@urun-sh/openai 0.5.5 → 0.5.6

package/dist/proxy/index.d.cts

@@ -1,10 +1,10 @@
 import { Server, IncomingMessage } from 'node:http';
-import { a as ProxyHandlerOptions, P as ProxyClients } from '../responses-turn-OrO4euEN.cjs';
-export { b as ProxyIdentity, c as ProxyVideoOutLane, S as SessionGoneError, U as UnknownModelError } from '../responses-turn-OrO4euEN.cjs';
-export { h as GEMINI_LIVE_PATH, V as VIDEO_OUT_SETUP_KEY } from '../translator-CcDBEfvm.cjs';
+import { f as ProxyHandlerOptions, P as ProxyClients } from '../responses-turn-tNUHq5b0.cjs';
+export { D as DialObserver, a as DialReport, g as ProxyIdentity, h as ProxyVideoOutLane, S as SessionGoneError, U as UnknownModelError } from '../responses-turn-tNUHq5b0.cjs';
+export { h as GEMINI_LIVE_PATH, V as VIDEO_OUT_SETUP_KEY } from '../translator-Bh0Bp_Ie.cjs';
 import { WebSocket } from 'ws';
-import '../models-NYMZrklp.cjs';
-import '../video-out-D20UuJ8G.cjs';
+import '../models-DUdx_Y6X.cjs';
+import '../video-out-CCksIcj8.cjs';
 import '../types-lsVTbNcH.cjs';
 
 /**
@@ -329,6 +329,14 @@ interface RealtimeBinding<S = unknown> {
         confidence?: number;
         startMs?: number;
     }>;
+    /**
+     * Release THIS attachment when the socket closes (the hosted
+     * conversation-factory seam — the twin of the Gemini Live binding's
+     * `release`): per-attachment release, native detach underneath, the named
+     * session itself stays live. Absent (the local fixed-clients lane) = the
+     * pooled sessions are the process's own and outlive the socket.
+     */
+    release?: () => Promise<void> | void;
 }
 interface OpenAIRealtimeOptions<S = unknown> {
     resolver: OpenAIRealtimeResolver<S>;
@@ -422,11 +430,29 @@ declare class OpenAIRealtimeConnection<S = unknown> {
      * SdkTransport-style) OR `clients.createResponse(params)` (the canonical
      * ProxyClients seam the shared conversation factory hands out). A binding
      * with neither was already rejected at upgrade time; this is the loud
-     * backstop.
+     * backstop. The ProxyClients seam is PROMISE-VALUED
+     * (`rehomingCreateResponse` dials inside an async function — its
+     * private-backhaul cold wake can park the promise for seconds), so it is
+     * awaited here; awaiting the SdkTransport generator is a no-op. The cast
+     * states the seam's REAL union: typing it sync-iterable is what let the
+     * unawaited promise reach `for await` as "R is not async iterable" (the
+     * prod-usw2 speak-turn kill, 2026-09-14).
      */
     private upstreamStream;
     private startResponse;
-    /** Bridge voice OUT frame → `response.output_audio.delta` on the open turn. */
+    /**
+     * Bridge voice OUT frame → `response.output_audio.delta` on the open turn.
+     *
+     * OUT-OF-TURN frames split by CONTENT, per the platform's own liveness
+     * model (urun-python #1812/#1827): the runtime's egress holds a DECLARED
+     * quiet track — `CtxMediaTransport`'s silence keepalive emits Opus silence
+     * whenever the app is silent, including before the first response.create —
+     * so an all-quiet frame with no open turn is that muted egress, not voice.
+     * It is dropped quietly (the same reading as a negotiated muted WebRTC
+     * track). A NON-quiet frame with no open turn is real voice the adapter
+     * cannot attribute — still the loud, once-per-connection
+     * `out_of_turn_audio` error that caught the speak-lane bug class.
+     */
     private onOutputAudioFrame;
     /** One assistant message item per turn (text and audio parts share it). */
     private messageItemAdded;
@@ -468,4 +494,54 @@ declare class OpenAIRealtimeConnection<S = unknown> {
     private teardown;
 }
 
-export { type GeminiLiveCall, type GeminiLiveClients, OPENAI_REALTIME_PATH, type OpenAIProxyOptions, OpenAIRealtimeConnection, type OpenAIRealtimeOptions, type OpenAIRealtimeResolver, ProxyClients, type RealtimeBinding, ResumptionRegistry, attachGeminiLive, attachOpenAIRealtime, createOpenAIProxy, serveGeminiLiveSocket };
+/**
+ * THE REALTIME BINDING COMPOSITION — the ONE seam every composition rides to
+ * turn an upgraded `/v1/realtime` request into a {@link RealtimeBinding}.
+ *
+ * The modality is read from the CATALOG (`task` stt | tts, invariant across
+ * placements) through the SAME router-backed gate seam the Images lane uses
+ * (`ProxyClients.supportsAudio` — `ModelRouter.supportsAudio`). A non-audio
+ * model is refused LOUDLY with a 404 before anything is opened; a voice
+ * binding acquires its PRIVATE audio bridge from `ProxyClients.openAudio`
+ * (the pooled session's native `audio` lane at 24 kHz — the ONE canonical
+ * session-audio wiring, `transport/media.ts` AUDIO_STREAM). There is no env-var
+ * switch and no chat-default fall-through: an absent model resolves through
+ * the router exactly like every other lane, and a chat default is refused
+ * exactly like an explicitly named chat model.
