_KEYS,C as SUPPORTED_LOCALES,L as STT_KEYS,V as I18N_KEYS,u as DEFAULT_LOCALE,I as CONVERSATIONAL_KEYS};

package/dist/node/i18n/keys.js

@@ -1 +0,0 @@
-var b={"prompt.tts.system":"prompt.tts.system","prompt.pace.sceneMatched":"prompt.pace.sceneMatched","prompt.emphasis.system":"prompt.emphasis.system"},g={"pace.intro.description":"pace.intro.description","pace.problem.description":"pace.problem.description","pace.solution.description":"pace.solution.description","pace.metric.description":"pace.metric.description","pace.cta.description":"pace.cta.description","pace.transition.description":"pace.transition.description"},h={"stt.transcribing":"stt.transcribing","stt.diarization.speaker":"stt.diarization.speaker","stt.subtitle.timestamp":"stt.subtitle.timestamp"},j={"conv.session.started":"conv.session.started","conv.turn.user":"conv.turn.user","conv.turn.agent":"conv.turn.agent","conv.session.ended":"conv.session.ended"},m={...b,...g,...h,...j};export{b as TTS_PROMPT_KEYS,g as TTS_PACE_KEYS,h as STT_KEYS,m as I18N_KEYS,j as CONVERSATIONAL_KEYS};

package/dist/node/i18n/locale.js

@@ -1 +0,0 @@
-import{DEFAULT_LOCALE as g,isSupportedLocale as h,resolveLocale as j,SUPPORTED_LOCALES as k}from"@contractspec/lib.contracts-spec/translations";export{j as resolveLocale,h as isSupportedLocale,k as SUPPORTED_LOCALES,g as DEFAULT_LOCALE};

package/dist/node/i18n/messages.js

@@ -1,7 +0,0 @@
-import{defineTranslation as q}from"@contractspec/lib.contracts-spec/translations";var j=q({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"All user-facing, LLM-facing, and developer-facing strings for the voice package",owners:["platform"],stability:"experimental"},locale:"en",fallback:"en",messages:{"prompt.tts.system":{value:`You are a voice narration script writer.
-Analyze the content and produce a narration script with pacing directives.
-Return JSON with segments, each having sceneId, text, and contentType.`,description:"TTS system prompt for LLM-enhanced script generation"},"prompt.pace.sceneMatched":{value:"Match voice pacing to scene durations. Adjust rate and emphasis per segment to fit the video timeline.",description:"Prompt for scene-matched pacing strategy"},"prompt.emphasis.system":{value:"You are a voice director. For each segment, determine the optimal emphasis, tone, and speaking rate.",description:"Emphasis planner LLM system prompt"},"pace.intro.description":{value:"Authoritative opening at a measured pace",description:"Description for intro pacing"},"pace.problem.description":{value:"Urgent emphasis on the challenge",description:"Description for problem pacing"},"pace.solution.description":{value:"Calm, clear delivery of the solution",description:"Description for solution pacing"},"pace.metric.description":{value:"Excited emphasis on key results",description:"Description for metric pacing"},"pace.cta.description":{value:"Authoritative call to action",description:"Description for CTA pacing"},"pace.transition.description":{value:"Quick neutral transition",description:"Description for transition pacing"},"stt.transcribing":{value:"Transcribing audio...",description:"Status message during transcription"},"stt.diarization.speaker":{value:"Speaker {index}",description:"Default speaker label",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Subtitle timestamp format",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Voice session started",description:"Session start notification"},"conv.turn.user":{value:"User is speaking",description:"User turn indicator"},"conv.turn.agent":{value:"Agent is responding",description:"Agent turn indicator"},"conv.session.ended":{value:"Voice session ended. Duration: {durationMs}ms",description:"Session end notification",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as u}from"@contractspec/lib.contracts-spec/translations";var k=u({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"Spanish translations for the voice package",owners:["platform"],stability:"experimental"},locale:"es",fallback:"en",messages:{"prompt.tts.system":{value:`Eres un escritor de guiones de narracion vocal.
-Analiza el contenido y produce un guion de narracion con directivas de ritmo.
-Devuelve JSON con segmentos, cada uno con sceneId, text y contentType.`,description:"Prompt del sistema TTS para generacion de guiones mejorada con LLM"},"prompt.pace.sceneMatched":{value:"Ajusta el ritmo vocal a la duracion de las escenas. Modifica la velocidad y el enfasis por segmento para coincidir con la linea de tiempo del video.",description:"Prompt para la estrategia de ritmo sincronizado con escenas"},"prompt.emphasis.system":{value:"Eres un director vocal. Para cada segmento, determina el enfasis, tono y velocidad optimos.",description:"Prompt del sistema LLM para el planificador de enfasis"},"pace.intro.description":{value:"Apertura autoritaria a un ritmo medido",description:"Descripcion del ritmo de introduccion"},"pace.problem.description":{value:"Enfasis urgente en el desafio",description:"Descripcion del ritmo de problema"},"pace.solution.description":{value:"Entrega calmada y clara de la solucion",description:"Descripcion del ritmo de solucion"},"pace.metric.description":{value:"Enfasis entusiasta en los resultados clave",description:"Descripcion del ritmo de metrica"},"pace.cta.description":{value:"Llamada a la accion autoritaria",description:"Descripcion del ritmo CTA"},"pace.transition.description":{value:"Transicion rapida y neutral",description:"Descripcion del ritmo de transicion"},"stt.transcribing":{value:"Transcribiendo audio...",description:"Mensaje de estado durante la transcripcion"},"stt.diarization.speaker":{value:"Hablante {index}",description:"Etiqueta de hablante predeterminada",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Formato de marca de tiempo para subtitulos",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Sesion de voz iniciada",description:"Notificacion de inicio de sesion"},"conv.turn.user":{value:"El usuario esta hablando",description:"Indicador de turno del usuario"},"conv.turn.agent":{value:"El agente esta respondiendo",description:"Indicador de turno del agente"},"conv.session.ended":{value:"Sesion de voz finalizada. Duracion: {durationMs}ms",description:"Notificacion de fin de sesion",placeholders:[{name:"durationMs",type:"number"}]}}});import{defineTranslation as v}from"@contractspec/lib.contracts-spec/translations";var m=v({meta:{key:"voice.messages",version:"1.0.0",domain:"voice",description:"French translations for the voice package",owners:["platform"],stability:"experimental"},locale:"fr",fallback:"en",messages:{"prompt.tts.system":{value:`Vous etes un redacteur de scripts de narration vocale.
-Analysez le contenu et produisez un script de narration avec des directives de rythme.
-Retournez du JSON avec des segments, chacun ayant sceneId, text et contentType.`,description:"Prompt systeme TTS pour la generation de scripts amelioree par LLM"},"prompt.pace.sceneMatched":{value:"Adaptez le rythme vocal aux durees des scenes. Ajustez le debit et l'emphase par segment pour correspondre a la timeline video.",description:"Prompt pour la strategie de rythme alignee sur les scenes"},"prompt.emphasis.system":{value:"Vous etes un directeur vocal. Pour chaque segment, determinez l'emphase, le ton et le debit optimaux.",description:"Prompt systeme LLM pour le planificateur d'emphase"},"pace.intro.description":{value:"Ouverture autoritaire a un rythme mesure",description:"Description du rythme d'introduction"},"pace.problem.description":{value:"Emphase urgente sur le defi",description:"Description du rythme de probleme"},"pace.solution.description":{value:"Livraison calme et claire de la solution",description:"Description du rythme de solution"},"pace.metric.description":{value:"Emphase enthousiaste sur les resultats cles",description:"Description du rythme de metrique"},"pace.cta.description":{value:"Appel a l'action autoritaire",description:"Description du rythme CTA"},"pace.transition.description":{value:"Transition rapide et neutre",description:"Description du rythme de transition"},"stt.transcribing":{value:"Transcription en cours...",description:"Message de statut pendant la transcription"},"stt.diarization.speaker":{value:"Locuteur {index}",description:"Label de locuteur par defaut",placeholders:[{name:"index",type:"number"}]},"stt.subtitle.timestamp":{value:"{start} --> {end}",description:"Format de timestamp pour les sous-titres",placeholders:[{name:"start",type:"string"},{name:"end",type:"string"}]},"conv.session.started":{value:"Session vocale demarree",description:"Notification de debut de session"},"conv.turn.user":{value:"L'utilisateur parle",description:"Indicateur de tour de l'utilisateur"},"conv.turn.agent":{value:"L'agent repond",description:"Indicateur de tour de l'agent"},"conv.session.ended":{value:"Session vocale terminee. Duree : {durationMs}ms",description:"Notification de fin de session",placeholders:[{name:"durationMs",type:"number"}]}}});import{createI18nFactory as w}from"@contractspec/lib.contracts-spec/translations";var h=w({specKey:"voice.messages",catalogs:[j,m,k]}),K=h.create,L=h.getDefault,N=h.resetRegistry;export{N as resetI18nRegistry,L as getDefaultI18n,K as createVoiceI18n};

