@urun-sh/openai 0.5.5 → 0.6.0

package/dist/gemini-live.cjs

@@ -1,2 +1,2 @@
-"use strict";var he=Object.create;var E=Object.defineProperty;var be=Object.getOwnPropertyDescriptor;var ke=Object.getOwnPropertyNames;var _e=Object.getPrototypeOf,Oe=Object.prototype.hasOwnProperty;var Ae=(e,t)=>{for(var n in t)E(e,n,{get:t[n],enumerable:!0})},F=(e,t,n,o)=>{if(t&&typeof t=="object"||typeof t=="function")for(let r of ke(t))!Oe.call(e,r)&&r!==n&&E(e,r,{get:()=>t[r],enumerable:!(o=be(t,r))||o.enumerable});return e};var $=(e,t,n)=>(n=e!=null?he(_e(e)):{},F(t||!e||!e.__esModule?E(n,"default",{value:e,enumerable:!0}):n,e)),ve=e=>F(E({},"__esModule",{value:!0}),e);var Pe={};Ae(Pe,{AudioBridge:()=>A,CLOSE_PROTOCOL:()=>B,CLOSE_SESSION_GONE:()=>ne,CLOSE_UNKNOWN_MODEL:()=>te,CLOSE_UPSTREAM:()=>U,GEMINI_AUDIO_IN_RATE:()=>_,GEMINI_AUDIO_OUT_MIME:()=>H,GEMINI_AUDIO_OUT_RATE:()=>V,GEMINI_LIVE_PATH:()=>ee,GeminiLiveError:()=>s,NATIVE_AUDIO_SAMPLE_RATE:()=>I,OUTBOUND_AUDIO_FLUSH_B64:()=>le,RT_VIDEO_IN_STREAM:()=>S,RT_VIDEO_OUT_STREAM:()=>k,SdkTransport:()=>M,VIDEO_FRAME_MESSAGE_BUDGET:()=>R,VIDEO_OUT_DEFAULT_CODEC:()=>T,VIDEO_OUT_FRAMING:()=>P,VIDEO_OUT_SETUP_KEY:()=>b,VIDEO_OUT_WS_FRAMING:()=>C,consumeVideoOut:()=>X,decodeResponsesLane:()=>v,decodeVideoOutRecord:()=>x,enableSessionAudio:()=>W,enableSessionVideo:()=>G,encodeVideoOutRecord:()=>K,goAwayMsg:()=>de,negotiateSessionVideoOutQuic:()=>Q,nodeAudioBackend:()=>q,openSessionVideoOutLane:()=>z,parseClientMessage:()=>oe,resamplePcm16:()=>j,responseCreateToEnvelope:()=>D,responsesCreateToEnvelope:()=>N,serverContentOfAudio:()=>ge,serverContentOfDelta:()=>me,sessionResumptionUpdateMsg:()=>ue,translateClientContent:()=>ce,translateRealtimeInput:()=>fe,translateSetup:()=>ae,translateToolResponse:()=>pe,turnEndOfCompleted:()=>ye,videoOutAck:()=>ie,videoOutChunkMsg:()=>se,weriftAudioBackend:()=>J});module.exports=ve(Pe);var I=24e3,A=class{constructor(t,n){this.backend=t;this.opts=n}backend;opts;source=null;sink=null;outHandlers=new Set;async startOutbound(){return this.source=this.backend.createSource(),this.source.track}appendInputAudio(t){if(!this.source)throw new Error("startOutbound() not called");let n=Buffer.from(t,"base64"),o=Math.floor(n.byteLength/2),r=new Int16Array(o);for(let i=0;i<o;i++)r[i]=n.readInt16LE(i*2);this.source.onData(r)}async startInbound(t){if(t==null)throw new Error('startInbound: no downstream audio track \u2014 the runtime audio producer has not been consumed yet (session.stream("rt-audio-out").track is null)');this.sink=this.backend.createSink(t),this.sink.onframe=n=>{let o=Buffer.from(n.samples.buffer,n.samples.byteOffset,n.samples.byteLength).toString("base64");for(let r of this.outHandlers)r(o)}}onOutputAudio(t){return this.outHandlers.add(t),()=>this.outHandlers.delete(t)}};async function W(e,t,n=1e4){let o=new A(t,{sampleRate:24e3}),r=await o.startOutbound();await e.stream("rt-audio-in").attach(r);let i=await Ee(e,n);return await o.startInbound(i),o}function Ee(e,t){let n=e.stream("rt-audio-out");return n.track!=null?Promise.resolve(n.track):new Promise((o,r)=>{let i,a=setTimeout(()=>{i?.(),r(new Error(`enableSessionAudio: no downstream audio track within ${t}ms (runtime produced no voice OUT / SFU audio consumer never attached)`))},t),c=d=>{d!=null&&(clearTimeout(a),i?.(),o(d))};i=n.on?.("track",c),n.track!=null&&c(n.track)})}var S="rt-video-in",R=256e3;function G(e){let t=e.stream(S),n=t.emit;if(typeof n!="function")throw new Error(`enableSessionVideo: session.stream('${S}') has no emit() \u2014 the named-DATA stream produce seam (@urun-sh/core SessionStream.emit, contract \xA75) is required for the video frame lane`);return{sendInputFrame:async o=>{if(o.byteLength===0)throw new Error("enableSessionVideo: refusing to send an empty video frame");if(o.byteLength>R)throw new Error(`enableSessionVideo: frame is ${o.byteLength} bytes \u2014 over the ${R}-byte data-channel budget (the SFU's 262144-byte SCTP ceiling drops oversize messages SILENTLY; send smaller/lower-quality frames)`);await n.call(t,o)}}}var Se=24e3,O=480,Re=111;async function Ie(){let e=await import("opusscript"),t=e.default??e;return new t(Se,1,2048)}async function J(){let e=await import("werift"),t=await Ie();return{createSource:()=>{let n=new e.MediaStreamTrack({kind:"audio"}),o=Math.random()*4294967295>>>0,r=Math.random()*65535&65535,i=Math.random()*4294967295>>>0,a=new Int16Array(0);return{track:n,onData:c=>{let d=new Int16Array(a.length+c.length);d.set(a,0),d.set(c,a.length);let u=0;for(;d.length-u>=O;){let l=d.subarray(u,u+O);u+=O;let f=Buffer.from(l.buffer,l.byteOffset,l.byteLength),p=t.encode(f,O);r=r+1&65535,i=i+O>>>0;let g=new e.RtpHeader({version:2,payloadType:Re,sequenceNumber:r,timestamp:i,ssrc:o,marker:!1}),w=new e.RtpPacket(g,p);n.writeRtp(w)}a=d.slice(u)}}},createSink:n=>{let o={onframe:null};return n.onReceiveRtp.subscribe(i=>{let a;try{a=t.decode(i.payload)}catch{return}let c=new Int16Array(a.length/2);for(let d=0;d<c.length;d++)c[d]=a.readInt16LE(d*2);o.onframe?.({samples:c})}),o}}}async function q(){let e=await import("@roamhq/wrtc");return{createSource:()=>{let t=new e.nonstandard.RTCAudioSource;return{track:t.createTrack(),onData:o=>t.onData({samples:o,sampleRate:24e3})}},createSink:t=>{let n={onframe:null},o=new e.nonstandard.RTCAudioSink(t);return o.ondata=r=>{n.onframe?.(r)},n}}}var y=require("@urun-sh/core"),k="rt-video-out",P="urun.videoOut.v1",T="h264";function Y(e,t){if(!Number.isInteger(e.seq)||e.seq<0)throw new Error(`${t}: header seq must be a non-negative integer (got ${e.seq})`);if(!Number.isInteger(e.ptsUs))throw new Error(`${t}: header ptsUs must be an integer microsecond timestamp (got ${e.ptsUs})`);if(typeof e.codec!="string"||e.codec.length===0)throw new Error(`${t}: header codec must be a non-empty string`);if(typeof e.key!="boolean")throw new Error(`${t}: header key must be a boolean`)}function K(e,t){if(Y(e,"encodeVideoOutRecord"),t.byteLength===0)throw new Error("encodeVideoOutRecord: refusing to encode an empty access unit");let n=new TextEncoder().encode(JSON.stringify({v:1,seq:e.seq,ptsUs:e.ptsUs,codec:e.codec,key:e.key}));if(n.byteLength>65535)throw new Error(`encodeVideoOutRecord: header JSON is ${n.byteLength} bytes \u2014 over the u16 length prefix`);let o=new Uint8Array(2+n.byteLength+t.byteLength);return o[0]=n.byteLength&255,o[1]=n.byteLength>>8,o.set(n,2),o.set(t,2+n.byteLength),o}function x(e){if(e.byteLength<2)throw new Error(`decodeVideoOutRecord: truncated record (${e.byteLength} byte(s), no header length)`);let t=e[0]|e[1]<<8;if(e.byteLength<2+t)throw new Error(`decodeVideoOutRecord: truncated record (header says ${t} bytes, ${e.byteLength-2} present)`);let n;try{n=JSON.parse(new TextDecoder().decode(e.subarray(2,2+t)))}catch(a){throw new Error(`decodeVideoOutRecord: header is not valid JSON (${String(a)})`)}let o=n;if(o===null||typeof o!="object"||o.v!==1)throw new Error(`decodeVideoOutRecord: header is not spec v1 (v=${JSON.stringify(o?.v)}) \u2014 a newer producer must be met with a newer consumer, never silently misread`);let r=o;Y(r,"decodeVideoOutRecord");let i=e.subarray(2+t);if(i.byteLength===0)throw new Error("decodeVideoOutRecord: empty access unit \u2014 a record with no AU bytes is a producer bug");return{header:r,au:i.slice()}}async function Q(e){let t=e.wtDataEndpoint;if(typeof t!="function")throw new Error("negotiateSessionVideoOutQuic: the core session exposes no wtDataEndpoint() \u2014 the \xA75-over-QUIC dial seam (urun-infra#1608 TransportOfferResponse.webtransport, offered with wtData: true) is required for the video OUT lane; there is no WS fallback");let n=await t.call(e),o=n.sessionId??e.sessionId;if(typeof o!="string"||o.length===0)throw new Error("negotiateSessionVideoOutQuic: no session id \u2014 the WT hello requires one and neither the negotiated endpoint nor the core session carries it");return{transport:"webtransport",url:n.url,certHashes:n.certHashes,token:n.token,sessionId:o,stream:k,codec:T,framing:P}}var C="urun.videoOut.v2-ws";function z(e){let t=e.stream(k);if(typeof t.messages!="function")throw new Error(`openSessionVideoOutLane: session.stream('${k}') has no messages() \u2014 the named-DATA stream consume seam (@urun-sh/core SessionStream.messages, contract \xA75) is required for the video OUT lane`);let n=new Set,o=new Set,r=new Set,i=d=>{for(let u of o)u(d)},a=!1,c=async()=>{for await(let d of t.messages()){if(!(d instanceof Uint8Array)){i(new Error(`openSessionVideoOutLane: "${k}" delivered a non-BINARY \xA75 payload (${typeof d}) \u2014 spec v1 records ride BINARY envelopes only`));continue}let u;try{u=x(d)}catch(l){i(l instanceof Error?l:new Error(String(l)));continue}for(let l of n)l(u)}for(let d of r)d()};return{codec:T,onOutputFrame:d=>(n.add(d),a||(a=!0,c().catch(u=>i(u instanceof Error?u:new Error(String(u))))),()=>n.delete(d)),onLaneError:d=>(o.add(d),()=>o.delete(d)),onEnd:d=>(r.add(d),()=>r.delete(d))}}function Te(e){if(e.length%2!==0||/[^0-9a-fA-F]/.test(e))throw new Error(`consumeVideoOut: certHashes entry is not hex sha-256 (${JSON.stringify(e)})`);let t=new Uint8Array(e.length/2);for(let n=0;n<t.length;n++)t[n]=parseInt(e.slice(n*2,n*2+2),16);return t}async function X(e,t,n){let o=n??(typeof globalThis.WebTransport=="function"?