+ *
+ * Used by BOTH compositions — the local `urun compat proxy` (fixed clients,
+ * optional fixed apiKey) and the hosted multi-tenant endpoint (per-request
+ * org resolution + per-conversation private backhauls) — so there is exactly
+ * ONE binding implementation and no copy-pasted second handler.
+ */
+
+/**
+ * The realtime lane's model name — the GA `?model=` query parameter
+ * (`wss://…/v1/realtime?model=…`). Absent = the router's configured default
+ * resolves it (locally the startup app; hosted, a loud 404 — there IS no
+ * default app on the hosted lane).
+ */
+declare function realtimeModelOf(req: IncomingMessage): string | undefined;
+/**
+ * Build the binding for `model` against `clients`:
+ *   - NO `supportsAudio` seam ⇒ 501 (loud, never a silent allow-all).
+ *   - {@link UnknownModelError} ⇒ 404 with the router's own loud message.
+ *   - A non-audio verdict ⇒ 404 — the realtime lane never falls through to
+ *     the chat lane, in either direction.
+ *   - A voice verdict with no `openAudio` seam ⇒ 501 (a voice binding cannot
+ *     be served without its bridge; a voiceless downgrade is never silent).
+ */
+declare function realtimeBindingFor(clients: Pick<ProxyClients, 'createResponse'> & Partial<Pick<ProxyClients, 'supportsAudio' | 'openAudio'>>, model: string | undefined): Promise<RealtimeBinding<ProxyClients>>;
+/**
+ * The LOCAL lane resolver: fixed pooled clients, optional fixed apiKey (the
+ * same shared-secret rule the local Gemini Live lane applies — a mismatched
+ * or missing Bearer is a 401; with no configured key the header is not
+ * consulted). The HOSTED lane builds its own resolver per request (org
+ * resolution + per-conversation private backhaul) and reuses
+ * {@link realtimeBindingFor} for the binding itself.
+ */
+declare function fixedClientsResolver(clients: ProxyClients, apiKey?: string): {
+    resolve(req: IncomingMessage): Promise<RealtimeBinding<ProxyClients>>;
+};
+
+export { type GeminiLiveCall, type GeminiLiveClients, OPENAI_REALTIME_PATH, type OpenAIProxyOptions, OpenAIRealtimeConnection, type OpenAIRealtimeOptions, type OpenAIRealtimeResolver, ProxyClients, type RealtimeBinding, ResumptionRegistry, attachGeminiLive, attachOpenAIRealtime, createOpenAIProxy, fixedClientsResolver, realtimeBindingFor, realtimeModelOf, serveGeminiLiveSocket };

package/dist/proxy/index.d.ts

@@ -1,10 +1,10 @@
 import { Server, IncomingMessage } from 'node:http';
-import { a as ProxyHandlerOptions, P as ProxyClients } from '../responses-turn-KOAoIqZ-.js';
-export { b as ProxyIdentity, c as ProxyVideoOutLane, S as SessionGoneError, U as UnknownModelError } from '../responses-turn-KOAoIqZ-.js';
-export { h as GEMINI_LIVE_PATH, V as VIDEO_OUT_SETUP_KEY } from '../translator-C9uPKypK.js';
+import { f as ProxyHandlerOptions, P as ProxyClients } from '../responses-turn-Dr37N3kH.js';
+export { D as DialObserver, a as DialReport, g as ProxyIdentity, h as ProxyVideoOutLane, S as SessionGoneError, U as UnknownModelError } from '../responses-turn-Dr37N3kH.js';
+export { h as GEMINI_LIVE_PATH, V as VIDEO_OUT_SETUP_KEY } from '../translator-BCoRFaTs.js';
 import { WebSocket } from 'ws';
-import '../models-NYMZrklp.js';
-import '../video-out-CWesbk12.js';
+import '../models-DUdx_Y6X.js';
+import '../video-out-IGQ4YQ-c.js';
 import '../types-lsVTbNcH.js';
 
 interface OpenAIProxyOptions extends ProxyHandlerOptions {
@@ -87,6 +87,8 @@ interface RealtimeBinding<S = unknown> {
         confidence?: number;
         startMs?: number;
     }>;
+    
+    release?: () => Promise<void> | void;
 }
 interface OpenAIRealtimeOptions<S = unknown> {
     resolver: OpenAIRealtimeResolver<S>;
@@ -173,4 +175,12 @@ declare class OpenAIRealtimeConnection<S = unknown> {
     private teardown;
 }
 
-export { type GeminiLiveCall, type GeminiLiveClients, OPENAI_REALTIME_PATH, type OpenAIProxyOptions, OpenAIRealtimeConnection, type OpenAIRealtimeOptions, type OpenAIRealtimeResolver, ProxyClients, type RealtimeBinding, ResumptionRegistry, attachGeminiLive, attachOpenAIRealtime, createOpenAIProxy, serveGeminiLiveSocket };
+declare function realtimeModelOf(req: IncomingMessage): string | undefined;
+
+declare function realtimeBindingFor(clients: Pick<ProxyClients, 'createResponse'> & Partial<Pick<ProxyClients, 'supportsAudio' | 'openAudio'>>, model: string | undefined): Promise<RealtimeBinding<ProxyClients>>;
+
+declare function fixedClientsResolver(clients: ProxyClients, apiKey?: string): {
+    resolve(req: IncomingMessage): Promise<RealtimeBinding<ProxyClients>>;
+};
+
+export { type GeminiLiveCall, type GeminiLiveClients, OPENAI_REALTIME_PATH, type OpenAIProxyOptions, OpenAIRealtimeConnection, type OpenAIRealtimeOptions, type OpenAIRealtimeResolver, ProxyClients, type RealtimeBinding, ResumptionRegistry, attachGeminiLive, attachOpenAIRealtime, createOpenAIProxy, fixedClientsResolver, realtimeBindingFor, realtimeModelOf, serveGeminiLiveSocket };

package/dist/proxy/index.js