package/dist/node/index.js

@@ -1,13 +0,0 @@
-class A{concatenate(q){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[J]=q;if(!J)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(q.length===1)return{...J};let{format:U,sampleRateHz:K}=J,X=J.channels??1;for(let V of q){if(V.format!==U)throw Error(`Format mismatch: expected ${U}, got ${V.format}`);if(V.sampleRateHz!==K)throw Error(`Sample rate mismatch: expected ${K}, got ${V.sampleRateHz}`)}let Q=q.reduce((V,L)=>V+L.data.length,0),Z=new Uint8Array(Q),$=0;for(let V of q)Z.set(V.data,$),$+=V.data.length;let Y=q.reduce((V,L)=>V+(L.durationMs??0),0);return{data:Z,format:U,sampleRateHz:K,durationMs:Y,channels:X}}}class G{static DEFAULT_WPM=150;estimateSeconds(q,J){let U=J??G.DEFAULT_WPM,K=q.split(/\s+/).filter(Boolean).length;return Math.ceil(K/U*60)}estimateMs(q,J){let U=J??G.DEFAULT_WPM,K=q.split(/\s+/).filter(Boolean).length;return Math.ceil(K/U*60*1000)}estimateWordCount(q,J){let U=J??G.DEFAULT_WPM;return Math.round(q/60*U)}}class P{convert(q,J){if(q.format===J)return q;return{...q,format:J}}isSupported(q,J){if(q===J)return!0;return{wav:["mp3","ogg","pcm","opus"],mp3:["wav"],ogg:["wav"],pcm:["wav"],opus:["wav"]}[q]?.includes(J)??!1}}class E{generate(q,J="wav",U=44100,K=1){let X=Math.ceil(U*q/1000),Q=2,Z=X*2*K;return{data:new Uint8Array(Z),format:J,sampleRateHz:U,durationMs:q,channels:K}}}class O{stt;llm;tts;modelSelector;selectionContext;conversationHistory=[];constructor(q,J,U){if("stt"in q)this.stt=q.stt,this.llm=q.llm,this.tts=q.tts,this.modelSelector=q.modelSelector,this.selectionContext=q.selectionContext;else{if(!J||!U)throw Error("ResponseOrchestrator requires llm and tts when constructed with positional arguments");this.stt=q,this.llm=J,this.tts=U}}async*processUserTurn(q,J){let K=(await this.stt.transcribe({audio:q,language:J.language,wordTimestamps:!1})).text;yield{type:"user_speech_ended",transcript:K},yield{type:"transcript",role:"user",text:K,timestamp:Date.now()},this.conversationHistory.push({role:"user",content:K});let X=await this.resolveModel(J.llmModel),Z=(await this.llm.chat([{role:"system",content:[{type:"text",text:J.systemPrompt}]},...this.conversationHistory.map((V)=>({role:V.role,content:[{type:"text",text:V.content}]}))],{model:X})).message.content.find((V)=>V.type==="text"),$=Z&&Z.type==="text"?Z.text:"I apologize, I could not generate a response.";this.conversationHistory.push({role:"assistant",content:$}),yield{type:"agent_speech_started",text:$},yield{type:"agent_audio",audio:(await this.tts.synthesize({text:$,voiceId:J.voiceId,language:J.language,format:J.outputFormat})).audio.data},yield{type:"agent_speech_ended"},yield{type:"transcript",role:"agent",text:$,timestamp:Date.now()}}async resolveModel(q){if(q)return q;if(this.modelSelector){let J=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(J)).modelId}return}reset(){this.conversationHistory.length=0}}class I{turns=[];currentTurn=null;sessionStartMs=Date.now();getTranscript(){return[...this.turns]}toText(){return this.turns.map((q)=>`[${q.role}] ${q.text}`).join(`
-`)}getTurnCount(){return this.turns.length}processEvent(q){switch(q.type){case"session_started":this.sessionStartMs=Date.now();break;case"user_speech_started":this.currentTurn={role:"user",startMs:Date.now()-this.sessionStartMs};break;case"user_speech_ended":if(this.currentTurn&&this.currentTurn.role==="user")this.currentTurn.text=q.transcript,this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"agent_speech_started":this.currentTurn={role:"agent",text:q.text,startMs:Date.now()-this.sessionStartMs};break;case"agent_speech_ended":if(this.currentTurn&&this.currentTurn.role==="agent")this.currentTurn.endMs=Date.now()-this.sessionStartMs,this.turns.push(this.currentTurn),this.currentTurn=null;break;case"transcript":break;default:break}}reset(){this.turns.length=0,this.currentTurn=null,this.sessionStartMs=Date.now()}}class k{silenceThresholdMs;energyThreshold;silenceStartMs=null;constructor(q=800,J=0.01){this.silenceThresholdMs=q,this.energyThreshold=J}processChunk(q,J){if(this.calculateEnergy(q)>this.energyThreshold)return this.silenceStartMs=null,!1;if(this.silenceStartMs===null)this.silenceStartMs=J;return J-this.silenceStartMs>=this.silenceThresholdMs}reset(){this.silenceStartMs=null}calculateEnergy(q){if(q.length<2)return 0;let J=0,U=Math.floor(q.length/2);for(let K=0;K<q.length-1;K+=2){let X=q[K]??0,Q=q[K+1]??0,$=((X|Q<<8)<<16>>16)/32768;J+=$*$}return Math.sqrt(J/U)}}class D{provider;constructor(q){this.provider=q.conversational}async startSession(q){let J=new I,U=await this.provider.startSession({voiceId:q.voiceId,language:q.language,systemPrompt:q.systemPrompt,llmModel:q.llmModel,inputFormat:q.inputFormat,outputFormat:q.outputFormat,turnDetection:q.turnDetection,silenceThresholdMs:q.silenceThresholdMs,maxDurationSeconds:q.maxDurationSeconds}),K={sessionId:"",status:"connecting",currentTurn:"idle",turnCount:0,durationMs:0,transcript:[]},X=this.wrapEvents(U.events,K,J);return{state:K,sendAudio:(Q)=>U.sendAudio(Q),sendText:(Q)=>U.sendText(Q),interrupt:()=>U.interrupt(),close:async()=>{let Q=await U.close();return K.status="ended",Q},events:X}}async*wrapEvents(q,J,U){for await(let K of q){switch(U.processEvent(K),K.type){case"session_started":J.sessionId=K.sessionId,J.status="active";break;case"user_speech_started":J.currentTurn="user";break;case"user_speech_ended":J.currentTurn="idle",J.turnCount+=1;break;case"agent_speech_started":J.currentTurn="agent";break;case"agent_speech_ended":J.currentTurn="idle",J.turnCount+=1;break;case"session_ended":J.status="ended",J.durationMs=K.durationMs;break}J.transcript=U.getTranscript(),yield K}}}class T{map(q,J="Speaker"){let U=[],K=new Map;for(let Z of q)if(Z.speakerId&&!U.includes(Z.speakerId))U.push(Z.speakerId),K.set(Z.speakerId,{segmentCount:0,totalSpeakingMs:0});let X=q.map((Z)=>{if(!Z.speakerId)return Z;let $=U.indexOf(Z.speakerId),Y=`${J} ${$+1}`,V=K.get(Z.speakerId);if(!V)return{...Z,speakerLabel:Y};return V.segmentCount+=1,V.totalSpeakingMs+=Z.endMs-Z.startMs,{...Z,speakerLabel:Y}}),Q=U.map((Z,$)=>{let Y=K.get(Z);return{id:Z,label:`${J} ${$+1}`,segmentCount:Y?.segmentCount??0,totalSpeakingMs:Y?.totalSpeakingMs??0}});return{segments:X,speakers:Q}}}class C{static DEFAULT_MAX_CHUNK_MS=300000;split(q,J=C.DEFAULT_MAX_CHUNK_MS){let U=q.durationMs??this.estimateDurationMs(q);if(U<=J)return[q];let K=[],X=q.data.length/Math.max(U,1),Q=0;while(Q<U){let Z=Math.min(J,U-Q),$=Math.floor(Q*X),Y=Math.floor((Q+Z)*X);K.push({data:q.data.slice($,Y),format:q.format,sampleRateHz:q.sampleRateHz,durationMs:Z,channels:q.channels}),Q+=Z}return K}estimateDurationMs(q){let U=q.channels??1,K=q.data.length/(2*U);return Math.ceil(K/q.sampleRateHz*1000)}}class N{toSRT(q){return q.map((J,U)=>{let K=this.formatTimeSRT(J.startMs),X=this.formatTimeSRT(J.endMs),Q=J.speakerLabel?`[${J.speakerLabel}] `:"";return`${U+1}
-${K} --> ${X}
-${Q}${J.text}`}).join(`
-
-`)}toVTT(q){return`WEBVTT
-
-`+q.map((K,X)=>{let Q=this.formatTimeVTT(K.startMs),Z=this.formatTimeVTT(K.endMs),$=K.speakerLabel?`<v ${K.speakerLabel}>`:"";return`${X+1}
-${Q} --> ${Z}
-${$}${K.text}`}).join(`
-
-`)}formatTimeSRT(q){let J=Math.floor(q/3600000),U=Math.floor(q%3600000/60000),K=Math.floor(q%60000/1000),X=q%1000;return`${this.pad(J,2)}:${this.pad(U,2)}:${this.pad(K,2)},${this.pad(X,3)}`}formatTimeVTT(q){let J=Math.floor(q/3600000),U=Math.floor(q%3600000/60000),K=Math.floor(q%60000/1000),X=q%1000;return`${this.pad(J,2)}:${this.pad(U,2)}:${this.pad(K,2)}.${this.pad(X,3)}`}pad(q,J){return q.toString().padStart(J,"0")}}class x{stt;segmentSplitter=new C;diarizationMapper=new T;subtitleFormatter=new N;constructor(q){this.stt=q.stt}async transcribe(q){let J=b(),U=this.segmentSplitter.split(q.audio),K=[],X="",Q=0,Z=0;for(let W of U){let R=await this.stt.transcribe({audio:W,language:q.language,diarize:q.diarize,speakerCount:q.speakerCount,wordTimestamps:!0,vocabularyHints:q.vocabularyHints}),h=R.segments.map((H)=>({text:H.text,startMs:H.startMs+Z,endMs:H.endMs+Z,speakerId:H.speakerId,speakerName:H.speakerName,confidence:H.confidence}));K.push(...h),X+=(X?" ":"")+R.text,Q+=R.durationMs,Z+=W.durationMs??0}let $=K,Y;if(q.diarize){let W=this.diarizationMapper.map(K);$=W.segments,Y=W.speakers}let V={text:X,segments:$,language:q.language??"en",durationMs:Q},L,_=q.subtitleFormat??"none";if(_==="srt")L=this.subtitleFormatter.toSRT($);else if(_==="vtt")L=this.subtitleFormatter.toVTT($);return{id:J,transcript:V,subtitles:L,speakers:Y}}async*transcribeStream(q,J){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let U=this.stt.transcribeStream(q,{language:J?.language,diarize:J?.diarize,speakerCount:J?.speakerCount,wordTimestamps:!0,vocabularyHints:J?.vocabularyHints});for await(let K of U)yield{text:K.text,startMs:K.startMs,endMs:K.endMs,speakerId:K.speakerId,speakerLabel:K.speakerName,confidence:K.confidence}}}function b(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`stt_${q}_${J}`}class B{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(q,J){let U=[],K=q.segments.map((X)=>{let Q=J.get(X.sceneId);if(Q===void 0)return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:X.recommendedSceneDurationInFrames,voiceDurationInFrames:X.durationInFrames,action:"no_change",finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X;let Z=X.durationInFrames/Q;if(Z>B.UPPER_THRESHOLD){let $=Math.min(Z,B.MAX_RATE);return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:Z>B.