(u,l)=>new globalThis.WebTransport(u,l):void 0);if(!o)throw new Error("consumeVideoOut: no WebTransport implementation \u2014 this platform has no globalThis.WebTransport; inject a connect factory (the third argument)");let r=o(e.url,{serverCertificateHashes:e.certHashes.map(u=>({algorithm:"sha-256",value:Te(u)}))});await r.ready;let i=await r.createBidirectionalStream();await i.writable.getWriter().write((0,y.encodeWtControlFrame)({t:"hello",token:e.token,sessionId:e.sessionId}));let a=i.readable.getReader(),c=new y.WtControlDecoder,d=!1;for(;!d;){let{value:u,done:l}=await a.read();if(l)throw new Error("consumeVideoOut: gateway closed the control stream before hello-ack");if(u)for(let f of c.push(u)){if(f.t==="hello-ack"){d=!0;break}throw new Error(`consumeVideoOut: gateway rejected hello (got ${JSON.stringify(f.t)} before hello-ack)`)}}await(0,y.consumeWtDataStreams)(r.incomingUnidirectionalStreams,{onData:(u,l)=>{if(u===e.stream){if(l===y.STREAM_END){t.onEnd?.();return}if(!(l instanceof Uint8Array)){t.onLaneError(new Error(`consumeVideoOut: "${e.stream}" delivered a non-BINARY \xA75 payload (${typeof l}) \u2014 spec v1 records ride BINARY envelopes only`));return}try{t.onFrame(x(l))}catch(f){t.onLaneError(f instanceof Error?f:new Error(String(f)))}}},onLaneError:u=>t.onLaneError(u)})}var ee="/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent",B=1007,te=1008,U=1011,ne=1008,s=class extends Error{constructor(n,o=B){super(n);this.closeCode=o}closeCode},m=e=>typeof e=="object"&&e!==null&&!Array.isArray(e)?e:null,L=["setup","clientContent","realtimeInput","toolResponse"];function oe(e){let t;try{t=JSON.parse(e)}catch{throw new s("client message is not valid JSON")}let n=m(t);if(!n)throw new s("client message must be a JSON object");let o=L.filter(c=>n[c]!==void 0),r=Object.keys(n).filter(c=>!L.includes(c));if(r.length>0)throw new s(`unknown client message field(s) ${JSON.stringify(r)} \u2014 expected exactly one of ${L.join(", ")}`);if(o.length!==1)throw new s(`client message must contain exactly one of ${L.join(", ")} (got ${o.length})`);let i=o[0],a=m(n[i]);if(!a)throw new s(`"${i}" must be a JSON object`);return{kind:i,[i]:a}}function re(e,t){if(typeof e=="string")return e;let n=m(e),o=n?n.parts:void 0;if(!Array.isArray(o))throw new s(`${t}: expected a Content with a parts array`);let r=[];for(let i of o){let a=m(i);if(a&&typeof a.text=="string"){r.push(a.text);continue}throw a&&(a.inlineData!==void 0||a.fileData!==void 0)?new s(`${t}: media parts (inlineData/fileData) are a declared follow-up slice (audio/video lanes over the proxy transport AudioBridge) \u2014 only text parts are supported in this slice`):a&&(a.functionCall!==void 0||a.functionResponse!==void 0)?new s(`${t}: functionCall/functionResponse parts are not valid here \u2014 the Live API exchanges tool traffic via the dedicated toolCall/toolResponse messages`):new s(`${t}: unsupported part ${JSON.stringify(i)}`)}return r.join(`
-`)}var xe={manualActivity:!1,activityHandling:"START_OF_ACTIVITY_INTERRUPTS"},b="urun.videoOut";function ie(e){return{framing:C,codec:e}}function se(e){let{seq:t,ptsUs:n,codec:o,key:r}=e.header;return{[b]:{v:1,seq:t,ptsUs:n,codec:o,key:r,data:Buffer.from(e.au).toString("base64")}}}var Ce={inputAudioTranscription:"setup.inputAudioTranscription is part of the declared audio follow-up slice",outputAudioTranscription:"setup.outputAudioTranscription is part of the declared audio follow-up slice"},Le=["model","generationConfig","systemInstruction","tools","realtimeInputConfig","sessionResumption",b],Z=["START_OF_ACTIVITY_INTERRUPTS","NO_INTERRUPTION"];function Ue(e){let t=m(e);if(!t)throw new s("setup.realtimeInputConfig must be an object");for(let r of Object.keys(t))if(!(r==="automaticActivityDetection"||r==="activityHandling"))throw new s(r==="turnCoverage"?"realtimeInputConfig.turnCoverage is part of the declared audio follow-up slice":`unsupported realtimeInputConfig field "${r}"`);let n=m(t.automaticActivityDetection);if(t.automaticActivityDetection!==void 0&&!n)throw new s("realtimeInputConfig.automaticActivityDetection must be an object");for(let r of Object.keys(n??{}))if(r!=="disabled")throw new s(`automaticActivityDetection.${r} tunes AUTOMATIC (server-side) activity detection \u2014 the declared platform follow-up (serve turn_detection events over the session stream lane); only "disabled: true" (manual activity mode) is supported in this slice`);if(n?.disabled!==!0)throw new s("automatic (server-side) activity detection is the declared platform follow-up (serve turn_detection turn_start/turn_end events over the session stream lane, not proxy DSP) \u2014 set realtimeInputConfig.automaticActivityDetection.disabled: true for manual activity mode");let o="START_OF_ACTIVITY_INTERRUPTS";if(t.activityHandling!==void 0){if(!Z.includes(t.activityHandling))throw new s(`unsupported activityHandling ${JSON.stringify(t.activityHandling)} \u2014 expected one of ${Z.join(", ")}`);o=t.activityHandling}return{manualActivity:!0,activityHandling:o}}var Ve=["temperature","maxOutputTokens","responseModalities"];function ae(e){for(let[f,p]of Object.entries(Ce))if(e[f]!==void 0)throw new s(p);for(let f of Object.keys(e))if(!Le.includes(f))throw new s(`unsupported setup field "${f}"`);if(typeof e.model!="string"||e.model.length===0)throw new s('setup.model is required (format "models/{model}")');let t=e.model.replace(/^models\//,""),n,o,r=!1,i=m(e.generationConfig);if(e.generationConfig!==void 0&&!i)throw new s("setup.generationConfig must be an object");if(i){for(let f of Object.keys(i))if(!Ve.includes(f))throw new s(`unsupported generationConfig field "${f}"`);if(i.temperature!==void 0){if(typeof i.temperature!="number")throw new s("generationConfig.temperature must be a number");n=i.temperature}if(i.maxOutputTokens!==void 0){if(typeof i.maxOutputTokens!="number")throw new s("generationConfig.maxOutputTokens must be a number");o=i.maxOutputTokens}if(i.responseModalities!==void 0){let f=Array.isArray(i.responseModalities)?i.responseModalities:[i.responseModalities];for(let p of f){let g=String(p).toUpperCase();if(g==="AUDIO"){r=!0;continue}if(g!=="TEXT")throw new s(`responseModalities ${JSON.stringify(p)} is not supported \u2014 this surface speaks TEXT and AUDIO (audio-out rides the native AudioBridge rt-audio-out lane)`)}}}let a=[];e.systemInstruction!==void 0&&a.push({role:"system",content:re(e.systemInstruction,"setup.systemInstruction")});let c;if(e.tools!==void 0){if(!Array.isArray(e.tools))throw new s("setup.tools must be an array");c=[];for(let f of e.tools){let p=m(f);if(!p)throw new s("setup.tools entries must be objects");for(let g of Object.keys(p))if(g!=="functionDeclarations")throw new s(`unsupported tool "${g}" \u2014 this slice translates functionDeclarations only (no googleSearch/codeExecution)`);if(!Array.isArray(p.functionDeclarations))throw new s("tool.functionDeclarations must be an array");for(let g of p.functionDeclarations){let w=m(g);if(!w||typeof w.name!="string")throw new s("functionDeclarations entries require a string name");c.push({type:"function",name:w.name,description:w.description,parameters:w.parameters})}}c.length===0&&(c=void 0)}let d={enabled:!1,handle:null};if(e.sessionResumption!==void 0){let f=m(e.sessionResumption);if(!f)throw new s("setup.sessionResumption must be an object (SessionResumptionConfig)");for(let p of Object.keys(f))if(p!=="handle")throw new s(`unsupported sessionResumption field "${p}" \u2014 this lane implements SessionResumptionConfig.handle only`);if(f.handle!==void 0&&(typeof f.handle!="string"||f.handle.length===0))throw new s("sessionResumption.handle must be a non-empty string when present");d={enabled:!0,handle:typeof f.handle=="string"?f.handle:null}}let u=e.realtimeInputConfig===void 0?xe:Ue(e.realtimeInputConfig),l=!1;if(e[b]!==void 0){if(!m(e[b]))throw new s(`setup["${b}"] must be an object (the v1 opt-in is {})`);l=!0}return{model:t,systemItems:a,tools:c,temperature:n,maxOutputTokens:o,audioModality:r,activity:u,sessionResumption:d,videoOut:l}}function ue(e){return{sessionResumptionUpdate:{newHandle:e,resumable:!0}}}function de(e){let t=Math.max(0,e??0)/1e3;return{goAway:{timeLeft:`${Math.round(t*1e3)/1e3}s`}}}function ce(e){for(let o of Object.keys(e))if(o!=="turns"&&o!=="turnComplete")throw new s(`unsupported clientContent field "${o}"`);let t=e.turns===void 0?[]:e.turns;if(!Array.isArray(t))throw new s("clientContent.turns must be an array");let n=[];for(let o of t){let r=m(o);if(!r)throw new s("clientContent.turns entries must be Content objects");let i=r.role===void 0?"user":String(r.role).toLowerCase();if(i!=="user"&&i!=="model")throw new s(`clientContent turn role must be "user" or "model" (got ${JSON.stringify(r.role)})`);n.push({role:i==="model"?"assistant":"user",content:re(r,"clientContent.turns")})}return{items:n,turnComplete:e.turnComplete===!0}}var _=16e3,V=24e3,H=`audio/pcm;rate=${V}`,le=Buffer.alloc(960).toString("base64");function j(e,t,n){if(t===n)return e;let o=Math.floor(e.length*n/t),r=new Int16Array(o),i=t/n,a=e.length-1;for(let c=0;c<o;c++){let d=c*i,u=Math.floor(d);if(u>=a){r[c]=e[a];continue}let l=d-u;r[c]=Math.round(e[u]+(e[u+1]-e[u])*l)}return r}function De(e){let t=m(e);if(!t)throw new s("realtimeInput.video must be a Blob object {mimeType, data}");if(typeof t.mimeType!="string"||!