@@ -1,6 +1 @@
-import{a as ee,b as te,c as ne,g as ie}from"../chunk-QVF7NF7G.js";import{b as H,c as K}from"../chunk-DMD5UENK.js";import{a as Z,h as Q}from"../chunk-5NXM4IO3.js";import"../chunk-CWJRDBDC.js";import{c as g}from"../chunk-YSFSRI3D.js";g();g();g();import{WebSocketServer as ce}from"ws";g();g();import{randomUUID as se}from"crypto";function r(a){return`${a}_${se().replace(/-/g,"").slice(0,24)}`}var F=4*1024*1024,$=8*1024*1024,U=512,P=20*1024*1024;function f(a,e,t){return{type:"invalid_request_error",code:a,message:e,param:t??null}}function E(a,e){return{type:"server_error",code:a,message:e}}function m(a){return{type:"error",event_id:r("event"),error:a}}function D(){return{outputModalities:["text"],turnDetection:null,inputAudioTranscription:!1}}var u=class extends Error{constructor(t,i,n="invalid_value"){super(t);this.param=i;this.code=n;this.name="SessionUpdateError"}param;code},oe={model:!0,instructions:!0,modalities:!0,voice:!0,input_audio_format:!0,output_audio_format:!0,temperature:!0,max_output_tokens:!0,max_response_output_tokens:!0,tool_choice:!0,tools:!0,metadata:!0,turn_detection:!0,input_audio_transcription:!0,input_audio_noise_reduction:!0,audio:!0},re={format:!0,transcription:!0,turn_detection:!0,noise_reduction:!0},ae={format:!0,voice:!0,speed:!0};function W(a,e){let t=a;if(!t||typeof t!="object"||t.type!=="audio/pcm"||t.rate!==void 0&&t.rate!==24e3)throw new u(`${e} must be {"type":"audio/pcm","rate":24000} \u2014 the session audio lane is 24 kHz mono PCM16 only`,e)}function S(a,e,t){if(!a||typeof a!="object"||Array.isArray(a))throw new u("session must be an object","session","invalid_request");let i=a,n=Object.keys(i).filter(p=>!(p in oe));if(n.length>0)throw new u(`session field(s) not supported by this backend: ${n.join(", ")}`,n[0],"unsupported_field");let s={...e,outputModalities:[...e.outputModalities]};for(let[p,o]of Object.entries(i))switch(p){case"model":if(o!==null&&typeof o!="string")throw new u("model must be a string or null","model");s.model=o??void 0;break;case"instructions":if(o!==null&&typeof o!="string")throw new u("instructions must be a string or null","instructions");s.instructions=o??void 0;break;case"modalities":{if(!Array.isArray(o)||o.length===0)throw new u("modalities must be a non-empty array","modalities");let l=[];for(let d of o){if(d!=="text"&&d!=="audio")throw new u(`modality ${JSON.stringify(d)} is not "text" or "audio"`,"modalities");if(d==="audio"&&!t.voice)throw new u('modality "audio" refused: the resolver did not provide an authoritative voice binding for the resolved backend \u2014 this model cannot speak on this session',"modalities","audio_not_supported");l.includes(d)||l.push(d)}s.outputModalities=l;break}case"voice":if(o===null){delete s.voice;break}if(typeof o!="string")throw new u("voice must be a string or null","voice");if(!t.voice)throw new u("voice refused: no authoritative voice binding for the resolved backend","voice","audio_not_supported");s.voice=o;break;case"input_audio_format":case"output_audio_format":if(o!=="pcm16")throw new u(`${p} must be "pcm16" \u2014 the session audio lane is 24 kHz mono PCM16 only`,p);break;case"temperature":if(o!==null&&(typeof o!="number"||!Number.isFinite(o)))throw new u("temperature must be a finite number or null","temperature");s.temperature=o??void 0;break;case"max_output_tokens":case"max_response_output_tokens":if(o!==null&&o!=="inf"&&(typeof o!="number"||!Number.isInteger(o)||o<1))throw new u(`${p} must be a positive integer, "inf" or null`,p);s.maxOutputTokens=o===null?void 0:o;break;case"tools":if(!Array.isArray(o))throw new u("tools must be an array","tools");s.tools=o;break;case"tool_choice":s.toolChoice=o;break;case"metadata":if(o!==null&&(!o||typeof o!="object"||Array.isArray(o)))throw new u("metadata must be an object or null","metadata");s.metadata=o??null;break;case"turn_detection":{if(o===null){s.turnDetection=null;break}if(!t.turns)throw new u("turn_detection is refused: the resolved backend does not wire its own turn boundary lane (stt stream with turn_start/turn_end) \u2014 there is no native VAD source to drive responses from. Drive turns manually: input_audio_buffer.commit + response.create.","turn_detection","unsupported_field");let l=o;if(!l||typeof l!="object"||Array.isArray(l))throw new u("turn_detection must be an object or null","turn_detection");let d=String(l.type??"");if(d!=="server_vad"&&d!=="semantic_vad")throw new u(`turn_detection.type ${JSON.stringify(l.type)} is not "server_vad"/"semantic_vad"`,"turn_detection.type");s.turnDetection={createResponse:l.create_response!==!1,interruptResponse:l.interrupt_response!==!1};break}case"input_audio_transcription":{if(o===null){s.inputAudioTranscription=!1;break}if(!t.turns)throw new u("input_audio_transcription is refused: the resolved backend does not wire a transcript lane \u2014 no native transcription source exists behind this session","input_audio_transcription","unsupported_field");if(!o||typeof o!