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:$,finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X}if(Z<B.LOWER_THRESHOLD){let $=Math.max(Z,B.MIN_RATE);return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:"pad_silence",suggestedRate:$,finalSceneDurationInFrames:Q}),{...X,recommendedSceneDurationInFrames:Q}}return U.push({sceneId:X.sceneId,originalSceneDurationInFrames:Q,voiceDurationInFrames:X.durationInFrames,action:"no_change",finalSceneDurationInFrames:X.recommendedSceneDurationInFrames}),X});return{timingMap:{...q,segments:K},adjustments:U}}}class S{durationEstimator=new G;adapt(q){let J=q.scenes.filter((Q)=>Q.narrationText&&Q.narrationText.trim().length>0),U=J.map((Q,Z)=>{let $=Q.narrationText??"";return{sceneId:Q.id,text:$,estimatedDurationSeconds:this.durationEstimator.estimateSeconds($),contentType:this.inferContentType(Z,J.length)}}),K=U.map((Q)=>Q.text).join(" "),X=U.reduce((Q,Z)=>Q+Z.estimatedDurationSeconds,0);return{fullText:K,segments:U,estimatedDurationSeconds:X}}inferContentType(q,J){if(q===0)return"intro";if(q===J-1)return"cta";if(q===1&&J>3)return"problem";if(q===J-2&&J>3)return"metric";return"solution"}}class w{calculate(q,J,U=1.15){let K=q.map((Q)=>{let Z=Math.ceil(Q.durationMs/1000*J),$=Math.ceil(Z*U),Y=Q.wordTimings?.map((V)=>({word:V.word,startMs:V.startMs,endMs:V.endMs}));return{sceneId:Q.sceneId,durationMs:Q.durationMs,durationInFrames:Z,recommendedSceneDurationInFrames:$,wordTimings:Y}});return{totalDurationMs:q.reduce((Q,Z)=>Q+Z.durationMs,0),segments:K,fps:J}}recalculateForFps(q,J){let U=q.segments.map((K)=>{let X=Math.ceil(K.durationMs/1000*J),Q=K.recommendedSceneDurationInFrames/Math.max(K.durationInFrames,1);return{...K,durationInFrames:X,recommendedSceneDurationInFrames:Math.ceil(X*Q)}});return{...q,segments:U,fps:J}}}class z{concatenator=new A;silenceGenerator=new E;assemble(q,J,U=500){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[K]=q;if(!K)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let X=new Map(J.map(($)=>[$.sceneId,$])),Q=K.audio,Z=[];for(let $=0;$<q.length;$++){let Y=q[$];if(!Y)continue;let V=X.get(Y.sceneId),L=V?.leadingSilenceMs??0;if(L>0)Z.push(this.silenceGenerator.generate(L,Q.format,Q.sampleRateHz,Q.channels??1));Z.push(Y.audio);let _=V?.trailingSilenceMs??($<q.length-1?U:0);if(_>0)Z.push(this.silenceGenerator.generate(_,Q.format,Q.sampleRateHz,Q.channels??1))}return this.concatenator.concatenate(Z)}}var v={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class j{analyze(q,J=1){return q.map((U)=>{let K=v[U.contentType];return{sceneId:U.sceneId,rate:K.rate*J,emphasis:K.emphasis,tone:K.tone,leadingSilenceMs:K.leadingSilenceMs,trailingSilenceMs:K.trailingSilenceMs}})}getDefaults(q){return{...v[q]}}}class F{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(q){this.llm=q?.llm,this.model=q?.model,this.modelSelector=q?.modelSelector,this.selectionContext=q?.selectionContext,this.paceAnalyzer=new j}async plan(q,J=1){if(!this.llm)return this.paceAnalyzer.analyze(q,J);try{return await this.planWithLlm(q,J)}catch{return this.paceAnalyzer.analyze(q,J)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let q=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(q)).modelId}return}async planWithLlm(q,J){if(!this.llm)return this.paceAnalyzer.analyze(q,J);let U=await this.resolveModel(),X=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(q.map((Z)=>({sceneId:Z.sceneId,text:Z.text,contentType:Z.contentType})))}]}],{model:U,temperature:0.3,responseFormat:"json"})).message.content.find((Z)=>Z.type==="text");if(!X||X.type!=="text")return this.paceAnalyzer.analyze(q,J);return JSON.parse(X.text).map((Z)=>({...Z,rate:Z.rate*J}))}}class y{tts;constructor(q){this.tts=q}async synthesizeAll(q,J,U){let K=new Map(U.map((Q)=>[Q.sceneId,Q]));return await Promise.all(q.map((Q)=>this.synthesizeOne(Q,J,K.get(Q.sceneId))))}async synthesizeOne(q,J,U){let K=await this.tts.synthesize({text:q.text,voiceId:J.voiceId,language:J.language,style:J.style,stability:J.stability,rate:U?.rate,emphasis:U?.emphasis});return{sceneId:q.sceneId,audio:K.audio,durationMs:K.audio.durationMs??0,wordTimings:K.wordTimings?.map((X)=>({word:X.word,startMs:X.startMs,endMs:X.endMs}))}}}class M{segmentSynthesizer;emphasisPlanner;audioAssembler=new z;durationEstimator=new G;paceAnalyzer=new j;options;constructor(q){this.options=q,this.segmentSynthesizer=new y(q.tts),this.emphasisPlanner=new F({llm:q.llm,model:q.model})}async synthesize(q){let J=this.buildScript(q);return this.executePipeline(J,q.voice,q.pacing)}async synthesizeForVideo(q){let J=this.buildScriptFromScenePlan(q);return this.executePipeline(J,q.voice,q.pacing,q.fps)}async executePipeline(q,J,U,K){let X=f(),Q=U?.baseRate??1,Z=await this.emphasisPlanner.plan(q.segments,Q),$=await this.segmentSynthesizer.synthesizeAll(q.segments,J,Z),Y=U?.segmentPauseMs??500,V=this.audioAssembler.assemble($,Z,Y),L=K??this.options.fps??30,_=U?.breathingRoomFactor??1.15,W=this.buildTimingMap($,L,_);return{id:X,script:q,pacingDirectives:Z,segments:$,assembledAudio:V,timingMap:W}}buildScript(q){let J=[],U=`${q.content.title}. ${q.content.summary}`;if(J.push({sceneId:"intro",text:U,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(U),contentType:"intro"}),q.content.problems.length>0){let Q=q.content.problems.join(". ");J.push({sceneId:"problems",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"problem"})}if(q.content.solutions.length>0){let Q=q.content.solutions.join(". ");J.push({sceneId:"solutions",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"solution"})}if(q.content.metrics&&q.content.metrics.length>0){let Q=q.content.metrics.join(". ");J.push({sceneId:"metrics",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"metric"})}if(q.content.callToAction)J.push({sceneId:"cta",text:q.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(q.content.callToAction),contentType:"cta"});let K=J.map((Q)=>Q.text).join(" "),X=J.reduce((Q,Z)=>Q+Z.estimatedDurationSeconds,0);return{fullText:K,segments:J,estimatedDurationSeconds:X}}buildScriptFromScenePlan(q){let J=q.scenePlan.scenes.filter((X)=>X.narrationText).map((X)=>{let Q=X.narrationText??"";return{sceneId:X.id,text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"intro"}}),U=J.map((X)=>X.text).join(" "),K=J.reduce((X,Q)=>X+Q.estimatedDurationSeconds,0);return{fullText:U,segments:J,estimatedDurationSeconds:K}}buildTimingMap(q,J,U){let K=q.map((Q)=>{let Z=Math.ceil(Q.durationMs/1000*J);return{sceneId:Q.sceneId,durationMs:Q.durationMs,durationInFrames:Z,recommendedSceneDurationInFrames:Math.ceil(Z*U),wordTimings:Q.wordTimings?.map(($)=>({word:$.word,startMs:$.startMs,endMs:$.endMs}))}});return{totalDurationMs:q.reduce((Q,Z)=>Q+Z.durationMs,0),segments:K,fps:J}}}function f(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`tts_${q}_${J}`}import{defineFeature as p}from"@contractspec/lib.contracts-spec/features";var gq=p({meta:{key:"libs.voice",version:"1.0.0",title:"Voice",description:"Voice capabilities: TTS, STT, and conversational AI",domain:"voice",owners:["@contractspec-core"],tags:["package","libs","voice"],stability:"experimental"}});export{M as VoiceSynthesizer,D as VoiceSessionManager,gq as VoiceFeature,k as TurnDetector,I as TranscriptBuilder,x as Transcriber,w as TimingCalculator,N as SubtitleFormatter,E as SilenceGenerator,y as SegmentSynthesizer,C as SegmentSplitter,S as SceneAdapter,O as ResponseOrchestrator,j as PaceAnalyzer,P as FormatConverter,F as EmphasisPlanner,B as DurationNegotiator,G as DurationEstimator,T as DiarizationMapper,A as AudioConcatenator,z as AudioAssembler};