/^image\/jpeg\s*(;|$)/.test(t.mimeType))throw new s(`realtimeInput.video.mimeType must be image/jpeg (the Live video input stream is individual JPEG frames), got ${JSON.stringify(t.mimeType)}`);if(typeof t.data!="string"||t.data.length===0)throw new s("realtimeInput.video.data must be non-empty base64");let n=Buffer.from(t.data,"base64");if(n.length<3||n[0]!==255||n[1]!==216)throw new s("realtimeInput.video.data does not decode to a JPEG frame (missing FF D8 SOI marker)");return new Uint8Array(n)}function Ne(e){let t=m(e);if(!t)throw new s("realtimeInput.audio must be a Blob object {mimeType, data}");if(typeof t.mimeType!="string"||!/^audio\/pcm\s*(;|$)/.test(t.mimeType))throw new s(`realtimeInput.audio.mimeType must be audio/pcm (16-bit LE PCM @16kHz mono per the Live spec), got ${JSON.stringify(t.mimeType)}`);let n=/;\s*rate=(\d+)/.exec(t.mimeType);if(n&&Number(n[1])!==_)throw new s(`realtimeInput.audio.mimeType declares rate=${n[1]} \u2014 the Live spec input format is ${_} Hz (audio/pcm;rate=${_})`);if(typeof t.data!="string"||t.data.length===0)throw new s("realtimeInput.audio.data must be non-empty base64");let o=Buffer.from(t.data,"base64");if(o.length===0||o.length%2!==0)throw new s(`realtimeInput.audio.data must decode to whole 16-bit samples (got ${o.length} bytes \u2014 not even)`);let r=new Int16Array(o.length/2);for(let a=0;a<r.length;a++)r[a]=o.readInt16LE(a*2);let i=j(r,_,V);return Buffer.from(i.buffer,i.byteOffset,i.byteLength).toString("base64")}function fe(e){if(e.mediaChunks!==void 0)throw new s('realtimeInput.mediaChunks is deprecated by the Live API \u2014 send the audio stream via realtimeInput.audio ({mimeType:"audio/pcm;rate=16000", data}) instead');if(e.activityStart!==void 0||e.activityEnd!==void 0){if(Object.keys(e).length!==1||e.activityStart!==void 0&&e.activityEnd!==void 0)throw new s("realtimeInput must carry exactly one field when signaling activity (activityStart XOR activityEnd, nothing else)");let r=e.activityStart!==void 0?"activityStart":"activityEnd",i=m(e[r]);if(!i||Object.keys(i).length!==0)throw new s(`realtimeInput.${r} must be an empty object (ActivityStart/ActivityEnd are bare markers)`);return{kind:r}}let t=["text","audio","audioStreamEnd","video"];for(let o of Object.keys(e))if(!t.includes(o))throw new s(`unsupported realtimeInput field "${o}"`);let n=t.filter(o=>e[o]!==void 0);if(n.length!==1)throw new s(`realtimeInput must carry exactly one of ${t.join(", ")} (got ${n.length})`);if(e.audio!==void 0)return{kind:"audio",base64Pcm24k:Ne(e.audio)};if(e.video!==void 0)return{kind:"video",jpegFrame:De(e.video)};if(e.audioStreamEnd!==void 0){if(e.audioStreamEnd!==!0)throw new s("realtimeInput.audioStreamEnd must be true when present");return{kind:"audioStreamEnd"}}if(typeof e.text!="string"||e.text.length===0)throw new s("realtimeInput.text must be a non-empty string");return{kind:"text",text:e.text}}function pe(e){for(let t of Object.keys(e))if(t!=="functionResponses")throw new s(`unsupported toolResponse field "${t}"`);if(!Array.isArray(e.functionResponses)||e.functionResponses.length===0)throw new s("toolResponse.functionResponses must be a non-empty array");return e.functionResponses.map(t=>{let n=m(t);if(!n||typeof n.id!="string"||n.id.length===0)throw new s("functionResponses entries require the string id matched from the toolCall");return{type:"function_call_output",call_id:n.id,output:JSON.stringify(n.response??null)}})}function me(e){return e.type!=="response.output_text.delta"||typeof e.delta!="string"?null:{serverContent:{modelTurn:{role:"model",parts:[{text:e.delta}]}}}}function ge(e){return{serverContent:{modelTurn:{role:"model",parts:[{inlineData:{mimeType:H,data:e}}]}}}}function ye(e){let t=m(e)??{},n=Array.isArray(t.output)?t.output:[],o=[],r=[],i=typeof t.output_text=="string"?t.output_text:"";if(!i)for(let u of n){let l=m(u);if(!(l?.type!=="message"||!Array.isArray(l.content)))for(let f of l.content){let p=m(f);p?.type==="output_text"&&typeof p.text=="string"&&(i+=p.text)}}i&&r.push({role:"assistant",content:i});for(let u of n){let l=m(u);if(l?.type!=="function_call")continue;if(typeof l.call_id!="string"||typeof l.name!="string")throw new s(`upstream function_call item is missing call_id/name: ${JSON.stringify(u)}`,U);let f=typeof l.arguments=="string"&&l.arguments.length>0?l.arguments:"{}",p;try{p=JSON.parse(f)}catch{throw new s(`upstream function_call arguments are not valid JSON: ${f}`,U)}o.push({id:l.call_id,name:l.name,args:p}),r.push({type:"function_call",call_id:l.call_id,name:l.name,arguments:f})}let a=[];o.length>0&&a.push({toolCall:{functionCalls:o}}),a.push({serverContent:{generationComplete:!0}});let c=m(t.usage),d={serverContent:{turnComplete:!0}};if(c){let u={};typeof c.input_tokens=="number"&&(u.promptTokenCount=c.input_tokens),typeof c.output_tokens=="number"&&(u.responseTokenCount=c.output_tokens),typeof c.total_tokens=="number"&&(u.totalTokenCount=c.total_tokens),Object.keys(u).length>0&&(d.usageMetadata=u)}return a.push(d),{messages:a,assistantItems:r}}function D(e,t,n,o){let r=e.response??{},i={request_id:n,consumer_id:o,stream:!0,kind:"chat",messages:t};return typeof r.instructions=="string"&&(i.instructions=r.instructions),Array.isArray(r.modalities)&&(i.modalities=r.modalities),typeof r.temperature=="number"&&(i.temperature=r.temperature),typeof r.max_output_tokens=="number"&&(i.max_output_tokens=r.max_output_tokens),r.tools!==void 0&&(i.tools=r.tools),i}function N(e,t,n){let o={request_id:t,consumer_id:n,stream:!!e.stream,kind:"responses",input:e.input};return e.model&&(o.model=e.model),e.tools!==void 0&&(o.tools=e.tools),e.tool_choice!==void 0&&(o.tool_choice=e.tool_choice),typeof e.temperature=="number"&&(o.temperature=e.temperature),typeof e.max_output_tokens=="number"&&(o.max_output_tokens=e.max_output_tokens),typeof e.top_p=="number"&&(o.top_p=e.top_p),Array.isArray(e.stop)&&e.stop.length>0&&(o.stop=e.stop),typeof e.reasoning_effort=="string"&&(o.reasoning_effort=e.reasoning_effort),e.chat_template_kwargs!==void 0&&(o.chat_template_kwargs=e.chat_template_kwargs),o}function we(e){if(e instanceof Error)return{type:"error",error:{type:"urun_error",code:e.name||null,message:e.message}};if(e&&typeof e=="object"&&e.t==="error"){let t=e,n=t.body??{};return{type:"error",error:{type:"urun_error",code:t.code??null,message:n.message??"unknown error"}}}return{type:"error",error:{type:"urun_error",code:null,message:String(e)}}}var Me="llm-resp";async function*v(e,t){let n=`${Me}:${t}`,o=`resp_${t}`;yield{type:"response.created",response:{id:o,status:"in_progress"}};let r=new Map;for await(let i of e.stream(n).messages()){let a=i;if(a.t==="delta"){if(typeof a.delta=="string"&&(yield{type:"response.output_text.delta",item_id:o,delta:a.delta}),typeof a.reasoning=="string"&&(yield{type:"response.reasoning_text.delta",item_id:o,delta:a.reasoning}),Array.isArray(a.tool_calls))for(let c of a.tool_calls){let d=typeof c.index=="number"?c.index:0,u=r.get(d)??{};c.id&&(u.id=c.id),c.function?.name&&(u.name=c.function.name),r.set(d,u),yield{type:"response.function_call_arguments.delta",item_id:`fc_${t}_${d}`,tool_index:d,call_id:u.id,name:u.name,delta:c.function?.arguments??""}}}else if(a.t==="response"){yield{type:"response.completed",response:{id:o,status:"completed",...a.body}};return}else if(a.t==="error"){yield we(a);return}}}var $e="llm",M=class{constructor(t){this.session=t;this.sessionTag=t.sessionId??globalThis.crypto.randomUUID()}session;sessionTag;get consumerId(){return this.session.consumerId}write(t){t.session_tag=this.sessionTag,this.session.doc($e).set({requests:{[t.request_id]:{payload:t,consumer_id:t.consumer_id,stream:t.stream}}})}sendResponses(t,n){let o=N(t,n,this.consumerId);return this.write(o),v(this.session,n)}sendResponseCreate(t,n,o){let r=D(t,n,o,this.consumerId);return this.write(r),v(this.session,o)}};0&&(module.exports={AudioBridge,CLOSE_PROTOCOL,CLOSE_SESSION_GONE,CLOSE_UNKNOWN_MODEL,CLOSE_UPSTREAM,GEMINI_AUDIO_IN_RATE,GEMINI_AUDIO_OUT_MIME,GEMINI_AUDIO_OUT_RATE,GEMINI_LIVE_PATH,GeminiLiveError,NATIVE_AUDIO_SAMPLE_RATE,OUTBOUND_AUDIO_FLUSH_B64,RT_VIDEO_IN_STREAM,RT_VIDEO_OUT_STREAM,SdkTransport,VIDEO_FRAME_MESSAGE_BUDGET,VIDEO_OUT_DEFAULT_CODEC,VIDEO_OUT_FRAMING,VIDEO_OUT_SETUP_KEY,VIDEO_OUT_WS_FRAMING,consumeVideoOut,decodeResponsesLane,decodeVideoOutRecord,enableSessionAudio,enableSessionVideo,encodeVideoOutRecord,goAwayMsg,negotiateSessionVideoOutQuic,nodeAudioBackend,openSessionVideoOutLane,parseClientMessage,resamplePcm16,responseCreateToEnvelope,responsesCreateToEnvelope,serverContentOfAudio,serverContentOfDelta,sessionResumptionUpdateMsg,translateClientContent,translateRealtimeInput,translateSetup,translateToolResponse,turnEndOfCompleted,videoOutAck,videoOutChunkMsg,weriftAudioBackend});