="object"||Array.isArray(o))throw new u("input_audio_transcription must be an object or null","input_audio_transcription");for(let l of Object.keys(o))if(!["model","language","languages","prompt"].includes(l))throw new u(`input_audio_transcription field not honored by this backend: ${l} (transcription comes from the voice engine itself)`,`input_audio_transcription.${l}`,"unsupported_field");s.inputAudioTranscription=!0;break}case"input_audio_noise_reduction":if(o!==null)throw new u("input_audio_noise_reduction must be null on this backend: no noise-reduction primitive exists behind the session audio bridge","input_audio_noise_reduction","unsupported_field");break;case"audio":{if(o===null)break;let l=o;if(!l||typeof l!="object"||Array.isArray(l))throw new u("audio must be an object or null","audio");if(l.input!==void 0&&l.input!==null){let d=l.input;if(!d||typeof d!="object"||Array.isArray(d))throw new u("audio.input must be an object","audio.input");for(let c of Object.keys(d))if(!(c in re))throw new u(`audio.input field not supported by this backend: ${c}`,`audio.input.${c}`,"unsupported_field");if(d.format!==void 0&&W(d.format,"audio.input.format"),d.transcription!==null&&d.transcription!==void 0){let c=S({input_audio_transcription:d.transcription},s,t);s.inputAudioTranscription=c.inputAudioTranscription}if(d.turn_detection!==null&&d.turn_detection!==void 0){let c=S({turn_detection:d.turn_detection},s,t);s.turnDetection=c.turnDetection}if(d.noise_reduction!==null&&d.noise_reduction!==void 0)throw new u("audio.input.noise_reduction must be null on this backend: no noise-reduction primitive","audio.input.noise_reduction","unsupported_field")}if(l.output!==void 0&&l.output!==null){let d=l.output;if(!d||typeof d!="object"||Array.isArray(d))throw new u("audio.output must be an object","audio.output");for(let c of Object.keys(d))if(!(c in ae))throw new u(`audio.output field not supported by this backend: ${c}`,`audio.output.${c}`,"unsupported_field");if(d.format!==void 0&&W(d.format,"audio.output.format"),d.speed!==void 0)throw new u("audio.output.speed is not supported on this backend: playback rate is fixed","audio.output.speed","unsupported_field");if(d.voice!==void 0)if(d.voice===null)delete s.voice;else{if(typeof d.voice!="string")throw new u("audio.output.voice must be a string or null","audio.output.voice");if(!t.voice)throw new u("audio.output.voice refused: no authoritative voice binding for the resolved backend","audio.output.voice","audio_not_supported");s.voice=d.voice}d.format!==void 0&&!s.outputModalities.includes("audio")&&(s.outputModalities=["text","audio"])}break}default:break}return s}function j(a,e){return{id:e,object:"realtime.session",model:a.model??null,instructions:a.instructions??"",modalities:[...a.outputModalities],voice:a.voice??null,input_audio_format:"pcm16",output_audio_format:"pcm16",input_audio_transcription:a.inputAudioTranscription?{}:null,input_audio_noise_reduction:null,turn_detection:a.turnDetection?{type:"server_vad",create_response:a.turnDetection.createResponse,interrupt_response:a.turnDetection.interruptResponse}:null,temperature:a.temperature??null,max_response_output_tokens:a.maxOutputTokens??"inf",tool_choice:a.toolChoice??"auto",tools:a.tools??[],metadata:a.metadata??null}}function q(a,e){let t=a;if(!t||typeof t!="object")throw new u("item must be an object","item","invalid_request");if(t.type==="function_call_output"){if(typeof t.call_id!="string"||typeof t.output!="string")throw new u("function_call_output item requires string call_id and output","item.call_id","invalid_request");return{id:typeof t.id=="string"?t.id:r("item"),type:"function_call_output",callId:t.call_id,output:t.output}}if(t.type==="message"){let i=t.role;if(i!=="user"&&i!=="assistant"&&i!=="system")throw new u('message item role must be "user", "assistant" or "system"',"item.role","invalid_request");if(!Array.isArray(t.content))throw new u("message item content must be an array","item.content","invalid_request");let n=[];for(let s of t.content){let p=s;if(!p||typeof p!="object")throw new u("content part must be an object","item.content","invalid_request");if(p.type==="input_text"||i==="assistant"&&p.type==="output_text"){if(typeof p.text!="string")throw new u(`${String(p.type)} part requires string text`,"item.content.text","invalid_request");n.push({type:p.type,text:p.text});continue}if(p.type==="input_audio"){if(!e)throw new u("input_audio content refused: no authoritative voice binding for the resolved backend","item.content.type","audio_not_supported");if(typeof p.audio!="string")throw new u("input_audio part requires base64 audio","item.content.audio","invalid_request");n.push({type:"input_audio",...typeof p.transcript=="string"?