package/dist/node/stt/diarization-mapper.js

@@ -1 +0,0 @@
-class D{map(z,A="Speaker"){let q=[],v=new Map;for(let h of z)if(h.speakerId&&!q.includes(h.speakerId))q.push(h.speakerId),v.set(h.speakerId,{segmentCount:0,totalSpeakingMs:0});let B=z.map((h)=>{if(!h.speakerId)return h;let w=q.indexOf(h.speakerId),j=`${A} ${w+1}`,y=v.get(h.speakerId);if(!y)return{...h,speakerLabel:j};return y.segmentCount+=1,y.totalSpeakingMs+=h.endMs-h.startMs,{...h,speakerLabel:j}}),C=q.map((h,w)=>{let j=v.get(h);return{id:h,label:`${A} ${w+1}`,segmentCount:j?.segmentCount??0,totalSpeakingMs:j?.totalSpeakingMs??0}});return{segments:B,speakers:C}}}export{D as DiarizationMapper};

package/dist/node/stt/index.js

@@ -1,11 +0,0 @@
-class W{map(T,q="Speaker"){let A=[],R=new Map;for(let w of T)if(w.speakerId&&!A.includes(w.speakerId))A.push(w.speakerId),R.set(w.speakerId,{segmentCount:0,totalSpeakingMs:0});let E=T.map((w)=>{if(!w.speakerId)return w;let J=A.indexOf(w.speakerId),K=`${q} ${J+1}`,N=R.get(w.speakerId);if(!N)return{...w,speakerLabel:K};return N.segmentCount+=1,N.totalSpeakingMs+=w.endMs-w.startMs,{...w,speakerLabel:K}}),G=A.map((w,J)=>{let K=R.get(w);return{id:w,label:`${q} ${J+1}`,segmentCount:K?.segmentCount??0,totalSpeakingMs:K?.totalSpeakingMs??0}});return{segments:E,speakers:G}}}class U{static DEFAULT_MAX_CHUNK_MS=300000;split(T,q=U.DEFAULT_MAX_CHUNK_MS){let A=T.durationMs??this.estimateDurationMs(T);if(A<=q)return[T];let R=[],E=T.data.length/Math.max(A,1),G=0;while(G<A){let w=Math.min(q,A-G),J=Math.floor(G*E),K=Math.floor((G+w)*E);R.push({data:T.data.slice(J,K),format:T.format,sampleRateHz:T.sampleRateHz,durationMs:w,channels:T.channels}),G+=w}return R}estimateDurationMs(T){let A=T.channels??1,R=T.data.length/(2*A);return Math.ceil(R/T.sampleRateHz*1000)}}class X{toSRT(T){return T.map((q,A)=>{let R=this.formatTimeSRT(q.startMs),E=this.formatTimeSRT(q.endMs),G=q.speakerLabel?`[${q.speakerLabel}] `:"";return`${A+1}
-${R} --> ${E}
-${G}${q.text}`}).join(`
-
-`)}toVTT(T){return`WEBVTT
-
-`+T.map((R,E)=>{let G=this.formatTimeVTT(R.startMs),w=this.formatTimeVTT(R.endMs),J=R.speakerLabel?`<v ${R.speakerLabel}>`:"";return`${E+1}
-${G} --> ${w}
-${J}${R.text}`}).join(`
-
-`)}formatTimeSRT(T){let q=Math.floor(T/3600000),A=Math.floor(T%3600000/60000),R=Math.floor(T%60000/1000),E=T%1000;return`${this.pad(q,2)}:${this.pad(A,2)}:${this.pad(R,2)},${this.pad(E,3)}`}formatTimeVTT(T){let q=Math.floor(T/3600000),A=Math.floor(T%3600000/60000),R=Math.floor(T%60000/1000),E=T%1000;return`${this.pad(q,2)}:${this.pad(A,2)}:${this.pad(R,2)}.${this.pad(E,3)}`}pad(T,q){return T.toString().padStart(q,"0")}}class ${stt;segmentSplitter=new U;diarizationMapper=new W;subtitleFormatter=new X;constructor(T){this.stt=T.stt}async transcribe(T){let q=z(),A=this.segmentSplitter.split(T.audio),R=[],E="",G=0,w=0;for(let Q of A){let Z=await this.stt.transcribe({audio:Q,language:T.language,diarize:T.diarize,speakerCount:T.speakerCount,wordTimestamps:!0,vocabularyHints:T.vocabularyHints}),j=Z.segments.map((L)=>({text:L.text,startMs:L.startMs+w,endMs:L.endMs+w,speakerId:L.speakerId,speakerName:L.speakerName,confidence:L.confidence}));R.push(...j),E+=(E?" ":"")+Z.text,G+=Z.durationMs,w+=Q.durationMs??0}let J=R,K;if(T.diarize){let Q=this.diarizationMapper.map(R);J=Q.segments,K=Q.speakers}let N={text:E,segments:J,language:T.language??"en",durationMs:G},Y,_=T.subtitleFormat??"none";if(_==="srt")Y=this.subtitleFormatter.toSRT(J);else if(_==="vtt")Y=this.subtitleFormatter.toVTT(J);return{id:q,transcript:N,subtitles:Y,speakers:K}}async*transcribeStream(T,q){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let A=this.stt.transcribeStream(T,{language:q?.language,diarize:q?.diarize,speakerCount:q?.speakerCount,wordTimestamps:!0,vocabularyHints:q?.vocabularyHints});for await(let R of A)yield{text:R.text,startMs:R.startMs,endMs:R.endMs,speakerId:R.speakerId,speakerLabel:R.speakerName,confidence:R.confidence}}}function z(){let T=Date.now().toString(36),q=Math.random().toString(36).slice(2,8);return`stt_${T}_${q}`}export{$ as Transcriber,X as SubtitleFormatter,U as SegmentSplitter,W as DiarizationMapper};

package/dist/node/stt/segment-splitter.js

@@ -1 +0,0 @@
-class F{static DEFAULT_MAX_CHUNK_MS=300000;split(j,z=F.DEFAULT_MAX_CHUNK_MS){let q=j.durationMs??this.estimateDurationMs(j);if(q<=z)return[j];let w=[],E=j.data.length/Math.max(q,1),v=0;while(v<q){let A=Math.min(z,q-v),G=Math.floor(v*E),H=Math.floor((v+A)*E);w.push({data:j.data.slice(G,H),format:j.format,sampleRateHz:j.sampleRateHz,durationMs:A,channels:j.channels}),v+=A}return w}estimateDurationMs(j){let q=j.channels??1,w=j.data.length/(2*q);return Math.ceil(w/j.sampleRateHz*1000)}}export{F as SegmentSplitter};

package/dist/node/stt/subtitle-formatter.js

@@ -1,11 +0,0 @@
-class A{toSRT(f){return f.map((j,w)=>{let k=this.formatTimeSRT(j.startMs),q=this.formatTimeSRT(j.endMs),x=j.speakerLabel?`[${j.speakerLabel}] `:"";return`${w+1}
-${k} --> ${q}
-${x}${j.text}`}).join(`
-
-`)}toVTT(f){return`WEBVTT
-
-`+f.map((k,q)=>{let x=this.formatTimeVTT(k.startMs),y=this.formatTimeVTT(k.endMs),z=k.speakerLabel?`<v ${k.speakerLabel}>`:"";return`${q+1}
-${x} --> ${y}
-${z}${k.text}`}).join(`
-
-`)}formatTimeSRT(f){let j=Math.floor(f/3600000),w=Math.floor(f%3600000/60000),k=Math.floor(f%60000/1000),q=f%1000;return`${this.pad(j,2)}:${this.pad(w,2)}:${this.pad(k,2)},${this.pad(q,3)}`}formatTimeVTT(f){let j=Math.floor(f/3600000),w=Math.floor(f%3600000/60000),k=Math.floor(f%60000/1000),q=f%1000;return`${this.pad(j,2)}:${this.pad(w,2)}:${this.pad(k,2)}.${this.pad(q,3)}`}pad(f,j){return f.toString().padStart(j,"0")}}export{A as SubtitleFormatter};