+"use strict";var _e=Object.create;var E=Object.defineProperty;var be=Object.getOwnPropertyDescriptor;var ve=Object.getOwnPropertyNames;var Oe=Object.getPrototypeOf,Ae=Object.prototype.hasOwnProperty;var Ee=(e,t)=>{for(var n in t)E(e,n,{get:t[n],enumerable:!0})},W=(e,t,n,o)=>{if(t&&typeof t=="object"||typeof t=="function")for(let r of ve(t))!Ae.call(e,r)&&r!==n&&E(e,r,{get:()=>t[r],enumerable:!(o=be(t,r))||o.enumerable});return e};var $=(e,t,n)=>(n=e!=null?_e(Oe(e)):{},W(t||!e||!e.__esModule?E(n,"default",{value:e,enumerable:!0}):n,e)),Se=e=>W(E({},"__esModule",{value:!0}),e);var Be={};Ee(Be,{AudioBridge:()=>O,CLOSE_PROTOCOL:()=>H,CLOSE_SESSION_GONE:()=>re,CLOSE_UNKNOWN_MODEL:()=>oe,CLOSE_UPSTREAM:()=>U,GEMINI_AUDIO_IN_RATE:()=>b,GEMINI_AUDIO_OUT_MIME:()=>F,GEMINI_AUDIO_OUT_RATE:()=>D,GEMINI_LIVE_PATH:()=>ne,GeminiLiveError:()=>s,NATIVE_AUDIO_SAMPLE_RATE:()=>R,OUTBOUND_AUDIO_FLUSH_B64:()=>pe,RT_VIDEO_IN_STREAM:()=>S,RT_VIDEO_OUT_STREAM:()=>_,SdkTransport:()=>P,VIDEO_FRAME_MESSAGE_BUDGET:()=>T,VIDEO_OUT_DEFAULT_CODEC:()=>I,VIDEO_OUT_FRAMING:()=>B,VIDEO_OUT_SETUP_KEY:()=>k,VIDEO_OUT_WS_FRAMING:()=>C,consumeVideoOut:()=>ee,decodeResponsesLane:()=>A,decodeVideoOutRecord:()=>x,enableSessionAudio:()=>J,enableSessionVideo:()=>q,encodeVideoOutRecord:()=>z,goAwayMsg:()=>le,negotiateSessionVideoOutQuic:()=>X,nodeAudioBackend:()=>K,openSessionVideoOutLane:()=>Z,parseClientMessage:()=>ie,resamplePcm16:()=>j,responseCreateToEnvelope:()=>V,responsesCreateToEnvelope:()=>M,serverContentOfAudio:()=>we,serverContentOfDelta:()=>ye,sessionResumptionUpdateMsg:()=>ce,translateClientContent:()=>fe,translateRealtimeInput:()=>me,translateSetup:()=>de,translateToolResponse:()=>ge,turnEndOfCompleted:()=>he,videoOutAck:()=>ae,videoOutChunkMsg:()=>ue,weriftAudioBackend:()=>Y});module.exports=Se(Be);var R=24e3,G="audio",O=class{constructor(t,n){this.backend=t;this.opts=n}backend;opts;source=null;sink=null;outHandlers=new Set;async startOutbound(){return this.source=this.backend.createSource(),this.source.track}appendInputAudio(t){if(!this.source)throw new Error("startOutbound() not called");let n=Buffer.from(t,"base64"),o=Math.floor(n.byteLength/2),r=new Int16Array(o);for(let i=0;i<o;i++)r[i]=n.readInt16LE(i*2);this.source.onData(r)}async startInbound(t){if(t==null)throw new Error('startInbound: no downstream audio track \u2014 the runtime audio producer has not been consumed yet (session.stream("audio").track is null)');this.sink=this.backend.createSink(t),this.sink.onframe=n=>{let o=Buffer.from(n.samples.buffer,n.samples.byteOffset,n.samples.byteLength).toString("base64");for(let r of this.outHandlers)r(o)}}onOutputAudio(t){return this.outHandlers.add(t),()=>this.outHandlers.delete(t)}};async function J(e,t,n=1e4){let o=new O(t,{sampleRate:24e3}),r=await o.startOutbound();await e.stream(G).attach(r);let i=await Te(e,n);return await o.startInbound(i),o}function Te(e,t){let n=e.stream(G);return n.track!=null?Promise.resolve(n.track):new Promise((o,r)=>{let i,a=setTimeout(()=>{i?.(),r(new Error(`enableSessionAudio: no downstream audio track within ${t}ms (runtime produced no voice OUT / SFU audio consumer never attached)`))},t),c=d=>{d!=null&&(clearTimeout(a),i?.(),o(d))};i=n.on?.("track",c),n.track!=null&&c(n.track)})}var S="rt-video-in",T=256e3;function q(e){let t=e.stream(S),n=t.emit;if(typeof n!="function")throw new Error(`enableSessionVideo: session.stream('${S}') has no emit() \u2014 the named-DATA stream produce seam (@urun-sh/core SessionStream.emit, contract \xA75) is required for the video frame lane`);return{sendInputFrame:async o=>{if(o.byteLength===0)throw new Error("enableSessionVideo: refusing to send an empty video frame");if(o.byteLength>T)throw new Error(`enableSessionVideo: frame is ${o.byteLength} bytes \u2014 over the ${T}-byte data-channel budget (the SFU's 262144-byte SCTP ceiling drops oversize messages SILENTLY; send smaller/lower-quality frames)`);await n.call(t,o)}}}var Re=24e3,v=480,Ie=111;async function xe(){let e=await import("opusscript"),t=e.default??e;return new t(Re,1,2048)}async function Y(){let e=await import("werift"),t=await xe();return{createSource:()=>{let n=new e.MediaStreamTrack({kind:"audio"}),o=Math.random()*4294967295>>>0,r=Math.random()*65535&65535,i=Math.random()*4294967295>>>0,a=new Int16Array(0);return{track:n,onData:c=>{let d=new Int16Array(a.length+c.length);d.set(a,0),d.set(c,a.length);let u=0;for(;d.length-u>=v;){let l=d.subarray(u,u+v);u+=v;let f=Buffer.from(l.buffer,l.byteOffset,l.byteLength),p=t.encode(f,v);r=r+1&65535,i=i+v>>>0;let g=new e.RtpHeader({version:2,payloadType:Ie,sequenceNumber:r,timestamp:i,ssrc:o,marker:!1}),h=new e.RtpPacket(g,p);n.writeRtp(h)}a=d.slice(u)}}},createSink:n=>{let o={onframe:null};return n.onReceiveRtp.subscribe(i=>{let a;try{a=t.decode(i.payload)}catch{return}let c=new Int16Array(a.length/2);for(let d=0;d<c.length;d++)c[d]=a.readInt16LE(d*2);o.onframe?.({samples:c})}),o}}}async function K(){let e=await import("@roamhq/wrtc");return{createSource:()=>{let t=new e.nonstandard.RTCAudioSource;return{track:t.createTrack(),onData:o=>t.onData({samples:o,sampleRate:24e3})}},createSink:t=>{let n={onframe:null},o=new e.nonstandard.RTCAudioSink(t);return o.ondata=r=>{n.onframe?.(r)},n}}}var w=require("@urun-sh/core"),_="rt-video-out",B="urun.videoOut.v1",I="h264";function Q(e,t){if(!Number.isInteger(e.seq)||e.seq<0)throw new Error(`${t}: header seq must be a non-negative integer (got ${e.seq})`);if(!Number.isInteger(e.ptsUs))throw new Error(`${t}: header ptsUs must be an integer microsecond timestamp (got ${e.ptsUs})`);if(typeof e.codec!="string"||e.codec.length===0)throw new Error(`${t}: header codec must be a non-empty string`);if(typeof e.key!="boolean")throw new Error(`${t}: header key must be a boolean`)}function z(e,t){if(Q(e,"encodeVideoOutRecord"),t.byteLength===0)throw new Error("encodeVideoOutRecord: refusing to encode an empty access unit");let n=new TextEncoder().encode(JSON.stringify({v:1,seq:e.seq,ptsUs:e.ptsUs,codec:e.codec,key:e.key}));if(n.byteLength>65535)throw new Error(`encodeVideoOutRecord: header JSON is ${n.byteLength} bytes \u2014 over the u16 length prefix`);let o=new Uint8Array(2+n.byteLength+t.byteLength);return o[0]=n.byteLength&255,o[1]=n.byteLength>>8,o.set(n,2),o.set(t,2+n.byteLength),o}function x(e){if(e.byteLength<2)throw new Error(`decodeVideoOutRecord: truncated record (${e.byteLength} byte(s), no header length)`);let t=e[0]|e[1]<<8;if(e.byteLength<2+t)throw new Error(`decodeVideoOutRecord: truncated record (header says ${t} bytes, ${e.byteLength-2} present)`);let n;try{n=JSON.parse(new TextDecoder().decode(e.subarray(2,2+t)))}catch(a){throw new Error(`decodeVideoOutRecord: header is not valid JSON (${String(a)})`)}let o=n;if(o===null||typeof o!="object"||o.v!==1)throw new Error(`decodeVideoOutRecord: header is not spec v1 (v=${JSON.stringify(o?.v)}) \u2014 a newer producer must be met with a newer consumer, never silently misread`);let r=o;Q(r,"decodeVideoOutRecord");let i=e.subarray(2+t);if(i.byteLength===0)throw new Error("decodeVideoOutRecord: empty access unit \u2014 a record with no AU bytes is a producer bug");return{header:r,au:i.slice()}}async function X(e){let t=e.wtDataEndpoint;if(typeof t!="function")throw new Error("negotiateSessionVideoOutQuic: the core session exposes no wtDataEndpoint() \u2014 the \xA75-over-QUIC dial seam (urun-infra#1608 TransportOfferResponse.webtransport, offered with wtData: true) is required for the video OUT lane; there is no WS fallback");let n=await t.call(e),o=n.sessionId??e.sessionId;if(typeof o!="string"||o.length===0)throw new Error("negotiateSessionVideoOutQuic: no session id \u2014 the WT hello requires one and neither the negotiated endpoint nor the core session carries it");return{transport:"webtransport",url:n.url,certHashes:n.certHashes,token:n.token,sessionId:o,stream:_,codec:I,framing:B}}var C="urun.videoOut.v2-ws";function Z(e){let t=e.stream(_);if(typeof t.messages!="function")throw new Error(`openSessionVideoOutLane: session.stream('${_}') has no messages() \u2014 the named-DATA stream consume seam (@urun-sh/core SessionStream.messages, contract \xA75) is required for the video OUT lane`);let n=new Set,o=new Set,r=new Set,i=d=>{for(let u of o)u(d)},a=!1,c=async()=>{for await(let d of t.messages()){if(!(d instanceof Uint8Array)){i(new Error(`openSessionVideoOutLane: "${_}" delivered a non-BINARY \xA75 payload (${typeof d}) \u2014 spec v1 records ride BINARY envelopes only`));continue}let u;try{u=x(d)}catch(l){i(l instanceof Error?l:new Error(String(l)));continue}for(let l of n)l(u)}for(let d of r)d()};return{codec:I,onOutputFrame:d=>(n.add(d),a||(a=!0,c().catch(u=>i(u instanceof Error?u:new Error(String(u))))),()=>n.delete(d)),onLaneError:d=>(o.add(d),()=>o.delete(d)),onEnd:d=>(r.add(d),()=>r.delete(d))}}function Ce(e){if(e.length%2!==0||/[^0-9a-fA-F]/.test(e))throw new Error(`consumeVideoOut: certHashes entry is not hex sha-256 (${JSON.stringify(e)})`);let t=new Uint8Array(e.length/2);for(let n=0;n<t.length;n++)t[n]=parseInt(e.slice(n*2,n*2+2),16);return t}async function ee(e,t,n){let o=n??(typeof globalThis.WebTransport=="function"?(u,l)=>new globalThis.WebTransport(u,l):void 0);if(!o)throw new Error("consumeVideoOut: no WebTransport implementation \u2014 this platform has no globalThis.WebTransport; inject a connect factory (the third argument)");let r=o(e.url,{serverCertificateHashes:e.certHashes.map(u=>({algorithm:"sha-256",value:Ce(u)}))});await r.ready;let i=await r.createBidirectionalStream();await i.writable.getWriter().write((0,w.encodeWtControlFrame)({t:"hello",token:e.token,sessionId:e.sessionId}));let a=i.readable.getReader(),c=new w.WtControlDecoder,d=!1;for(;!d;){let{value:u,done:l}=await a.read();if(l)throw new Error("consumeVideoOut: gateway closed the control stream before hello-ack");if(u)for(let f of c.push(u)){if(f.t==="hello-ack"){d=!0;break}throw new Error(`consumeVideoOut: gateway rejected hello (got ${JSON.stringify(f.t)} before hello-ack)`)}}await(0,w.consumeWtDataStreams)(r.incomingUnidirectionalStreams,{onData:(u,l)=>{if(u===e.stream){if(l===w.STREAM_END){t.onEnd?.