{transcript:p.transcript}:{}});continue}throw new u(`content part type ${String(p.type)} is not supported on message items here (input_text / output_text / input_audio)`,"item.content.type","invalid_request")}return{id:typeof t.id=="string"?t.id:r("item"),type:"message",role:i,content:n}}throw new u(`item type ${String(t.type)} is not supported \u2014 message and function_call_output items only`,"item.type","invalid_request")}function L(a){let e=[],t=!1;for(let i of a){if(i.type==="function_call"){e.push({type:"function_call",call_id:i.callId,name:i.name,arguments:i.arguments??"{}"});continue}if(i.type==="function_call_output"){e.push({type:"function_call_output",call_id:i.callId,output:i.output});continue}let n=[];for(let s of i.content??[]){if(s.type==="input_text"||s.type==="output_text"){n.push({type:s.type,text:s.text});continue}if(s.type==="input_audio"){let p=s.transcript;if(typeof p=="string"&&p.trim().length>0){n.push({type:"input_text",text:p});continue}t=!0;continue}}e.push({role:i.role,content:n})}return{input:e,sawAudioOnlyItem:t}}var ue=1001,de=1011,X=a=>a.slice(0,120),pe=/^[A-Za-z0-9+/=]*$/,R=class{id=r("conn");ws;binding;onClosed;config=D();sessionId=r("sess");conversationId=r("conv");items=[];pendingAudioBytes=[];pendingAudioTotalBytes=0;frozenAudio=[];inFlight=null;activeResponseId=null;responseFlushed=!1;cancelled=!1;voiceClaimed=!1;audioLaneOwner=!1;audioUnsubscribe=null;outOfTurnAudioWarned=!1;audioOnlyTranscriptGapWarned=!1;closed=!1;turnLaneAbort=null;turnTranscript={open:!1,text:""};lastCommittedTranscript=null;constructor(e){this.ws=e.ws,this.binding=e.binding,this.onClosed=e.onClosed,this.binding.model&&(this.config.model=this.binding.model),this.ws.on("message",t=>{this.onMessage(t)}),this.ws.on("close",()=>this.teardown("client disconnected")),this.ws.on("error",()=>this.teardown("client socket error")),this.send({type:"session.created",event_id:r("event"),session:j(this.config,this.sessionId)}),this.send({type:"conversation.created",event_id:r("event"),conversation:{id:this.conversationId,object:"realtime.conversation"}}),this.binding.turns&&this.startTurnLane()}startTurnLane(){let e=new AbortController;this.turnLaneAbort=e;let t=this.binding.turns;(async()=>{try{for await(let i of t()){if(this.closed||e.signal.aborted)return;this.onTurnEvent(i)}}catch(i){!this.closed&&!e.signal.aborted&&this.send(m(E("turn_lane_fault",`the backend turn/transcript lane failed: ${i instanceof Error?i.message:String(i)}`)))}})()}onTurnEvent(e){switch(e.kind){case"delta":case"final":{let t=typeof e.text=="string"?e.text:"";if(t.length===0)return;this.turnTranscript.open=!0,this.turnTranscript.text+=t,this.config.inputAudioTranscription&&this.send({type:"conversation.item.input_audio_transcription.delta",event_id:r("event"),item_id:"input_audio_turn",content_index:0,delta:t});return}case"turn_start":{this.turnTranscript={open:!0,text:this.turnTranscript.text},this.send({type:"input_audio_buffer.speech_started",event_id:r("event"),item_id:"input_audio_turn",audio_start_ms:typeof e.startMs=="number"?e.startMs:0}),this.config.turnDetection?.interruptResponse&&this.inFlight&&this.cancelResponse();return}case"turn_end":{let t=this.turnTranscript.text.trim(),i=t.length>0?t:null;this.turnTranscript={open:!1,text:""},this.lastCommittedTranscript=i,this.frozenAudio=this.pendingAudioBytes.splice(0),this.pendingAudioTotalBytes=0;let n=r("item"),s=this.items.length>0?this.items[this.items.length-1].id:null;this.pushItem({id:n,type:"message",role:"user",content:[{type:"input_audio",...i!==null?{transcript:i}:{}}]}),this.send({type:"input_audio_buffer.committed",event_id:r("event"),previous_item_id:s,item_id:n}),this.send({type:"conversation.item.created",event_id:r("event"),previous_item_id:s,item:this.itemToProtocol(this.items[this.items.length-1])}),this.send({type:"input_audio_buffer.speech_stopped",event_id:r("event"),item_id:n,audio_end_ms:typeof e.startMs=="number"?e.startMs:0}),this.config.inputAudioTranscription&&i!==null&&this.send({type:"conversation.item.input_audio_transcription.completed",event_id:r("event"),item_id:n,content_index:0,transcript:i}),this.config.turnDetection?.createResponse&&!this.inFlight&&this.startResponse({});return}case"turn_eager_end":case"turn_resumed":return;default:return}}takeTurnTranscript(){let e=this.lastCommittedTranscript;return this.lastCommittedTranscript=null,e}async onMessage(e){if(this.closed)return;let t;try{t=e.toString()}catch{this.fail(f("decode_error","frame is not decodable text"),"invalid_frame");return}if(Buffer.byteLength(t,"utf8")>F){this.fail(f("frame_too_large",`frame exceeds the ${F}-byte limit`),"invalid_frame");return}let i;try{i=JSON.parse(t)}catch{this.fail(f("invalid_json","frame is not valid JSON"),"invalid_frame");return}if(typeof i.type!="string"||i.type.length===0){this.fail(f("missing_type",'event is missing a string "type"'),"invalid_frame");return}try{await this.handle(i)}catch(n){let s=n;if(s&&typeof s=="object"&&s.name==="SessionUpdateError"){this.send(m({type:"invalid_request_error",code:s.code??"invalid_value",message:s.message,param:s.param??null}));return}this.fail(E(null,n instanceof Error?n.message:String(n)),"handler_fault")}}async handle(e){switch(e.type){case"session.update":this.config=S(e.session,this.config,{voice:this.binding.voice,turns:this.binding.turns!