package/dist/node/stt/transcriber.js

@@ -1,11 +0,0 @@
-class Z{map(q,A="Speaker"){let G=[],w=new Map;for(let E of q)if(E.speakerId&&!G.includes(E.speakerId))G.push(E.speakerId),w.set(E.speakerId,{segmentCount:0,totalSpeakingMs:0});let J=q.map((E)=>{if(!E.speakerId)return E;let L=G.indexOf(E.speakerId),N=`${A} ${L+1}`,R=w.get(E.speakerId);if(!R)return{...E,speakerLabel:N};return R.segmentCount+=1,R.totalSpeakingMs+=E.endMs-E.startMs,{...E,speakerLabel:N}}),K=G.map((E,L)=>{let N=w.get(E);return{id:E,label:`${A} ${L+1}`,segmentCount:N?.segmentCount??0,totalSpeakingMs:N?.totalSpeakingMs??0}});return{segments:J,speakers:K}}}class W{static DEFAULT_MAX_CHUNK_MS=300000;split(q,A=W.DEFAULT_MAX_CHUNK_MS){let G=q.durationMs??this.estimateDurationMs(q);if(G<=A)return[q];let w=[],J=q.data.length/Math.max(G,1),K=0;while(K<G){let E=Math.min(A,G-K),L=Math.floor(K*J),N=Math.floor((K+E)*J);w.push({data:q.data.slice(L,N),format:q.format,sampleRateHz:q.sampleRateHz,durationMs:E,channels:q.channels}),K+=E}return w}estimateDurationMs(q){let G=q.channels??1,w=q.data.length/(2*G);return Math.ceil(w/q.sampleRateHz*1000)}}class _{toSRT(q){return q.map((A,G)=>{let w=this.formatTimeSRT(A.startMs),J=this.formatTimeSRT(A.endMs),K=A.speakerLabel?`[${A.speakerLabel}] `:"";return`${G+1}
-${w} --> ${J}
-${K}${A.text}`}).join(`
-
-`)}toVTT(q){return`WEBVTT
-
-`+q.map((w,J)=>{let K=this.formatTimeVTT(w.startMs),E=this.formatTimeVTT(w.endMs),L=w.speakerLabel?`<v ${w.speakerLabel}>`:"";return`${J+1}
-${K} --> ${E}
-${L}${w.text}`}).join(`
-
-`)}formatTimeSRT(q){let A=Math.floor(q/3600000),G=Math.floor(q%3600000/60000),w=Math.floor(q%60000/1000),J=q%1000;return`${this.pad(A,2)}:${this.pad(G,2)}:${this.pad(w,2)},${this.pad(J,3)}`}formatTimeVTT(q){let A=Math.floor(q/3600000),G=Math.floor(q%3600000/60000),w=Math.floor(q%60000/1000),J=q%1000;return`${this.pad(A,2)}:${this.pad(G,2)}:${this.pad(w,2)}.${this.pad(J,3)}`}pad(q,A){return q.toString().padStart(A,"0")}}class j{stt;segmentSplitter=new W;diarizationMapper=new Z;subtitleFormatter=new _;constructor(q){this.stt=q.stt}async transcribe(q){let A=z(),G=this.segmentSplitter.split(q.audio),w=[],J="",K=0,E=0;for(let U of G){let Y=await this.stt.transcribe({audio:U,language:q.language,diarize:q.diarize,speakerCount:q.speakerCount,wordTimestamps:!0,vocabularyHints:q.vocabularyHints}),V=Y.segments.map((Q)=>({text:Q.text,startMs:Q.startMs+E,endMs:Q.endMs+E,speakerId:Q.speakerId,speakerName:Q.speakerName,confidence:Q.confidence}));w.push(...V),J+=(J?" ":"")+Y.text,K+=Y.durationMs,E+=U.durationMs??0}let L=w,N;if(q.diarize){let U=this.diarizationMapper.map(w);L=U.segments,N=U.speakers}let R={text:J,segments:L,language:q.language??"en",durationMs:K},X,$=q.subtitleFormat??"none";if($==="srt")X=this.subtitleFormatter.toSRT(L);else if($==="vtt")X=this.subtitleFormatter.toVTT(L);return{id:A,transcript:R,subtitles:X,speakers:N}}async*transcribeStream(q,A){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let G=this.stt.transcribeStream(q,{language:A?.language,diarize:A?.diarize,speakerCount:A?.speakerCount,wordTimestamps:!0,vocabularyHints:A?.vocabularyHints});for await(let w of G)yield{text:w.text,startMs:w.startMs,endMs:w.endMs,speakerId:w.speakerId,speakerLabel:w.speakerName,confidence:w.confidence}}}function z(){let q=Date.now().toString(36),A=Math.random().toString(36).slice(2,8);return`stt_${q}_${A}`}export{j as Transcriber};

package/dist/node/sync/duration-negotiator.js

@@ -1 +0,0 @@
-class q{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(z,A){let v=[],B=z.segments.map((b)=>{let k=A.get(b.sceneId);if(k===void 0)return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:b.recommendedSceneDurationInFrames,voiceDurationInFrames:b.durationInFrames,action:"no_change",finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b;let x=b.durationInFrames/k;if(x>q.UPPER_THRESHOLD){let y=Math.min(x,q.MAX_RATE);return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:x>q.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:y,finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b}if(x<q.LOWER_THRESHOLD){let y=Math.max(x,q.MIN_RATE);return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:"pad_silence",suggestedRate:y,finalSceneDurationInFrames:k}),{...b,recommendedSceneDurationInFrames:k}}return v.push({sceneId:b.sceneId,originalSceneDurationInFrames:k,voiceDurationInFrames:b.durationInFrames,action:"no_change",finalSceneDurationInFrames:b.recommendedSceneDurationInFrames}),b});return{timingMap:{...z,segments:B},adjustments:v}}}export{q as DurationNegotiator};

package/dist/node/sync/index.js

@@ -1 +0,0 @@
-class O{static DEFAULT_WPM=150;estimateSeconds(A,B){let G=B??O.DEFAULT_WPM,H=A.split(/\s+/).filter(Boolean).length;return Math.ceil(H/G*60)}estimateMs(A,B){let G=B??O.DEFAULT_WPM,H=A.split(/\s+/).filter(Boolean).length;return Math.ceil(H/G*60*1000)}estimateWordCount(A,B){let G=B??O.DEFAULT_WPM;return Math.round(A/60*G)}}class L{static UPPER_THRESHOLD=1.1;static LOWER_THRESHOLD=0.7;static MAX_RATE=1.3;static MIN_RATE=0.8;negotiate(A,B){let G=[],H=A.segments.map((z)=>{let q=B.get(z.sceneId);if(q===void 0)return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:z.recommendedSceneDurationInFrames,voiceDurationInFrames:z.durationInFrames,action:"no_change",finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z;let J=z.durationInFrames/q;if(J>L.UPPER_THRESHOLD){let K=Math.min(J,L.MAX_RATE);return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:J>L.MAX_RATE?"extend_scene":"suggest_rate_change",suggestedRate:K,finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z}if(J<L.LOWER_THRESHOLD){let K=Math.max(J,L.MIN_RATE);return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:"pad_silence",suggestedRate:K,finalSceneDurationInFrames:q}),{...z,recommendedSceneDurationInFrames:q}}return G.push({sceneId:z.sceneId,originalSceneDurationInFrames:q,voiceDurationInFrames:z.durationInFrames,action:"no_change",finalSceneDurationInFrames:z.recommendedSceneDurationInFrames}),z});return{timingMap:{...A,segments:H},adjustments:G}}}class U{durationEstimator=new O;adapt(A){let B=A.scenes.filter((q)=>q.narrationText&&q.narrationText.trim().length>0),G=B.map((q,J)=>{let K=q.narrationText??"";return{sceneId:q.id,text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:this.inferContentType(J,B.length)}}),H=G.map((q)=>q.text).join(" "),z=G.reduce((q,J)=>q+J.estimatedDurationSeconds,0);return{fullText:H,segments:G,estimatedDurationSeconds:z}}inferContentType(A,B){if(A===0)return"intro";if(A===B-1)return"cta";if(A===1&&B>3)return"problem";if(A===B-2&&B>3)return"metric";return"solution"}}class X{calculate(A,B,G=1.15){let H=A.map((q)=>{let J=Math.ceil(q.durationMs/1000*B),K=Math.ceil(J*G),Y=q.wordTimings?.map((Q)=>({word:Q.word,startMs:Q.startMs,endMs:Q.endMs}));return{sceneId:q.sceneId,durationMs:q.durationMs,durationInFrames:J,recommendedSceneDurationInFrames:K,wordTimings:Y}});return{totalDurationMs:A.reduce((q,J)=>q+J.durationMs,0),segments:H,fps:B}}recalculateForFps(A,B){let G=A.segments.map((H)=>{let z=Math.ceil(H.durationMs/1000*B),q=H.recommendedSceneDurationInFrames/Math.max(H.durationInFrames,1);return{...H,durationInFrames:z,recommendedSceneDurationInFrames:Math.ceil(z*q)}});return{...A,segments:G,fps:B}}}export{X as TimingCalculator,U as SceneAdapter,L as DurationNegotiator};

package/dist/node/sync/scene-adapter.js

@@ -1 +0,0 @@
-class G{static DEFAULT_WPM=150;estimateSeconds(q,j){let z=j??G.DEFAULT_WPM,H=q.split(/\s+/).filter(Boolean).length;return Math.ceil(H/z*60)}estimateMs(q,j){let z=j??G.DEFAULT_WPM,H=q.split(/\s+/).filter(Boolean).length;return Math.ceil(H/z*60*1000)}estimateWordCount(q,j){let z=j??G.DEFAULT_WPM;return Math.round(q/60*z)}}class L{durationEstimator=new G;adapt(q){let j=q.scenes.filter((B)=>B.narrationText&&B.narrationText.trim().length>0),z=j.map((B,I)=>{let J=B.narrationText??"";return{sceneId:B.id,text:J,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(J),contentType:this.inferContentType(I,j.length)}}),H=z.map((B)=>B.text).join(" "),K=z.reduce((B,I)=>B+I.estimatedDurationSeconds,0);return{fullText:H,segments:z,estimatedDurationSeconds:K}}inferContentType(q,j){if(q===0)return"intro";if(q===j-1)return"cta";if(q===1&&j>3)return"problem";if(q===j-2&&j>3)return"metric";return"solution"}}export{L as SceneAdapter};

package/dist/node/sync/timing-calculator.js

@@ -1 +0,0 @@
-class z{calculate(h,j,l=1.15){let f=h.map((b)=>{let k=Math.ceil(b.durationMs/1000*j),x=Math.ceil(k*l),y=b.wordTimings?.map((v)=>({word:v.word,startMs:v.startMs,endMs:v.endMs}));return{sceneId:b.sceneId,durationMs:b.durationMs,durationInFrames:k,recommendedSceneDurationInFrames:x,wordTimings:y}});return{totalDurationMs:h.reduce((b,k)=>b+k.durationMs,0),segments:f,fps:j}}recalculateForFps(h,j){let l=h.segments.map((f)=>{let q=Math.ceil(f.durationMs/1000*j),b=f.recommendedSceneDurationInFrames/Math.max(f.durationInFrames,1);return{...f,durationInFrames:q,recommendedSceneDurationInFrames:Math.ceil(q*b)}});return{...h,segments:l,fps:j}}}export{z as TimingCalculator};