();return}if(!(l instanceof Uint8Array)){t.onLaneError(new Error(`consumeVideoOut: "${e.stream}" delivered a non-BINARY \xA75 payload (${typeof l}) \u2014 spec v1 records ride BINARY envelopes only`));return}try{t.onFrame(x(l))}catch(f){t.onLaneError(f instanceof Error?f:new Error(String(f)))}}},onLaneError:u=>t.onLaneError(u)})}var ne="/ws/google.ai.generativelanguage.v1beta.GenerativeService.BidiGenerateContent",H=1007,oe=1008,U=1011,re=1008,s=class extends Error{constructor(n,o=H){super(n);this.closeCode=o}closeCode},m=e=>typeof e=="object"&&e!==null&&!Array.isArray(e)?e:null,L=["setup","clientContent","realtimeInput","toolResponse"];function ie(e){let t;try{t=JSON.parse(e)}catch{throw new s("client message is not valid JSON")}let n=m(t);if(!n)throw new s("client message must be a JSON object");let o=L.filter(c=>n[c]!==void 0),r=Object.keys(n).filter(c=>!L.includes(c));if(r.length>0)throw new s(`unknown client message field(s) ${JSON.stringify(r)} \u2014 expected exactly one of ${L.join(", ")}`);if(o.length!==1)throw new s(`client message must contain exactly one of ${L.join(", ")} (got ${o.length})`);let i=o[0],a=m(n[i]);if(!a)throw new s(`"${i}" must be a JSON object`);return{kind:i,[i]:a}}function se(e,t){if(typeof e=="string")return e;let n=m(e),o=n?n.parts:void 0;if(!Array.isArray(o))throw new s(`${t}: expected a Content with a parts array`);let r=[];for(let i of o){let a=m(i);if(a&&typeof a.text=="string"){r.push(a.text);continue}throw a&&(a.inlineData!==void 0||a.fileData!==void 0)?new s(`${t}: media parts (inlineData/fileData) are a declared follow-up slice (audio/video lanes over the proxy transport AudioBridge) \u2014 only text parts are supported in this slice`):a&&(a.functionCall!==void 0||a.functionResponse!==void 0)?new s(`${t}: functionCall/functionResponse parts are not valid here \u2014 the Live API exchanges tool traffic via the dedicated toolCall/toolResponse messages`):new s(`${t}: unsupported part ${JSON.stringify(i)}`)}return r.join(`
+`)}var Le={manualActivity:!1,activityHandling:"START_OF_ACTIVITY_INTERRUPTS"},k="urun.videoOut";function ae(e){return{framing:C,codec:e}}function ue(e){let{seq:t,ptsUs:n,codec:o,key:r}=e.header;return{[k]:{v:1,seq:t,ptsUs:n,codec:o,key:r,data:Buffer.from(e.au).toString("base64")}}}var Ue={inputAudioTranscription:"setup.inputAudioTranscription is part of the declared audio follow-up slice",outputAudioTranscription:"setup.outputAudioTranscription is part of the declared audio follow-up slice"},De=["model","generationConfig","systemInstruction","tools","realtimeInputConfig","sessionResumption",k],te=["START_OF_ACTIVITY_INTERRUPTS","NO_INTERRUPTION"];function Ve(e){let t=m(e);if(!t)throw new s("setup.realtimeInputConfig must be an object");for(let r of Object.keys(t))if(!(r==="automaticActivityDetection"||r==="activityHandling"))throw new s(r==="turnCoverage"?"realtimeInputConfig.turnCoverage is part of the declared audio follow-up slice":`unsupported realtimeInputConfig field "${r}"`);let n=m(t.automaticActivityDetection);if(t.automaticActivityDetection!==void 0&&!n)throw new s("realtimeInputConfig.automaticActivityDetection must be an object");for(let r of Object.keys(n??{}))if(r!=="disabled")throw new s(`automaticActivityDetection.${r} tunes AUTOMATIC (server-side) activity detection \u2014 the declared platform follow-up (serve turn_detection events over the session stream lane); only "disabled: true" (manual activity mode) is supported in this slice`);if(n?.disabled!==!0)throw new s("automatic (server-side) activity detection is the declared platform follow-up (serve turn_detection turn_start/turn_end events over the session stream lane, not proxy DSP) \u2014 set realtimeInputConfig.automaticActivityDetection.disabled: true for manual activity mode");let o="START_OF_ACTIVITY_INTERRUPTS";if(t.activityHandling!==void 0){if(!te.includes(t.activityHandling))throw new s(`unsupported activityHandling ${JSON.stringify(t.activityHandling)} \u2014 expected one of ${te.join(", ")}`);o=t.activityHandling}return{manualActivity:!0,activityHandling:o}}var Me=["temperature","maxOutputTokens","responseModalities"];function de(e){for(let[f,p]of Object.entries(Ue))if(e[f]!==void 0)throw new s(p);for(let f of Object.keys(e))if(!De.includes(f))throw new s(`unsupported setup field "${f}"`);if(typeof e.model!="string"||e.model.length===0)throw new s('setup.model is required (format "models/{model}")');let t=e.model.replace(/^models\//,""),n,o,r=!1,i=m(e.generationConfig);if(e.generationConfig!==void 0&&!i)throw new s("setup.generationConfig must be an object");if(i){for(let f of Object.keys(i))if(!Me.includes(f))throw new s(`unsupported generationConfig field "${f}"`);if(i.temperature!==void 0){if(typeof i.temperature!="number")throw new s("generationConfig.temperature must be a number");n=i.temperature}if(i.maxOutputTokens!==void 0){if(typeof i.maxOutputTokens!="number")throw new s("generationConfig.maxOutputTokens must be a number");o=i.maxOutputTokens}if(i.responseModalities!==void 0){let f=Array.isArray(i.responseModalities)?i.responseModalities:[i.responseModalities];for(let p of f){let g=String(p).toUpperCase();if(g==="AUDIO"){r=!0;continue}if(g!=="TEXT")throw new s(`responseModalities ${JSON.stringify(p)} is not supported \u2014 this surface speaks TEXT and AUDIO (audio-out rides the native AudioBridge 'audio' lane)`)}}}let a=[];e.systemInstruction!==void 0&&a.push({role:"system",content:se(e.systemInstruction,"setup.systemInstruction")});let c;if(e.tools!==void 0){if(!Array.isArray(e.tools))throw new s("setup.tools must be an array");c=[];for(let f of e.tools){let p=m(f);if(!p)throw new s("setup.tools entries must be objects");for(let g of Object.keys(p))if(g!=="functionDeclarations")throw new s(`unsupported tool "${g}" \u2014 this slice translates functionDeclarations only (no googleSearch/codeExecution)`);if(!Array.isArray(p.functionDeclarations))throw new s("tool.functionDeclarations must be an array");for(let g of p.functionDeclarations){let h=m(g);if(!h||typeof h.name!="string")throw new s("functionDeclarations entries require a string name");c.push({type:"function",name:h.name,description:h.description,parameters:h.parameters})}}c.length===0&&(c=void 0)}let d={enabled:!1,handle:null};if(e.sessionResumption!==void 0){let f=m(e.sessionResumption);if(!f)throw new s("setup.sessionResumption must be an object (SessionResumptionConfig)");for(let p of Object.keys(f))if(p!=="handle")throw new s(`unsupported sessionResumption field "${p}" \u2014 this lane implements SessionResumptionConfig.handle only`);if(f.handle!==void 0&&(typeof f.handle!="string"||f.handle.length===0))throw new s("sessionResumption.handle must be a non-empty string when present");d={enabled:!0,handle:typeof f.handle=="string"?f.handle:null}}let u=e.realtimeInputConfig===void 0?Le:Ve(e.realtimeInputConfig),l=!1;if(e[k]!==void 0){if(!m(e[k]))throw new s(`setup["${k}"] must be an object (the v1 opt-in is {})`);l=!0}return{model:t,systemItems:a,tools:c,temperature:n,maxOutputTokens:o,audioModality:r,activity:u,sessionResumption:d,videoOut:l}}function ce(e){return{sessionResumptionUpdate:{newHandle:e,resumable:!0}}}function le(e){let t=Math.max(0,e??0)/1e3;return{goAway:{timeLeft:`${Math.round(t*1e3)/1e3}s`}}}function fe(e){for(let o of Object.keys(e))if(o!=="turns"&&o!=="turnComplete")throw new s(`unsupported clientContent field "${o}"`);let t=e.turns===void 0?[]:e.turns;if(!Array.isArray(t))throw new s("clientContent.turns must be an array");let n=[];for(let o of t){let r=m(o);if(!r)throw new s("clientContent.turns entries must be Content objects");let i=r.role===void 0?"user":String(r.role).toLowerCase();if(i!=="user"&&i!=="model")throw new s(`clientContent turn role must be "user" or "model" (got ${JSON.stringify(r.role)})`);n.push({role:i==="model"?"assistant":"user",content:se(r,"clientContent.turns")})}return{items:n,turnComplete:e.turnComplete===!0}}var b=16e3,D=24e3,F=`audio/pcm;rate=${D}`,pe=Buffer.alloc(960).toString("base64");function j(e,t,n){if(t===n)return e;let o=Math.floor(e.length*n/t),r=new Int16Array(o),i=t/n,a=e.length-1;for(let c=0;c<o;c++){let d=c*i,u=Math.floor(d);if(u>=a){r[c]=e[a];continue}let l=d-u;r[c]=Math.round(e[u]+(e[u+1]-e[u])*l)}return r}function Ne(e){let t=m(e);if(!t)throw new s("realtimeInput.video must be a Blob object {mimeType, data}");if(typeof t.mimeType!="string"||!/^image\/jpeg\s*(;|$)/.test(t.mimeType))throw new s(`realtimeInput.video.mimeType must be image/jpeg (the Live video input stream is individual JPEG frames), got ${JSON.stringify(t.mimeType)}`);if(typeof t.data!="string"||t.data.length===0)throw new s("realtimeInput.video.data must be non-empty base64");let n=Buffer.from(t.data,"base64");if(n.length<3||n[0]!==255||n[1]!==216)throw new s("realtimeInput.video.data does not decode to a JPEG frame (missing FF D8 SOI marker)");return new Uint8Array(n)}function Pe(e){let t=m(e);if(!t)throw new s("realtimeInput.audio must be a Blob object {mimeType, data}");if(typeof t.mimeType!="string"||!