==void 0}),this.send({type:"session.updated",event_id:r("event"),session:j(this.config,this.sessionId)});return;case"conversation.item.create":{let t=q(e.item,this.binding.voice),i=this.items.length>0?this.items[this.items.length-1].id:null;this.pushItem(t),this.send({type:"conversation.item.created",event_id:r("event"),previous_item_id:i,item:this.itemToProtocol(t)});return}case"conversation.item.delete":{let t=this.items.findIndex(i=>i.id===e.item_id);if(t<0){this.send(m(f("item_not_found",`no item with id ${JSON.stringify(e.item_id)}`,"item_id")));return}this.items.splice(t,1),this.send({type:"conversation.item.deleted",event_id:r("event"),item_id:e.item_id});return}case"conversation.item.retrieve":{let t=this.items.find(i=>i.id===e.item_id);if(!t){this.send(m(f("item_not_found",`no item with id ${JSON.stringify(e.item_id)}`,"item_id")));return}this.send({type:"conversation.item.retrieved",event_id:r("event"),item:this.itemToProtocol(t)});return}case"conversation.item.truncate":{let t=this.items.find(i=>i.id===e.item_id);if(!t||t.type!=="message"||t.role!=="assistant"){this.send(m(f("item_not_found",`no assistant message item with id ${JSON.stringify(e.item_id)}`,"item_id")));return}t.truncated=!0,this.send({type:"conversation.item.truncated",event_id:r("event"),item_id:t.id,content_index:typeof e.content_index=="number"?e.content_index:0,audio_end_ms:typeof e.audio_end_ms=="number"?e.audio_end_ms:0});return}case"input_audio_buffer.append":{if(typeof e.audio!="string"||e.audio.length===0){this.send(m(f("invalid_audio",'input_audio_buffer.append requires base64 "audio"',"audio")));return}if(!this.ensureVoice())return;if(!pe.test(e.audio)){this.send(m(f("invalid_audio","input_audio_buffer.append audio must be base64","audio")));return}let t=Buffer.byteLength(e.audio,"base64");if(this.pendingAudioBytes.length+t>P){this.send(m(f("buffer_full",`uncommitted input audio exceeds the ${P}-byte bound \u2014 commit or clear first`,"audio")));return}if(this.inFlight){this.binding.audioSession.audio.appendInputAudio(e.audio);return}this.pendingAudioBytes.push(e.audio),this.pendingAudioTotalBytes+=t;return}case"input_audio_buffer.clear":{if(!this.ensureVoice())return;this.pendingAudioBytes.length=0,this.pendingAudioTotalBytes=0,this.send({type:"input_audio_buffer.cleared",event_id:r("event")});return}case"input_audio_buffer.commit":{if(!this.ensureVoice())return;if(this.pendingAudioTotalBytes===0&&this.takeTurnTranscript()===null){this.send(m(f("empty_buffer","input_audio_buffer.commit with no buffered audio","audio")));return}this.frozenAudio=this.pendingAudioBytes.splice(0),this.pendingAudioTotalBytes=0;let t=this.takeTurnTranscript();t===null&&this.binding.turns===void 0&&this.warnAudioOnlyTranscriptGap();let i=r("item"),n=this.items.length>0?this.items[this.items.length-1].id:null;this.pushItem({id:i,type:"message",role:"user",content:[{type:"input_audio",...t!==null?{transcript:t}:{}}]}),this.send({type:"input_audio_buffer.committed",event_id:r("event"),previous_item_id:n,item_id:i}),this.send({type:"conversation.item.created",event_id:r("event"),previous_item_id:n,item:this.itemToProtocol(this.items[this.items.length-1])});return}case"output_audio_buffer.clear":{this.send(m(f("output_clear_unavailable","output_audio_buffer.clear has no server-side playback queue to flush on this adapter: output audio is delivered as response.output_audio.delta events and interrupted turns are stopped with response.cancel")));return}case"response.create":this.startResponse(e.response??{});return;case"response.cancel":this.cancelResponse();return;default:this.send(m(f("unknown_event",`event type ${e.type} is not supported by this adapter`)))}}ensureVoice(){return!this.binding.voice||!this.binding.audioSession?(this.send(m(f("audio_not_supported","audio events require a voice binding from the resolver \u2014 this session was resolved text-only","audio"))),!1):(this.voiceClaimed||(this.claimVoiceBinding(),this.audioUnsubscribe=this.binding.audioSession.audio.onOutputAudio(e=>{this.onOutputAudioFrame(this.activeResponseId??"",e)}),this.voiceClaimed=!0),!0)}claimVoiceBinding(){let e=this.binding.audioSession,t=e.openaiRealtimeClaims??