package/dist/node/tts/audio-assembler.js

@@ -1 +0,0 @@
-class V{concatenate(j){if(j.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[x]=j;if(!x)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(j.length===1)return{...x};let{format:N,sampleRateHz:E}=x,Q=x.channels??1;for(let k of j){if(k.format!==N)throw Error(`Format mismatch: expected ${N}, got ${k.format}`);if(k.sampleRateHz!==E)throw Error(`Sample rate mismatch: expected ${E}, got ${k.sampleRateHz}`)}let J=j.reduce((k,O)=>k+O.data.length,0),K=new Uint8Array(J),q=0;for(let k of j)K.set(k.data,q),q+=k.data.length;let U=j.reduce((k,O)=>k+(O.durationMs??0),0);return{data:K,format:N,sampleRateHz:E,durationMs:U,channels:Q}}}class W{generate(j,x="wav",N=44100,E=1){let Q=Math.ceil(N*j/1000),J=2,K=Q*2*E;return{data:new Uint8Array(K),format:x,sampleRateHz:N,durationMs:j,channels:E}}}class Y{concatenator=new V;silenceGenerator=new W;assemble(j,x,N=500){if(j.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[E]=j;if(!E)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let Q=new Map(x.map((q)=>[q.sceneId,q])),J=E.audio,K=[];for(let q=0;q<j.length;q++){let U=j[q];if(!U)continue;let k=Q.get(U.sceneId),O=k?.leadingSilenceMs??0;if(O>0)K.push(this.silenceGenerator.generate(O,J.format,J.sampleRateHz,J.channels??1));K.push(U.audio);let X=k?.trailingSilenceMs??(q<j.length-1?N:0);if(X>0)K.push(this.silenceGenerator.generate(X,J.format,J.sampleRateHz,J.channels??1))}return this.concatenator.concatenate(K)}}export{Y as AudioAssembler};

package/dist/node/tts/emphasis-planner.js

@@ -1,2 +0,0 @@
-var H={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class F{analyze(j,q=1){return j.map((B)=>{let w=H[B.contentType];return{sceneId:B.sceneId,rate:w.rate*q,emphasis:w.emphasis,tone:w.tone,leadingSilenceMs:w.leadingSilenceMs,trailingSilenceMs:w.trailingSilenceMs}})}getDefaults(j){return{...H[j]}}}class J{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(j){this.llm=j?.llm,this.model=j?.model,this.modelSelector=j?.modelSelector,this.selectionContext=j?.selectionContext,this.paceAnalyzer=new F}async plan(j,q=1){if(!this.llm)return this.paceAnalyzer.analyze(j,q);try{return await this.planWithLlm(j,q)}catch{return this.paceAnalyzer.analyze(j,q)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let j=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(j)).modelId}return}async planWithLlm(j,q){if(!this.llm)return this.paceAnalyzer.analyze(j,q);let B=await this.resolveModel(),D=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(j.map((v)=>({sceneId:v.sceneId,text:v.text,contentType:v.contentType})))}]}],{model:B,temperature:0.3,responseFormat:"json"})).message.content.find((v)=>v.type==="text");if(!D||D.type!=="text")return this.paceAnalyzer.analyze(j,q);return JSON.parse(D.text).map((v)=>({...v,rate:v.rate*q}))}}export{J as EmphasisPlanner};

package/dist/node/tts/index.js

@@ -1,2 +0,0 @@
-class H{concatenate(V){if(V.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[q]=V;if(!q)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(V.length===1)return{...q};let{format:J,sampleRateHz:K}=q,U=q.channels??1;for(let Z of V){if(Z.format!==J)throw Error(`Format mismatch: expected ${J}, got ${Z.format}`);if(Z.sampleRateHz!==K)throw Error(`Sample rate mismatch: expected ${K}, got ${Z.sampleRateHz}`)}let B=V.reduce((Z,$)=>Z+$.data.length,0),Q=new Uint8Array(B),X=0;for(let Z of V)Q.set(Z.data,X),X+=Z.data.length;let j=V.reduce((Z,$)=>Z+($.durationMs??0),0);return{data:Q,format:J,sampleRateHz:K,durationMs:j,channels:U}}}class O{static DEFAULT_WPM=150;estimateSeconds(V,q){let J=q??O.DEFAULT_WPM,K=V.split(/\s+/).filter(Boolean).length;return Math.ceil(K/J*60)}estimateMs(V,q){let J=q??O.DEFAULT_WPM,K=V.split(/\s+/).filter(Boolean).length;return Math.ceil(K/J*60*1000)}estimateWordCount(V,q){let J=q??O.DEFAULT_WPM;return Math.round(V/60*J)}}class _{generate(V,q="wav",J=44100,K=1){let U=Math.ceil(J*V/1000),B=2,Q=U*2*K;return{data:new Uint8Array(Q),format:q,sampleRateHz:J,durationMs:V,channels:K}}}class T{concatenator=new H;silenceGenerator=new _;assemble(V,q,J=500){if(V.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[K]=V;if(!K)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let U=new Map(q.map((X)=>[X.sceneId,X])),B=K.audio,Q=[];for(let X=0;X<V.length;X++){let j=V[X];if(!j)continue;let Z=U.get(j.sceneId),$=Z?.leadingSilenceMs??0;if($>0)Q.push(this.silenceGenerator.generate($,B.format,B.sampleRateHz,B.channels??1));Q.push(j.audio);let k=Z?.trailingSilenceMs??(X<V.length-1?J:0);if(k>0)Q.push(this.silenceGenerator.generate(k,B.format,B.sampleRateHz,B.channels??1))}return this.concatenator.concatenate(Q)}}var G={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class W{analyze(V,q=1){return V.map((J)=>{let K=G[J.contentType];return{sceneId:J.sceneId,rate:K.rate*q,emphasis:K.emphasis,tone:K.tone,leadingSilenceMs:K.leadingSilenceMs,trailingSilenceMs:K.trailingSilenceMs}})}getDefaults(V){return{...G[V]}}}class Y{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(V){this.llm=V?.llm,this.model=V?.model,this.modelSelector=V?.modelSelector,this.selectionContext=V?.selectionContext,this.paceAnalyzer=new W}async plan(V,q=1){if(!this.llm)return this.paceAnalyzer.analyze(V,q);try{return await this.planWithLlm(V,q)}catch{return this.paceAnalyzer.analyze(V,q)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let V=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(V)).modelId}return}async planWithLlm(V,q){if(!this.llm)return this.paceAnalyzer.analyze(V,q);let J=await this.resolveModel(),U=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(V.map((Q)=>({sceneId:Q.sceneId,text:Q.text,contentType:Q.contentType})))}]}],{model:J,temperature:0.3,responseFormat:"json"})).message.content.find((Q)=>Q.type==="text");if(!U||U.type!=="text")return this.paceAnalyzer.analyze(V,q);return JSON.parse(U.text).map((Q)=>({...Q,rate:Q.rate*q}))}}class L{tts;constructor(V){this.tts=V}async synthesizeAll(V,q,J){let K=new Map(J.map((B)=>[B.sceneId,B]));return await Promise.all(V.map((B)=>this.synthesizeOne(B,q,K.get(B.sceneId))))}async synthesizeOne(V,q,J){let K=await this.tts.synthesize({text:V.text,voiceId:q.voiceId,language:q.language,style:q.style,stability:q.stability,rate:J?.rate,emphasis:J?.emphasis});return{sceneId:V.sceneId,audio:K.audio,durationMs:K.audio.durationMs??0,wordTimings:K.wordTimings?.map((U)=>({word:U.word,startMs:U.startMs,endMs:U.endMs}))}}}class N{segmentSynthesizer;emphasisPlanner;audioAssembler=new T;durationEstimator=new O;paceAnalyzer=new W;options;constructor(V){this.options=V,this.segmentSynthesizer=new L(V.tts),this.emphasisPlanner=new Y({llm:V.llm,model:V.model})}async synthesize(V){let q=this.buildScript(V);return this.executePipeline(q,V.voice,V.pacing)}async synthesizeForVideo(V){let q=this.buildScriptFromScenePlan(V);return this.executePipeline(q,V.voice,V.pacing,V.fps)}async executePipeline(V,q,J,K){let U=F(),B=J?.baseRate??1,Q=await this.emphasisPlanner.plan(V.segments,B),X=await this.segmentSynthesizer.synthesizeAll(V.segments,q,Q),j=J?.segmentPauseMs??500,Z=this.audioAssembler.assemble(X,Q,j),$=K??this.options.fps??30,k=J?.breathingRoomFactor??1.15,C=this.buildTimingMap(X,$,k);return{id:U,script:V,pacingDirectives:Q,segments:X,assembledAudio:Z,timingMap:C}}buildScript(V){let q=[],J=`${V.content.title}. ${V.content.summary}`;if(q.push({sceneId:"intro",text:J,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(J),contentType:"intro"}),V.content.problems.length>0){let B=V.content.problems.join(". ");q.push({sceneId:"problems",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"problem"})}if(V.content.solutions.length>0){let B=V.content.solutions.join(". ");q.push({sceneId:"solutions",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"solution"})}if(V.content.metrics&&V.content.metrics.length>0){let B=V.content.metrics.join(". ");q.push({sceneId:"metrics",text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"metric"})}if(V.content.callToAction)q.push({sceneId:"cta",text:V.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(V.content.callToAction),contentType:"cta"});let K=q.map((B)=>B.text).join(" "),U=q.reduce((B,Q)=>B+Q.estimatedDurationSeconds,0);return{fullText:K,segments:q,estimatedDurationSeconds:U}}buildScriptFromScenePlan(V){let q=V.scenePlan.scenes.filter((U)=>U.narrationText).map((U)=>{let B=U.narrationText??"";return{sceneId:U.id,text:B,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(B),contentType:"intro"}}),J=q.map((U)=>U.text).join(" "),K=q.reduce((U,B)=>U+B.estimatedDurationSeconds,0);return{fullText:J,segments:q,estimatedDurationSeconds:K}}buildTimingMap(V,q,J){let K=V.map((B)=>{let Q=Math.ceil(B.durationMs/1000*q);return{sceneId:B.sceneId,durationMs:B.durationMs,durationInFrames:Q,recommendedSceneDurationInFrames:Math.ceil(Q*J),wordTimings:B.wordTimings?.map((X)=>({word:X.word,startMs:X.startMs,endMs:X.endMs}))}});return{totalDurationMs:V.reduce((B,Q)=>B+Q.durationMs,0),segments:K,fps:q}}}function F(){let V=Date.now().toString(36),q=Math.random().toString(36).slice(2,8);return`tts_${V}_${q}`}export{N as VoiceSynthesizer,L as SegmentSynthesizer,W as PaceAnalyzer,Y as EmphasisPlanner,T as AudioAssembler};