/^audio\/pcm\s*(;|$)/.test(t.mimeType))throw new s(`realtimeInput.audio.mimeType must be audio/pcm (16-bit LE PCM @16kHz mono per the Live spec), got ${JSON.stringify(t.mimeType)}`);let n=/;\s*rate=(\d+)/.exec(t.mimeType);if(n&&Number(n[1])!==b)throw new s(`realtimeInput.audio.mimeType declares rate=${n[1]} \u2014 the Live spec input format is ${b} Hz (audio/pcm;rate=${b})`);if(typeof t.data!="string"||t.data.length===0)throw new s("realtimeInput.audio.data must be non-empty base64");let o=Buffer.from(t.data,"base64");if(o.length===0||o.length%2!==0)throw new s(`realtimeInput.audio.data must decode to whole 16-bit samples (got ${o.length} bytes \u2014 not even)`);let r=new Int16Array(o.length/2);for(let a=0;a<r.length;a++)r[a]=o.readInt16LE(a*2);let i=j(r,b,D);return Buffer.from(i.buffer,i.byteOffset,i.byteLength).toString("base64")}function me(e){if(e.mediaChunks!==void 0)throw new s('realtimeInput.mediaChunks is deprecated by the Live API \u2014 send the audio stream via realtimeInput.audio ({mimeType:"audio/pcm;rate=16000", data}) instead');if(e.activityStart!==void 0||e.activityEnd!==void 0){if(Object.keys(e).length!==1||e.activityStart!==void 0&&e.activityEnd!==void 0)throw new s("realtimeInput must carry exactly one field when signaling activity (activityStart XOR activityEnd, nothing else)");let r=e.activityStart!==void 0?"activityStart":"activityEnd",i=m(e[r]);if(!i||Object.keys(i).length!==0)throw new s(`realtimeInput.${r} must be an empty object (ActivityStart/ActivityEnd are bare markers)`);return{kind:r}}let t=["text","audio","audioStreamEnd","video"];for(let o of Object.keys(e))if(!t.includes(o))throw new s(`unsupported realtimeInput field "${o}"`);let n=t.filter(o=>e[o]!==void 0);if(n.length!==1)throw new s(`realtimeInput must carry exactly one of ${t.join(", ")} (got ${n.length})`);if(e.audio!==void 0)return{kind:"audio",base64Pcm24k:Pe(e.audio)};if(e.video!==void 0)return{kind:"video",jpegFrame:Ne(e.video)};if(e.audioStreamEnd!==void 0){if(e.audioStreamEnd!==!0)throw new s("realtimeInput.audioStreamEnd must be true when present");return{kind:"audioStreamEnd"}}if(typeof e.text!="string"||e.text.length===0)throw new s("realtimeInput.text must be a non-empty string");return{kind:"text",text:e.text}}function ge(e){for(let t of Object.keys(e))if(t!=="functionResponses")throw new s(`unsupported toolResponse field "${t}"`);if(!Array.isArray(e.functionResponses)||e.functionResponses.length===0)throw new s("toolResponse.functionResponses must be a non-empty array");return e.functionResponses.map(t=>{let n=m(t);if(!n||typeof n.id!="string"||n.id.length===0)throw new s("functionResponses entries require the string id matched from the toolCall");return{type:"function_call_output",call_id:n.id,output:JSON.stringify(n.response??null)}})}function ye(e){return e.type!=="response.output_text.delta"||typeof e.delta!="string"?null:{serverContent:{modelTurn:{role:"model",parts:[{text:e.delta}]}}}}function we(e){return{serverContent:{modelTurn:{role:"model",parts:[{inlineData:{mimeType:F,data:e}}]}}}}function he(e){let t=m(e)??{},n=Array.isArray(t.output)?t.output:[],o=[],r=[],i=typeof t.output_text=="string"?t.output_text:"";if(!i)for(let u of n){let l=m(u);if(!(l?.type!=="message"||!Array.isArray(l.content)))for(let f of l.content){let p=m(f);p?.type==="output_text"&&typeof p.text=="string"&&(i+=p.text)}}i&&r.push({role:"assistant",content:i});for(let u of n){let l=m(u);if(l?.type!=="function_call")continue;if(typeof l.call_id!="string"||typeof l.name!="string")throw new s(`upstream function_call item is missing call_id/name: ${JSON.stringify(u)}`,U);let f=typeof l.arguments=="string"&&l.arguments.length>0?l.arguments:"{}",p;try{p=JSON.parse(f)}catch{throw new s(`upstream function_call arguments are not valid JSON: ${f}`,U)}o.push({id:l.call_id,name:l.name,args:p}),r.push({type:"function_call",call_id:l.call_id,name:l.name,arguments:f})}let a=[];o.length>0&&a.push({toolCall:{functionCalls:o}}),a.push({serverContent:{generationComplete:!0}});let c=m(t.usage),d={serverContent:{turnComplete:!0}};if(c){let u={};typeof c.input_tokens=="number"&&(u.promptTokenCount=c.input_tokens),typeof c.output_tokens=="number"&&(u.responseTokenCount=c.output_tokens),typeof c.total_tokens=="number"&&(u.totalTokenCount=c.total_tokens),Object.keys(u).length>0&&(d.usageMetadata=u)}return a.push(d),{messages:a,assistantItems:r}}function V(e,t,n,o){let r=e.response??{},i={request_id:n,consumer_id:o,stream:!0,kind:"chat",messages:t};return typeof r.instructions=="string"&&(i.instructions=r.instructions),Array.isArray(r.modalities)&&(i.modalities=r.modalities),typeof r.temperature=="number"&&(i.temperature=r.temperature),typeof r.max_output_tokens=="number"&&(i.max_output_tokens=r.max_output_tokens),r.tools!==void 0&&(i.tools=r.tools),i}function M(e,t,n){let o={request_id:t,consumer_id:n,stream:!!e.stream,kind:"responses",input:e.input};return e.model&&(o.model=e.model),e.tools!==void 0&&(o.tools=e.tools),e.tool_choice!==void 0&&(o.tool_choice=e.tool_choice),typeof e.temperature=="number"&&(o.temperature=e.temperature),typeof e.max_output_tokens=="number"&&(o.max_output_tokens=e.max_output_tokens),typeof e.top_p=="number"&&(o.top_p=e.top_p),Array.isArray(e.stop)&&e.stop.length>0&&(o.stop=e.stop),typeof e.reasoning_effort=="string"&&(o.reasoning_effort=e.reasoning_effort),typeof e.instructions=="string"&&(o.instructions=e.instructions),e.chat_template_kwargs!==void 0&&(o.chat_template_kwargs=e.chat_template_kwargs),e.response_format!==void 0&&(o.response_format=e.response_format),typeof e.voice=="string"&&(o.voice=e.voice),o}function ke(e){if(e instanceof Error)return{type:"error",error:{type:"urun_error",code:e.name||null,message:e.message}};if(e&&typeof e=="object"&&e.t==="error"){let t=e,n=t.body??{},o=n.error,r=o?.message??n.message??`unreadable serve error body: ${JSON.stringify(t.body??null)}`;return{type:"error",error:{type:o?.type??"urun_error",code:t.code??o?.code??null,message:r}}}return{type:"error",error:{type:"urun_error",code:null,message:String(e)}}}var $e="llm-resp";async function*A(e,t){let n=`${$e}:${t}`,o=`resp_${t}`;yield{type:"response.created",response:{id:o,status:"in_progress"}};let r=new Map;try{for await(let i of e.stream(n).messages()){let a=i;if(a.t==="delta"){if(typeof a.delta=="string"&&(yield{type:"response.output_text.delta",item_id:o,delta:a.delta}),typeof a.reasoning=="string"&&(yield{type:"response.reasoning_text.delta",item_id:o,delta:a.reasoning}),Array.isArray(a.tool_calls))for(let c of a.tool_calls){let d=typeof c.index=="number"?c.index:0,u=r.get(d)??{};c.id&&(u.id=c.id),c.function?.name&&(u.name=c.function.name),r.set(d,u),yield{type:"response.function_call_arguments.delta",item_id:`fc_${t}_${d}`,tool_index:d,call_id:u.id,name:u.name,delta:c.function?.arguments??""}}}else if(a.t==="response"){yield{type:"response.completed",response:{id:o,status:"completed",...a.body}};return}else if(a.t==="error"){yield ke(a);return}}}finally{try{e.doc("llm").set({requests:{[t]:null}})}catch{}}}var P=class e{constructor(t,n=e.DEFAULT_DOC_SYNC_GATE_TIMEOUT_MS){this.session=t;this.docSyncTimeoutMs=n;this.sessionTag=t.sessionId??globalThis.crypto.randomUUID()}session;docSyncTimeoutMs;sessionTag;static DEFAULT_DOC_SYNC_GATE_TIMEOUT_MS=15e3;get consumerId(){return this.session.consumerId}async ensureDocSynced(){let t=this.session.doc("llm");if(t.synced===!0||typeof t.onSynced!="function")return;let n;await new Promise((r,i)=>{let a=t.onSynced(()=>{clearTimeout(n),r()});n=setTimeout(()=>{a(),i(new Error(`[urun] llm doc plane never synced within ${this.docSyncTimeoutMs}ms \u2014 the request envelope write is withheld rather than raced into a possibly-lost pre-sync write (resumed-attach first-write-lost, prod-usw2 2026-09-23)`))},this.docSyncTimeoutMs)})}write(t){t.session_tag=this.sessionTag,this.session.doc("llm").set({requests:{[t.request_id]:{payload:t,consumer_id:t.consumer_id,stream:t.stream}}})}sendResponses(t,n){return this.writeGated(n,()=>{let o=M(t,n,this.consumerId);return this.write(o),A(this.session,n)})}sendResponseCreate(t,n,o){return this.writeGated(o,()=>{let r=V(t,n,o,this.consumerId);return this.write(r),A(this.session,o)})}async writeGated(t,n){try{await this.ensureDocSynced()}catch(o){throw new Error(`[urun] responses dial ${t} aborted before its envelope write: ${o instanceof Error?o.message:String(o)}`)}return n()}};0&&(module.exports={AudioBridge,CLOSE_PROTOCOL,CLOSE_SESSION_GONE,CLOSE_UNKNOWN_MODEL,CLOSE_UPSTREAM,GEMINI_AUDIO_IN_RATE,GEMINI_AUDIO_OUT_MIME,GEMINI_AUDIO_OUT_RATE,GEMINI_LIVE_PATH,GeminiLiveError,NATIVE_AUDIO_SAMPLE_RATE,OUTBOUND_AUDIO_FLUSH_B64,RT_VIDEO_IN_STREAM,RT_VIDEO_OUT_STREAM,SdkTransport,VIDEO_FRAME_MESSAGE_BUDGET,VIDEO_OUT_DEFAULT_CODEC,VIDEO_OUT_FRAMING,VIDEO_OUT_SETUP_KEY,VIDEO_OUT_WS_FRAMING,consumeVideoOut,decodeResponsesLane,decodeVideoOutRecord,enableSessionAudio,enableSessionVideo,encodeVideoOutRecord,goAwayMsg,negotiateSessionVideoOutQuic,nodeAudioBackend,openSessionVideoOutLane,parseClientMessage,resamplePcm16,responseCreateToEnvelope,responsesCreateToEnvelope,serverContentOfAudio,serverContentOfDelta,sessionResumptionUpdateMsg,translateClientContent,translateRealtimeInput,translateSetup,translateToolResponse,turnEndOfCompleted,videoOutAck,videoOutChunkMsg,weriftAudioBackend});