(e.openaiRealtimeClaims=new Set);if(t.has(e))throw new Error("voice binding already owned by another connection \u2014 the resolver handed one audio bridge to two realtime conversations, which must never happen (independent callers require private bindings)");t.add(e),this.audioLaneOwner=!0}releaseVoiceBinding(){let e=this.binding.audioSession;!e||!this.audioLaneOwner||(e.openaiRealtimeClaims?.delete(e),this.audioLaneOwner=!1)}warnAudioOnlyTranscriptGap(){this.audioOnlyTranscriptGapWarned||(this.audioOnlyTranscriptGapWarned=!0,this.send(m({type:"invalid_request_error",code:"audio_input_transcript_unavailable",message:"committed input audio is forwarded to the runtime audio lane but is NOT model-visible in the text turn lane: the resolved backend wires no turn/transcript lane (the voice engine's own stt stream with turn_start/turn_end + transcripts). Wire the transcribe/turn-detection leg upstream (the resolver supplies binding.turns) and committed audio gains its REAL transcript as the model-visible turn.",param:"input_audio_buffer.commit"})))}upstreamStream(e,t){let i=this.binding.session;if(i&&typeof i.sendResponseCreate=="function")return i.sendResponseCreate(e,e.input??[],t);let n=this.binding.clients;if(n&&typeof n.createResponse=="function")return n.createResponse(e);throw new Error("realtime binding provides neither session.sendResponseCreate nor clients.createResponse \u2014 no upstream turn seam")}startResponse(e){if(this.inFlight){this.send(m(f("response_in_progress","a response is already in progress \u2014 cancel it with response.cancel first","response.create")));return}let t=e??{};if(t.modalities!==void 0){if(!Array.isArray(t.modalities)||t.modalities.some(h=>h!=="text"&&h!=="audio")){this.send(m(f("invalid_modalities",'response.modalities must be "text"/"audio"',"response.modalities")));return}if(t.modalities.includes("audio")&&!this.ensureVoice())return}if(t.input!==void 0){this.send(m(f("response_input_unsupported","response.input is not supported by this adapter: build the context with conversation.item.create, then response.create","response.input")));return}this.cancelled=!1,this.responseFlushed=!1,this.messageItemAdded=!1;let i=r("req"),n=`resp_${i}`;this.activeResponseId=n;let s=this.items.length,{input:p,sawAudioOnlyItem:o}=L(this.items),l=typeof t.instructions=="string"?t.instructions:this.config.instructions,d=typeof t.temperature=="number"?t.temperature:this.config.temperature,c=t.max_output_tokens??this.config.maxOutputTokens,y=c==="inf"||c===void 0?void 0:c,x=t.tools!==void 0?t.tools:this.config.tools,C=t.tool_choice!==void 0?t.tool_choice:this.config.toolChoice,N={input:p,stream:!0,...this.config.model?{model:this.config.model}:{},...l!==void 0?{instructions:l}:{},...d!==void 0?{temperature:d}:{},...y!==void 0?{max_output_tokens:y}:{},...x!==void 0?{tools:x}:{},...C!==void 0?{tool_choice:C}:{}};for(let h of this.frozenAudio.splice(0))this.binding.audioSession?.audio.appendInputAudio(h);this.inFlight=(async()=>{let h=[],A=w=>{this.responseFlushed?this.send(w):h.push(w)},I=()=>{if(!this.responseFlushed){this.send({type:"response.created",event_id:r("event"),response:this.responseObject(n,"in_progress")}),this.send({type:"response.in_progress",event_id:r("event"),response:this.responseObject(n,"in_progress")});for(let w of h)this.send(w);h.length=0,this.responseFlushed=!0}},v={messageOpen:!1,messageText:"",tools:new Map,nextOutputIndex:0},Y=null;try{o&&this.binding.turns===void 0&&this.warnAudioOnlyTranscriptGap();let w;try{w=this.upstreamStream(N,i)}catch(O){this.emitUpstreamFault(O,n);return}try{for await(let O of w){if(this.cancelled)break;let _=O;if(_.type==="error"){let T=this.upstreamErrorBody(_);this.responseFlushed?this.send(m(T)):this.send({type:"response.failed",event_id:r("event"),response:this.failedResponse(n,T)});return}if(_.type==="response.output_text.delta"&&typeof _.delta=="string"){I(),v.messageOpen||(this.beginMessageItem(n,v.nextOutputIndex++),v.messageOpen=!0),A({type:"response.output_text.delta",event_id:r("event"),item_id:`${n}_msg_0`,response_id:n,output_index:0,content_index:0,delta:_.delta}),v.messageText+=_.delta;continue}if(_.type==="response.function_call_arguments.delta"){I();let T=typeof _.delta=="string"?_.delta:"",k=typeof _.call_id=="string"&&_.call_id.length>0?_.call_id:`${n}_fc_${v.nextOutputIndex}`,b=v.tools.get(k);if(!b){let z=typeof _.name=="string"&&_.name.length>0?_.name:void 0;b={outputIndex:v.nextOutputIndex++,name:z,args:""},v.tools.set(k,b),A({type:"response.output_item.added",event_id:r("event"),response_id:n,output_index:b.outputIndex,item:{id:k,object:"realtime.item",type:"function_call",call_id:k,name:z??"",arguments:"",status:"in_progress"}})}typeof _.name=="string"&&_.name.length>0&&(b.name=_.name),b.args+=T,A({type:"response.function_call_arguments.delta",event_id:r("event"),item_id:k,response_id:n,output_index:b.outputIndex,call_id:k,name:b.name??"",delta:T});continue}if(_.type==="response.completed"){Y=_.response??null;continue}}}catch(O){this.emitUpstreamFault(O,n);return}if(this.cancelled)return;this.closeOpenItems(n,v,s),I(),this.send({type:"response.done",event_id:r("event"),response:this.doneResponse(n,s,Y)})}finally{this.inFlight=null,this.activeResponseId=null}})()}onOutputAudioFrame(e,t){if(this.cancelled)return;if(this.activeResponseId!