package/dist/node/tts/pace-analyzer.js

@@ -1 +0,0 @@
-var v={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class x{analyze(k,w=1){return k.map((q)=>{let j=v[q.contentType];return{sceneId:q.sceneId,rate:j.rate*w,emphasis:j.emphasis,tone:j.tone,leadingSilenceMs:j.leadingSilenceMs,trailingSilenceMs:j.trailingSilenceMs}})}getDefaults(k){return{...v[k]}}}export{x as PaceAnalyzer};

package/dist/node/tts/segment-synthesizer.js

@@ -1 +0,0 @@
-class C{tts;constructor(j){this.tts=j}async synthesizeAll(j,k,A){let q=new Map(A.map((x)=>[x.sceneId,x]));return await Promise.all(j.map((x)=>this.synthesizeOne(x,k,q.get(x.sceneId))))}async synthesizeOne(j,k,A){let q=await this.tts.synthesize({text:j.text,voiceId:k.voiceId,language:k.language,style:k.style,stability:k.stability,rate:A?.rate,emphasis:A?.emphasis});return{sceneId:j.sceneId,audio:q.audio,durationMs:q.audio.durationMs??0,wordTimings:q.wordTimings?.map((B)=>({word:B.word,startMs:B.startMs,endMs:B.endMs}))}}}export{C as SegmentSynthesizer};

package/dist/node/tts/voice-synthesizer.js

@@ -1,2 +0,0 @@
-class Y{concatenate(q){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[J]=q;if(!J)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};if(q.length===1)return{...J};let{format:Q,sampleRateHz:U}=J,Z=J.channels??1;for(let V of q){if(V.format!==Q)throw Error(`Format mismatch: expected ${Q}, got ${V.format}`);if(V.sampleRateHz!==U)throw Error(`Sample rate mismatch: expected ${U}, got ${V.sampleRateHz}`)}let K=q.reduce((V,W)=>V+W.data.length,0),X=new Uint8Array(K),$=0;for(let V of q)X.set(V.data,$),$+=V.data.length;let j=q.reduce((V,W)=>V+(W.durationMs??0),0);return{data:X,format:Q,sampleRateHz:U,durationMs:j,channels:Z}}}class k{static DEFAULT_WPM=150;estimateSeconds(q,J){let Q=J??k.DEFAULT_WPM,U=q.split(/\s+/).filter(Boolean).length;return Math.ceil(U/Q*60)}estimateMs(q,J){let Q=J??k.DEFAULT_WPM,U=q.split(/\s+/).filter(Boolean).length;return Math.ceil(U/Q*60*1000)}estimateWordCount(q,J){let Q=J??k.DEFAULT_WPM;return Math.round(q/60*Q)}}class L{generate(q,J="wav",Q=44100,U=1){let Z=Math.ceil(Q*q/1000),K=2,X=Z*2*U;return{data:new Uint8Array(X),format:J,sampleRateHz:Q,durationMs:q,channels:U}}}class H{concatenator=new Y;silenceGenerator=new L;assemble(q,J,Q=500){if(q.length===0)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let[U]=q;if(!U)return{data:new Uint8Array(0),format:"wav",sampleRateHz:44100,durationMs:0,channels:1};let Z=new Map(J.map(($)=>[$.sceneId,$])),K=U.audio,X=[];for(let $=0;$<q.length;$++){let j=q[$];if(!j)continue;let V=Z.get(j.sceneId),W=V?.leadingSilenceMs??0;if(W>0)X.push(this.silenceGenerator.generate(W,K.format,K.sampleRateHz,K.channels??1));X.push(j.audio);let O=V?.trailingSilenceMs??($<q.length-1?Q:0);if(O>0)X.push(this.silenceGenerator.generate(O,K.format,K.sampleRateHz,K.channels??1))}return this.concatenator.concatenate(X)}}var N={intro:{rate:0.95,emphasis:"normal",tone:"authoritative",leadingSilenceMs:0,trailingSilenceMs:500},problem:{rate:0.9,emphasis:"strong",tone:"urgent",leadingSilenceMs:300,trailingSilenceMs:500},solution:{rate:1,emphasis:"normal",tone:"calm",leadingSilenceMs:300,trailingSilenceMs:500},metric:{rate:0.85,emphasis:"strong",tone:"excited",leadingSilenceMs:300,trailingSilenceMs:600},cta:{rate:0.9,emphasis:"strong",tone:"authoritative",leadingSilenceMs:400,trailingSilenceMs:0},transition:{rate:1.1,emphasis:"reduced",tone:"neutral",leadingSilenceMs:200,trailingSilenceMs:300}};class B{analyze(q,J=1){return q.map((Q)=>{let U=N[Q.contentType];return{sceneId:Q.sceneId,rate:U.rate*J,emphasis:U.emphasis,tone:U.tone,leadingSilenceMs:U.leadingSilenceMs,trailingSilenceMs:U.trailingSilenceMs}})}getDefaults(q){return{...N[q]}}}class _{llm;model;modelSelector;selectionContext;paceAnalyzer;constructor(q){this.llm=q?.llm,this.model=q?.model,this.modelSelector=q?.modelSelector,this.selectionContext=q?.selectionContext,this.paceAnalyzer=new B}async plan(q,J=1){if(!this.llm)return this.paceAnalyzer.analyze(q,J);try{return await this.planWithLlm(q,J)}catch{return this.paceAnalyzer.analyze(q,J)}}async resolveModel(){if(this.model)return this.model;if(this.modelSelector){let q=this.selectionContext??{taskDimension:"reasoning"};return(await this.modelSelector.select(q)).modelId}return}async planWithLlm(q,J){if(!this.llm)return this.paceAnalyzer.analyze(q,J);let Q=await this.resolveModel(),Z=(await this.llm.chat([{role:"system",content:[{type:"text",text:["You are a voice director planning emphasis and pacing for TTS narration.","For each segment, return a JSON array of directives.","Each directive has: sceneId, rate (0.7-1.3), emphasis (reduced|normal|strong),","tone (neutral|urgent|excited|calm|authoritative), leadingSilenceMs, trailingSilenceMs.","Return ONLY a JSON array, no other text."].join(`
-`)}]},{role:"user",content:[{type:"text",text:JSON.stringify(q.map((X)=>({sceneId:X.sceneId,text:X.text,contentType:X.contentType})))}]}],{model:Q,temperature:0.3,responseFormat:"json"})).message.content.find((X)=>X.type==="text");if(!Z||Z.type!=="text")return this.paceAnalyzer.analyze(q,J);return JSON.parse(Z.text).map((X)=>({...X,rate:X.rate*J}))}}class G{tts;constructor(q){this.tts=q}async synthesizeAll(q,J,Q){let U=new Map(Q.map((K)=>[K.sceneId,K]));return await Promise.all(q.map((K)=>this.synthesizeOne(K,J,U.get(K.sceneId))))}async synthesizeOne(q,J,Q){let U=await this.tts.synthesize({text:q.text,voiceId:J.voiceId,language:J.language,style:J.style,stability:J.stability,rate:Q?.rate,emphasis:Q?.emphasis});return{sceneId:q.sceneId,audio:U.audio,durationMs:U.audio.durationMs??0,wordTimings:U.wordTimings?.map((Z)=>({word:Z.word,startMs:Z.startMs,endMs:Z.endMs}))}}}class I{segmentSynthesizer;emphasisPlanner;audioAssembler=new H;durationEstimator=new k;paceAnalyzer=new B;options;constructor(q){this.options=q,this.segmentSynthesizer=new G(q.tts),this.emphasisPlanner=new _({llm:q.llm,model:q.model})}async synthesize(q){let J=this.buildScript(q);return this.executePipeline(J,q.voice,q.pacing)}async synthesizeForVideo(q){let J=this.buildScriptFromScenePlan(q);return this.executePipeline(J,q.voice,q.pacing,q.fps)}async executePipeline(q,J,Q,U){let Z=C(),K=Q?.baseRate??1,X=await this.emphasisPlanner.plan(q.segments,K),$=await this.segmentSynthesizer.synthesizeAll(q.segments,J,X),j=Q?.segmentPauseMs??500,V=this.audioAssembler.assemble($,X,j),W=U??this.options.fps??30,O=Q?.breathingRoomFactor??1.15,F=this.buildTimingMap($,W,O);return{id:Z,script:q,pacingDirectives:X,segments:$,assembledAudio:V,timingMap:F}}buildScript(q){let J=[],Q=`${q.content.title}. ${q.content.summary}`;if(J.push({sceneId:"intro",text:Q,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(Q),contentType:"intro"}),q.content.problems.length>0){let K=q.content.problems.join(". ");J.push({sceneId:"problems",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"problem"})}if(q.content.solutions.length>0){let K=q.content.solutions.join(". ");J.push({sceneId:"solutions",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"solution"})}if(q.content.metrics&&q.content.metrics.length>0){let K=q.content.metrics.join(". ");J.push({sceneId:"metrics",text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"metric"})}if(q.content.callToAction)J.push({sceneId:"cta",text:q.content.callToAction,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(q.content.callToAction),contentType:"cta"});let U=J.map((K)=>K.text).join(" "),Z=J.reduce((K,X)=>K+X.estimatedDurationSeconds,0);return{fullText:U,segments:J,estimatedDurationSeconds:Z}}buildScriptFromScenePlan(q){let J=q.scenePlan.scenes.filter((Z)=>Z.narrationText).map((Z)=>{let K=Z.narrationText??"";return{sceneId:Z.id,text:K,estimatedDurationSeconds:this.durationEstimator.estimateSeconds(K),contentType:"intro"}}),Q=J.map((Z)=>Z.text).join(" "),U=J.reduce((Z,K)=>Z+K.estimatedDurationSeconds,0);return{fullText:Q,segments:J,estimatedDurationSeconds:U}}buildTimingMap(q,J,Q){let U=q.map((K)=>{let X=Math.ceil(K.durationMs/1000*J);return{sceneId:K.sceneId,durationMs:K.durationMs,durationInFrames:X,recommendedSceneDurationInFrames:Math.ceil(X*Q),wordTimings:K.wordTimings?.map(($)=>({word:$.word,startMs:$.startMs,endMs:$.endMs}))}});return{totalDurationMs:q.reduce((K,X)=>K+X.durationMs,0),segments:U,fps:J}}}function C(){let q=Date.now().toString(36),J=Math.random().toString(36).slice(2,8);return`tts_${q}_${J}`}export{I as VoiceSynthesizer};