package/dist/gemini-live.d.cts

@@ -1,10 +1,12 @@
-export { A as ActivityConfig, a as ActivityHandling, C as CLOSE_PROTOCOL, b as CLOSE_SESSION_GONE, c as CLOSE_UNKNOWN_MODEL, d as CLOSE_UPSTREAM, e as ClientMessage, G as GEMINI_AUDIO_IN_RATE, f as GEMINI_AUDIO_OUT_MIME, g as GEMINI_AUDIO_OUT_RATE, h as GEMINI_LIVE_PATH, i as GeminiLiveError, L as LiveSessionConfig, O as OUTBOUND_AUDIO_FLUSH_B64, R as RealtimeInputAction, T as TurnEnd, V as VIDEO_OUT_SETUP_KEY, j as goAwayMsg, p as parseClientMessage, r as resamplePcm16, s as serverContentOfAudio, k as serverContentOfDelta, l as sessionResumptionUpdateMsg, t as translateClientContent, m as translateRealtimeInput, n as translateSetup, o as translateToolResponse, q as turnEndOfCompleted, v as videoOutAck, u as videoOutChunkMsg } from './translator-CcDBEfvm.cjs';
-import { b as UrunSessionLike } from './types-lsVTbNcH.cjs';
-export { L as LaneMessage, a as LaneToolCallDelta, U as UrunDocLike, c as UrunStreamLike } from './types-lsVTbNcH.cjs';
-export { A as AudioBackend, a as AudioBridge, b as AudioSink, c as AudioSource, N as NATIVE_AUDIO_SAMPLE_RATE, R as RT_VIDEO_IN_STREAM, d as RT_VIDEO_OUT_STREAM, V as VIDEO_FRAME_MESSAGE_BUDGET, e as VIDEO_OUT_DEFAULT_CODEC, f as VIDEO_OUT_FRAMING, g as VIDEO_OUT_WS_FRAMING, h as VideoFrameLane, i as VideoOutEndpoint, j as VideoOutLane, k as VideoOutRecord, l as VideoOutRecordHeader, m as VideoOutTransportSink, n as VideoOutWtConnect, o as VideoOutWtSession, p as consumeVideoOut, q as decodeVideoOutRecord, r as enableSessionAudio, s as enableSessionVideo, t as encodeVideoOutRecord, u as negotiateSessionVideoOutQuic, v as nodeAudioBackend, w as openSessionVideoOutLane, x as weriftAudioBackend } from './video-out-D20UuJ8G.cjs';
+export { A as ActivityConfig, a as ActivityHandling, C as CLOSE_PROTOCOL, b as CLOSE_SESSION_GONE, c as CLOSE_UNKNOWN_MODEL, d as CLOSE_UPSTREAM, e as ClientMessage, G as GEMINI_AUDIO_IN_RATE, f as GEMINI_AUDIO_OUT_MIME, g as GEMINI_AUDIO_OUT_RATE, h as GEMINI_LIVE_PATH, i as GeminiLiveError, L as LiveSessionConfig, O as OUTBOUND_AUDIO_FLUSH_B64, R as RealtimeInputAction, T as TurnEnd, V as VIDEO_OUT_SETUP_KEY, j as goAwayMsg, p as parseClientMessage, r as resamplePcm16, s as serverContentOfAudio, k as serverContentOfDelta, l as sessionResumptionUpdateMsg, t as translateClientContent, m as translateRealtimeInput, n as translateSetup, o as translateToolResponse, q as turnEndOfCompleted, v as videoOutAck, u as videoOutChunkMsg } from './translator-CO8W_hmJ.cjs';
+import { b as UrunSessionLike } from './types-CHPtJx6d.cjs';
+export { L as LaneMessage, a as LaneToolCallDelta, U as UrunDocLike, c as UrunStreamLike } from './types-CHPtJx6d.cjs';
+export { A as AudioBackend, a as AudioBridge, b as AudioSink, c as AudioSource, N as NATIVE_AUDIO_SAMPLE_RATE, R as RT_VIDEO_IN_STREAM, d as RT_VIDEO_OUT_STREAM, V as VIDEO_FRAME_MESSAGE_BUDGET, e as VIDEO_OUT_DEFAULT_CODEC, f as VIDEO_OUT_FRAMING, g as VIDEO_OUT_WS_FRAMING, h as VideoFrameLane, i as VideoOutEndpoint, j as VideoOutLane, k as VideoOutRecord, l as VideoOutRecordHeader, m as VideoOutTransportSink, n as VideoOutWtConnect, o as VideoOutWtSession, p as consumeVideoOut, q as decodeVideoOutRecord, r as enableSessionAudio, s as enableSessionVideo, t as encodeVideoOutRecord, u as negotiateSessionVideoOutQuic, v as nodeAudioBackend, w as openSessionVideoOutLane, x as weriftAudioBackend } from './video-out-BsuLqlID.cjs';
 
 declare class SdkTransport {
     private readonly session;
+    /** Test seam for the gate deadline; never an env knob or config surface. */
+    private readonly docSyncTimeoutMs;
     /**
      * Session-affinity tag (urun-python #1556): every envelope this transport
      * writes carries the SAME stable tag, so the serve engine adapter can
@@ -14,13 +16,43 @@ declare class SdkTransport {
      * persistent session, so a per-transport tag identifies the session too.
      */
     private readonly sessionTag;
-    constructor(session: UrunSessionLike);
+    /** The shipped gate deadline — the only production value (test seam overrides it). */
+    private static readonly DEFAULT_DOC_SYNC_GATE_TIMEOUT_MS;
+    constructor(session: UrunSessionLike, 
+    /** Test seam for the gate deadline; never an env knob or config surface. */
+    docSyncTimeoutMs?: number);
     get consumerId(): string;
+    /**
+     * THE FIRST-WRITE-LOST GATE (resumed-attach, prod-usw2 2026-09-23): a turn's
+     * request envelope is the llm doc's FIRST write after a cold (re-)attach, and
+     * the dial's `whenLive` gate resolves off the SIGNALING status frame — which
+     * routinely lands while the yjs doc provider is still dialing its room. A
+     * local write applied while the provider has no socket is relayed ONLY through
+     * the initial-sync diff exchange, and on a resumed session that exchange races
+     * the session-server's S3 hydration barrier. Observed live (local proxy against
+     * prod, dial resumed the long-parked shared-org session): the 396-byte envelope
+     * applied at `wsconnected=false synced=false` NEVER reached the room — the
+     * serve runtime drained the doc for the caller's whole timeout budget and
+     * never saw the request (`drain: 2 in doc, 0 pending`, no change events), while
+     * the client's provider reported `synced`. The write is silently stranded.
+     *
+     * So every request-envelope write first holds for the doc plane's OWN
+     * readiness — `synced`, the provider's level-triggered initial-sync bar.
+     * Post-sync writes ride the connected socket directly and are delivered.
+     * Bounded and loud on deadline: the caller (rehome's one re-dial, then the
+     * lane's loud 502) owns the verdict — never a silent loss. A session double
+     * that exposes no `onSynced` surface cannot be gated and is skipped honestly
+     * (same structural-seam contract as `onConnectionState`).
+     */
+    /** Hold until the llm doc plane has completed its first sync (bounded, loud). */
+    private ensureDocSynced;
     private write;
     sendResponses(params: {
         model?: string;
         input: unknown;
         stream?: boolean;
+        /** Per-request system instructions — forwarded verbatim to the serve envelope. */
+        instructions?: string;
         tools?: unknown;
         tool_choice?: unknown;
         temperature?: number;
@@ -29,14 +61,29 @@ declare class SdkTransport {
         max_output_tokens?: number;
         reasoning_effort?: string;
         chat_template_kwargs?: Record<string, unknown>;
-    }, requestId: string): AsyncIterable<unknown>;
+        /** ENG-310 structured output: forwarded verbatim; the serve runtime validates. */
+        response_format?: unknown;
+        /**
+         * ENG-331 speech lane: the per-request TTS voice, forwarded verbatim
+         * on the serve envelope (encode.ts `ServeEnvelope.voice`); absent →
+         * key absent — the engine's catalog-configured default voice applies.
+         */
+        voice?: string;
+    }, requestId: string): Promise<AsyncIterable<unknown>>;
     sendResponseCreate(ev: {
         type: 'response.create';
         response?: Record<string, unknown>;
     }, conversation: Array<{
         role: string;
         content: unknown;
-    }>, requestId: string): AsyncIterable<unknown>;
+    }>, requestId: string): Promise<AsyncIterable<unknown>>;
+    /**
+     * Run the turn's write + lane subscribe behind {@link ensureDocSynced}. The
+     * gate is per-turn: the doc plane may have re-dialed since the last turn
+     * (resume / pod move), so every turn re-checks — `synced` is true the moment
+     * the provider is warm, so the steady-state turn pays only a property read.
+     */
+    private writeGated;
 }
 
 /**
@@ -90,6 +137,18 @@ interface ServeEnvelope {
      */
     reasoning_effort?: string;
     chat_template_kwargs?: Record<string, unknown>;
+    /**
+     * OpenAI `response_format` — STRUCTURED OUTPUT (ENG-310). Forwarded VERBATIM
+     * to the serve envelope, where it becomes engine-native CONSTRAINED DECODING
+     * (vLLM `structured_outputs`, SGLang `json_schema`) — never a prompt-level
+     * "answer in JSON" hint. `unknown` on purpose: the serve runtime's
+     * `parse_request` is the ONE validator (it accepts both the Chat-Completions
+     * nested `json_schema` spelling and the Responses flat one, and rejects
+     * everything else with a typed 400), so the proxy passes the caller's value
+     * through instead of growing a second, drifting copy of those rules.
+     * Absent -> key absent: today's unconstrained request, byte for byte.
+     */
+    response_format?: unknown;
     /**
      * Session-affinity tag (urun-python #1556, additive like `tools` in the
      * #1557 lane pair): the stable identity of THIS persistent uRun session.
@@ -99,6 +158,15 @@ interface ServeEnvelope {
      * absent -> the server treats the request as stateless, today's behavior.
      */
     session_tag?: string;
+    /**
+     * The speech lane's per-request TTS voice (ENG-331, additive like
+     * `instructions`): the speaker id a `tts`-task engine synthesizes under
+     * (qwen3 CustomVoice timbre, e.g. "Ryan"), applied through the engine's
+     * own per-speak-session override seam (`SpeakSessionConfig.voice` ->
+     * `set_voice`, urun-python tts_engine.py). Absent -> key absent: the
+     * engine's catalog-configured default voice, today's behavior.
+     */
+    voice?: string;
 }
 /** Realtime: a response.create + the accumulated conversation turns → envelope. */
 declare function responseCreateToEnvelope(ev: {
@@ -121,6 +189,10 @@ declare function responsesCreateToEnvelope(params: {
     max_output_tokens?: number;
     reasoning_effort?: string;
     chat_template_kwargs?: Record<string, unknown>;
+    response_format?: unknown;
+    voice?: string;
+    /** Per-request system instructions (the realtime standing schema's only path to the serve envelope). */
+    instructions?: string;
 }, requestId: string, consumerId: string): ServeEnvelope;
 
 export { SdkTransport, type ServeEnvelope, UrunSessionLike, decodeResponsesLane, responseCreateToEnvelope, responsesCreateToEnvelope };