package/dist/node/voice.feature.js

@@ -1 +0,0 @@
-import{defineFeature as g}from"@contractspec/lib.contracts-spec/features";var j=g({meta:{key:"libs.voice",version:"1.0.0",title:"Voice",description:"Voice capabilities: TTS, STT, and conversational AI",domain:"voice",owners:["@contractspec-core"],tags:["package","libs","voice"],stability:"experimental"}});export{j as VoiceFeature};

package/dist/stt/diarization-mapper.d.ts

@@ -1,19 +0,0 @@
-import type { SpeakerMap, TranscriptionSegment } from './types';
-/**
- * Map raw speaker IDs from STT provider output to human-readable labels.
- *
- * Assigns sequential labels (Speaker 1, Speaker 2, ...) based on
- * first-appearance order. Calculates per-speaker statistics.
- */
-export declare class DiarizationMapper {
-    /**
-     * Map speaker IDs to labels and compute statistics.
-     *
-     * @param segments - Transcription segments with optional speakerId
-     * @param labelPrefix - Prefix for speaker labels (default "Speaker")
-     */
-    map(segments: TranscriptionSegment[], labelPrefix?: string): {
-        segments: TranscriptionSegment[];
-        speakers: SpeakerMap[];
-    };
-}

package/dist/stt/diarization-mapper.js

@@ -1,2 +0,0 @@
-// @bun
-class D{map(z,A="Speaker"){let q=[],v=new Map;for(let h of z)if(h.speakerId&&!q.includes(h.speakerId))q.push(h.speakerId),v.set(h.speakerId,{segmentCount:0,totalSpeakingMs:0});let B=z.map((h)=>{if(!h.speakerId)return h;let w=q.indexOf(h.speakerId),j=`${A} ${w+1}`,y=v.get(h.speakerId);if(!y)return{...h,speakerLabel:j};return y.segmentCount+=1,y.totalSpeakingMs+=h.endMs-h.startMs,{...h,speakerLabel:j}}),C=q.map((h,w)=>{let j=v.get(h);return{id:h,label:`${A} ${w+1}`,segmentCount:j?.segmentCount??0,totalSpeakingMs:j?.totalSpeakingMs??0}});return{segments:B,speakers:C}}}export{D as DiarizationMapper};

package/dist/stt/index.d.ts

@@ -1,5 +0,0 @@
-export { DiarizationMapper } from './diarization-mapper';
-export { SegmentSplitter } from './segment-splitter';
-export { SubtitleFormatter } from './subtitle-formatter';
-export { Transcriber } from './transcriber';
-export type { SpeakerMap, STTBrief, STTOptions, TranscriptionProject, TranscriptionResult, TranscriptionSegment, } from './types';

package/dist/stt/index.js

@@ -1,12 +0,0 @@
-// @bun
-class W{map(T,q="Speaker"){let A=[],R=new Map;for(let w of T)if(w.speakerId&&!A.includes(w.speakerId))A.push(w.speakerId),R.set(w.speakerId,{segmentCount:0,totalSpeakingMs:0});let E=T.map((w)=>{if(!w.speakerId)return w;let J=A.indexOf(w.speakerId),K=`${q} ${J+1}`,N=R.get(w.speakerId);if(!N)return{...w,speakerLabel:K};return N.segmentCount+=1,N.totalSpeakingMs+=w.endMs-w.startMs,{...w,speakerLabel:K}}),G=A.map((w,J)=>{let K=R.get(w);return{id:w,label:`${q} ${J+1}`,segmentCount:K?.segmentCount??0,totalSpeakingMs:K?.totalSpeakingMs??0}});return{segments:E,speakers:G}}}class U{static DEFAULT_MAX_CHUNK_MS=300000;split(T,q=U.DEFAULT_MAX_CHUNK_MS){let A=T.durationMs??this.estimateDurationMs(T);if(A<=q)return[T];let R=[],E=T.data.length/Math.max(A,1),G=0;while(G<A){let w=Math.min(q,A-G),J=Math.floor(G*E),K=Math.floor((G+w)*E);R.push({data:T.data.slice(J,K),format:T.format,sampleRateHz:T.sampleRateHz,durationMs:w,channels:T.channels}),G+=w}return R}estimateDurationMs(T){let A=T.channels??1,R=T.data.length/(2*A);return Math.ceil(R/T.sampleRateHz*1000)}}class X{toSRT(T){return T.map((q,A)=>{let R=this.formatTimeSRT(q.startMs),E=this.formatTimeSRT(q.endMs),G=q.speakerLabel?`[${q.speakerLabel}] `:"";return`${A+1}
-${R} --> ${E}
-${G}${q.text}`}).join(`
-
-`)}toVTT(T){return`WEBVTT
-
-`+T.map((R,E)=>{let G=this.formatTimeVTT(R.startMs),w=this.formatTimeVTT(R.endMs),J=R.speakerLabel?`<v ${R.speakerLabel}>`:"";return`${E+1}
-${G} --> ${w}
-${J}${R.text}`}).join(`
-
-`)}formatTimeSRT(T){let q=Math.floor(T/3600000),A=Math.floor(T%3600000/60000),R=Math.floor(T%60000/1000),E=T%1000;return`${this.pad(q,2)}:${this.pad(A,2)}:${this.pad(R,2)},${this.pad(E,3)}`}formatTimeVTT(T){let q=Math.floor(T/3600000),A=Math.floor(T%3600000/60000),R=Math.floor(T%60000/1000),E=T%1000;return`${this.pad(q,2)}:${this.pad(A,2)}:${this.pad(R,2)}.${this.pad(E,3)}`}pad(T,q){return T.toString().padStart(q,"0")}}class ${stt;segmentSplitter=new U;diarizationMapper=new W;subtitleFormatter=new X;constructor(T){this.stt=T.stt}async transcribe(T){let q=z(),A=this.segmentSplitter.split(T.audio),R=[],E="",G=0,w=0;for(let Q of A){let Z=await this.stt.transcribe({audio:Q,language:T.language,diarize:T.diarize,speakerCount:T.speakerCount,wordTimestamps:!0,vocabularyHints:T.vocabularyHints}),j=Z.segments.map((L)=>({text:L.text,startMs:L.startMs+w,endMs:L.endMs+w,speakerId:L.speakerId,speakerName:L.speakerName,confidence:L.confidence}));R.push(...j),E+=(E?" ":"")+Z.text,G+=Z.durationMs,w+=Q.durationMs??0}let J=R,K;if(T.diarize){let Q=this.diarizationMapper.map(R);J=Q.segments,K=Q.speakers}let N={text:E,segments:J,language:T.language??"en",durationMs:G},Y,_=T.subtitleFormat??"none";if(_==="srt")Y=this.subtitleFormatter.toSRT(J);else if(_==="vtt")Y=this.subtitleFormatter.toVTT(J);return{id:q,transcript:N,subtitles:Y,speakers:K}}async*transcribeStream(T,q){if(!this.stt.transcribeStream)throw Error("Streaming transcription not supported by the current STT provider");let A=this.stt.transcribeStream(T,{language:q?.language,diarize:q?.diarize,speakerCount:q?.speakerCount,wordTimestamps:!0,vocabularyHints:q?.vocabularyHints});for await(let R of A)yield{text:R.text,startMs:R.startMs,endMs:R.endMs,speakerId:R.speakerId,speakerLabel:R.speakerName,confidence:R.confidence}}}function z(){let T=Date.now().toString(36),q=Math.random().toString(36).slice(2,8);return`stt_${T}_${q}`}export{$ as Transcriber,X as SubtitleFormatter,U as SegmentSplitter,W as DiarizationMapper};

package/dist/stt/segment-splitter.d.ts

@@ -1,19 +0,0 @@
-import type { AudioData } from '../types';
-/**
- * Split long audio into processable chunks.
- *
- * Useful for providers with maximum audio length limits.
- * Splits at silence boundaries when possible (approximated by byte position).
- */
-export declare class SegmentSplitter {
-    /** Default maximum chunk duration in ms (5 minutes) */
-    private static readonly DEFAULT_MAX_CHUNK_MS;
-    /**
-     * Split audio into chunks of at most maxChunkMs duration.
-     *
-     * @param audio - Input audio data
-     * @param maxChunkMs - Maximum chunk duration in milliseconds
-     */
-    split(audio: AudioData, maxChunkMs?: number): AudioData[];
-    private estimateDurationMs;
-}