package/dist/gemini-live.d.ts

@@ -1,19 +1,30 @@
-export { A as ActivityConfig, a as ActivityHandling, C as CLOSE_PROTOCOL, b as CLOSE_SESSION_GONE, c as CLOSE_UNKNOWN_MODEL, d as CLOSE_UPSTREAM, e as ClientMessage, G as GEMINI_AUDIO_IN_RATE, f as GEMINI_AUDIO_OUT_MIME, g as GEMINI_AUDIO_OUT_RATE, h as GEMINI_LIVE_PATH, i as GeminiLiveError, L as LiveSessionConfig, O as OUTBOUND_AUDIO_FLUSH_B64, R as RealtimeInputAction, T as TurnEnd, V as VIDEO_OUT_SETUP_KEY, j as goAwayMsg, p as parseClientMessage, r as resamplePcm16, s as serverContentOfAudio, k as serverContentOfDelta, l as sessionResumptionUpdateMsg, t as translateClientContent, m as translateRealtimeInput, n as translateSetup, o as translateToolResponse, q as turnEndOfCompleted, v as videoOutAck, u as videoOutChunkMsg } from './translator-C9uPKypK.js';
-import { b as UrunSessionLike } from './types-lsVTbNcH.js';
-export { L as LaneMessage, a as LaneToolCallDelta, U as UrunDocLike, c as UrunStreamLike } from './types-lsVTbNcH.js';
-export { A as AudioBackend, a as AudioBridge, b as AudioSink, c as AudioSource, N as NATIVE_AUDIO_SAMPLE_RATE, R as RT_VIDEO_IN_STREAM, d as RT_VIDEO_OUT_STREAM, V as VIDEO_FRAME_MESSAGE_BUDGET, e as VIDEO_OUT_DEFAULT_CODEC, f as VIDEO_OUT_FRAMING, g as VIDEO_OUT_WS_FRAMING, h as VideoFrameLane, i as VideoOutEndpoint, j as VideoOutLane, k as VideoOutRecord, l as VideoOutRecordHeader, m as VideoOutTransportSink, n as VideoOutWtConnect, o as VideoOutWtSession, p as consumeVideoOut, q as decodeVideoOutRecord, r as enableSessionAudio, s as enableSessionVideo, t as encodeVideoOutRecord, u as negotiateSessionVideoOutQuic, v as nodeAudioBackend, w as openSessionVideoOutLane, x as weriftAudioBackend } from './video-out-CWesbk12.js';
+export { A as ActivityConfig, a as ActivityHandling, C as CLOSE_PROTOCOL, b as CLOSE_SESSION_GONE, c as CLOSE_UNKNOWN_MODEL, d as CLOSE_UPSTREAM, e as ClientMessage, G as GEMINI_AUDIO_IN_RATE, f as GEMINI_AUDIO_OUT_MIME, g as GEMINI_AUDIO_OUT_RATE, h as GEMINI_LIVE_PATH, i as GeminiLiveError, L as LiveSessionConfig, O as OUTBOUND_AUDIO_FLUSH_B64, R as RealtimeInputAction, T as TurnEnd, V as VIDEO_OUT_SETUP_KEY, j as goAwayMsg, p as parseClientMessage, r as resamplePcm16, s as serverContentOfAudio, k as serverContentOfDelta, l as sessionResumptionUpdateMsg, t as translateClientContent, m as translateRealtimeInput, n as translateSetup, o as translateToolResponse, q as turnEndOfCompleted, v as videoOutAck, u as videoOutChunkMsg } from './translator-Ddd65sXR.js';
+import { b as UrunSessionLike } from './types-CHPtJx6d.js';
+export { L as LaneMessage, a as LaneToolCallDelta, U as UrunDocLike, c as UrunStreamLike } from './types-CHPtJx6d.js';
+export { A as AudioBackend, a as AudioBridge, b as AudioSink, c as AudioSource, N as NATIVE_AUDIO_SAMPLE_RATE, R as RT_VIDEO_IN_STREAM, d as RT_VIDEO_OUT_STREAM, V as VIDEO_FRAME_MESSAGE_BUDGET, e as VIDEO_OUT_DEFAULT_CODEC, f as VIDEO_OUT_FRAMING, g as VIDEO_OUT_WS_FRAMING, h as VideoFrameLane, i as VideoOutEndpoint, j as VideoOutLane, k as VideoOutRecord, l as VideoOutRecordHeader, m as VideoOutTransportSink, n as VideoOutWtConnect, o as VideoOutWtSession, p as consumeVideoOut, q as decodeVideoOutRecord, r as enableSessionAudio, s as enableSessionVideo, t as encodeVideoOutRecord, u as negotiateSessionVideoOutQuic, v as nodeAudioBackend, w as openSessionVideoOutLane, x as weriftAudioBackend } from './video-out-Cawtm7SF.js';
 
 declare class SdkTransport {
     private readonly session;
     
+    private readonly docSyncTimeoutMs;
+    
     private readonly sessionTag;
-    constructor(session: UrunSessionLike);
+    
+    private static readonly DEFAULT_DOC_SYNC_GATE_TIMEOUT_MS;
+    constructor(session: UrunSessionLike, 
+    
+    docSyncTimeoutMs?: number);
     get consumerId(): string;
+    
+    
+    private ensureDocSynced;
     private write;
     sendResponses(params: {
         model?: string;
         input: unknown;
         stream?: boolean;
+        
+        instructions?: string;
         tools?: unknown;
         tool_choice?: unknown;
         temperature?: number;
@@ -22,14 +33,20 @@ declare class SdkTransport {
         max_output_tokens?: number;
         reasoning_effort?: string;
         chat_template_kwargs?: Record<string, unknown>;
-    }, requestId: string): AsyncIterable<unknown>;
+        
+        response_format?: unknown;
+        
+        voice?: string;
+    }, requestId: string): Promise<AsyncIterable<unknown>>;
     sendResponseCreate(ev: {
         type: 'response.create';
         response?: Record<string, unknown>;
     }, conversation: Array<{
         role: string;
         content: unknown;
-    }>, requestId: string): AsyncIterable<unknown>;
+    }>, requestId: string): Promise<AsyncIterable<unknown>>;
+    
+    private writeGated;
 }
 
 declare function decodeResponsesLane(session: UrunSessionLike, requestId: string): AsyncIterable<unknown>;
@@ -57,7 +74,11 @@ interface ServeEnvelope {
     reasoning_effort?: string;
     chat_template_kwargs?: Record<string, unknown>;
     
+    response_format?: unknown;
+    
     session_tag?: string;
+    
+    voice?: string;
 }
 
 declare function responseCreateToEnvelope(ev: {
@@ -80,6 +101,10 @@ declare function responsesCreateToEnvelope(params: {
     max_output_tokens?: number;
     reasoning_effort?: string;
     chat_template_kwargs?: Record<string, unknown>;
+    response_format?: unknown;
+    voice?: string;
+    
+    instructions?: string;
 }, requestId: string, consumerId: string): ServeEnvelope;
 
 export { SdkTransport, type ServeEnvelope, UrunSessionLike, decodeResponsesLane, responseCreateToEnvelope, responsesCreateToEnvelope };

package/dist/gemini-live.js

@@ -1 +1 @@
-import{a as c,b as C,c as L,d as M,e as R,f as D,g as f,h as N,i as v,j as g,k as x,l as k,m as y,n as G,o as B,p as F,q as P,r as H,s as W,t as b,u as w,v as K,w as h,x as Q}from"./chunk-5NXM4IO3.js";import{a as i,b as s,c as O,d,e as a,f as E,g as _,h as p,i as u,j as A,k as S,l as I,m as T,n as l,o as m,p as U,q as V}from"./chunk-CWJRDBDC.js";import{a as o,b as t,d as n,e as r}from"./chunk-OI2OY32M.js";import{c as e}from"./chunk-YSFSRI3D.js";e();export{s as AudioBridge,C as CLOSE_PROTOCOL,R as CLOSE_SESSION_GONE,L as CLOSE_UNKNOWN_MODEL,M as CLOSE_UPSTREAM,B as GEMINI_AUDIO_IN_RATE,P as GEMINI_AUDIO_OUT_MIME,F as GEMINI_AUDIO_OUT_RATE,c as GEMINI_LIVE_PATH,D as GeminiLiveError,i as NATIVE_AUDIO_SAMPLE_RATE,H as OUTBOUND_AUDIO_FLUSH_B64,d as RT_VIDEO_IN_STREAM,u as RT_VIDEO_OUT_STREAM,r as SdkTransport,a as VIDEO_FRAME_MESSAGE_BUDGET,S as VIDEO_OUT_DEFAULT_CODEC,A as VIDEO_OUT_FRAMING,N as VIDEO_OUT_SETUP_KEY,m as VIDEO_OUT_WS_FRAMING,V as consumeVideoOut,n as decodeResponsesLane,T as decodeVideoOutRecord,O as enableSessionAudio,E as enableSessionVideo,I as encodeVideoOutRecord,y as goAwayMsg,l as negotiateSessionVideoOutQuic,p as nodeAudioBackend,U as openSessionVideoOutLane,f as parseClientMessage,W as resamplePcm16,o as responseCreateToEnvelope,t as responsesCreateToEnvelope,h as serverContentOfAudio,K as serverContentOfDelta,k as sessionResumptionUpdateMsg,G as translateClientContent,b as translateRealtimeInput,x as translateSetup,w as translateToolResponse,Q as turnEndOfCompleted,v as videoOutAck,g as videoOutChunkMsg,_ as weriftAudioBackend};
+import{a as c,b as C,c as L,d as M,e as R,f as D,g as f,h as N,i as v,j as g,k as x,l as k,m as y,n as G,o as B,p as F,q as P,r as H,s as W,t as b,u as w,v as K,w as h,x as Q}from"./chunk-3JWYIKHM.js";import{a as i,b as s,c as O,d,e as a,f as E,g as _,h as p,i as u,j as A,k as S,l as I,m as T,n as l,o as m,p as U,q as V}from"./chunk-6M5JK4YC.js";import{a as o,b as t,d as n,e as r}from"./chunk-QTVTCMJU.js";import{c as e}from"./chunk-YSFSRI3D.js";e();export{s as AudioBridge,C as CLOSE_PROTOCOL,R as CLOSE_SESSION_GONE,L as CLOSE_UNKNOWN_MODEL,M as CLOSE_UPSTREAM,B as GEMINI_AUDIO_IN_RATE,P as GEMINI_AUDIO_OUT_MIME,F as GEMINI_AUDIO_OUT_RATE,c as GEMINI_LIVE_PATH,D as GeminiLiveError,i as NATIVE_AUDIO_SAMPLE_RATE,H as OUTBOUND_AUDIO_FLUSH_B64,d as RT_VIDEO_IN_STREAM,u as RT_VIDEO_OUT_STREAM,r as SdkTransport,a as VIDEO_FRAME_MESSAGE_BUDGET,S as VIDEO_OUT_DEFAULT_CODEC,A as VIDEO_OUT_FRAMING,N as VIDEO_OUT_SETUP_KEY,m as VIDEO_OUT_WS_FRAMING,V as consumeVideoOut,n as decodeResponsesLane,T as decodeVideoOutRecord,O as enableSessionAudio,E as enableSessionVideo,I as encodeVideoOutRecord,y as goAwayMsg,l as negotiateSessionVideoOutQuic,p as nodeAudioBackend,U as openSessionVideoOutLane,f as parseClientMessage,W as resamplePcm16,o as responseCreateToEnvelope,t as responsesCreateToEnvelope,h as serverContentOfAudio,K as serverContentOfDelta,k as sessionResumptionUpdateMsg,G as translateClientContent,b as translateRealtimeInput,x as translateSetup,w as translateToolResponse,Q as turnEndOfCompleted,v as videoOutAck,g as videoOutChunkMsg,_ as weriftAudioBackend};