==e){!this.outOfTurnAudioWarned&&!this.closed&&(this.outOfTurnAudioWarned=!0,this.send(m({type:"invalid_request_error",code:"out_of_turn_audio",message:"the runtime produced voice OUT outside a response turn; this adapter delivers output audio only within response.output_audio.delta of an open turn \u2014 wire server-driven turns (VAD) upstream or keep speech inside response turns"})));return}this.responseFlushed||(this.send({type:"response.created",event_id:r("event"),response:this.responseObject(e,"in_progress")}),this.send({type:"response.in_progress",event_id:r("event"),response:this.responseObject(e,"in_progress")}),this.responseFlushed=!0);let i=`${e}_msg_0`;this.messageItemOpenForAudio||(this.beginAudioMessageItem(e),this.messageItemOpenForAudio=!0),this.send({type:"response.output_audio.delta",event_id:r("event"),item_id:i,response_id:e,output_index:0,content_index:0,delta:t})}messageItemAdded=!1;messageItemOpenForAudio=!1;beginMessageItem(e,t){let i=`${e}_msg_0`;this.messageItemAdded||(this.messageItemAdded=!0,this.send({type:"response.output_item.added",event_id:r("event"),response_id:e,output_index:t,item:{id:i,object:"realtime.item",type:"message",role:"assistant",status:"in_progress",content:[]}})),this.send({type:"response.content_part.added",event_id:r("event"),item_id:i,response_id:e,output_index:t,content_index:0,part:{type:"output_text",text:""}})}beginAudioMessageItem(e){let t=`${e}_msg_0`;this.messageItemAdded||(this.messageItemAdded=!0,this.send({type:"response.output_item.added",event_id:r("event"),response_id:e,output_index:0,item:{id:t,object:"realtime.item",type:"message",role:"assistant",status:"in_progress",content:[]}})),this.send({type:"response.content_part.added",event_id:r("event"),item_id:t,response_id:e,output_index:0,content_index:0,part:{type:"output_audio",transcript:""}})}closeOpenItems(e,t,i){if(t.messageOpen||this.messageItemOpenForAudio){let n=`${e}_msg_0`,s=this.messageItemOpenForAudio;s&&this.send({type:"response.output_audio.done",event_id:r("event"),item_id:n,response_id:e,output_index:0,content_index:0}),t.messageText.length>0&&this.send({type:"response.output_text.done",event_id:r("event"),item_id:n,response_id:e,output_index:0,content_index:0,text:t.messageText}),this.send({type:"response.content_part.done",event_id:r("event"),item_id:n,response_id:e,output_index:0,content_index:0,part:s?{type:"output_audio",transcript:""}:{type:"output_text",text:t.messageText}});let p=[];s&&p.push({type:"output_audio"}),t.messageText.length>0&&p.push({type:"output_text",text:t.messageText});let o={id:n,type:"message",role:"assistant",content:p};this.pushItem(o),this.send({type:"response.output_item.done",event_id:r("event"),response_id:e,output_index:0,item:this.itemToProtocol(o)})}for(let[n,s]of t.tools){let p=s.args.length>0?s.args:"{}";this.send({type:"response.function_call_arguments.done",event_id:r("event"),item_id:n,response_id:e,output_index:s.outputIndex,call_id:n,name:s.name??"",arguments:p});let o={id:n,type:"function_call",callId:n,name:s.name,arguments:p};this.pushItem(o),this.send({type:"response.output_item.done",event_id:r("event"),response_id:e,output_index:s.outputIndex,item:this.itemToProtocol(o)})}this.messageItemOpenForAudio=!1,this.messageItemAdded=!1}responseObject(e,t,i){return{id:e,object:"realtime.response",status:t,conversation_id:this.conversationId,usage:null,...i??{}}}failedResponse(e,t){return this.responseObject(e,"failed",{status_details:{error:{code:t.code,message:t.message}}})}doneResponse(e,t,i){let n=i&&typeof i.usage=="object"&&i.usage!==null?i.usage:null,s=[];for(let p=t;p<this.items.length;p++)s.push(this.itemToProtocol(this.items[p]));return this.responseObject(e,"completed",{output:s,usage:n?{input_tokens:typeof n.input_tokens=="number"&&Number.isFinite(n.input_tokens)?n.input_tokens:0,output_tokens:typeof n.output_tokens=="number"&&Number.isFinite(n.output_tokens)?n.output_tokens:0,total_tokens:typeof n.total_tokens=="number"&&Number.isFinite(n.total_tokens)?n.total_tokens:(typeof n.input_tokens=="number"&&Number.isFinite(n.input_tokens)?n.input_tokens:0)+(typeof n.output_tokens=="number"&&Number.isFinite(n.output_tokens)?n.output_tokens:0),input_token_details:{cached_tokens:0,audio_tokens:0,text_tokens:0},output_token_details:{audio_tokens:0,text_tokens:0}}:null})}upstreamErrorBody(e){let t=e.error,i=t&&typeof t.message=="string"?t.message:JSON.stringify(e);return{type:"server_error",code:t&&typeof t.code=="string"?t.code:null,message:i}}emitUpstreamFault(e,t){if(this.cancelled)return;let i=E(null,e instanceof Error?e.message:String(e));this.responseFlushed?this.send(m(i)):this.send({type:"response.failed",event_id:r("event"),response:this.failedResponse(t,i)})}cancelResponse(){let e=this.inFlight;if(!e){this.send(m(f("no_response_to_cancel","no response is in progress","response.cancel")));return}this.cancelled=!0;let t=this.activeResponseId??"resp_";e.finally(()=>{this.send({type:"response.cancelled",event_id:r("event"),response:this.responseObject(t,"cancelled")})})}pushItem(e){this.items.push(e),this.items.length>U&&(this.items.shift(),this.send(m({type:"invalid_request_error",code:"transcript_evicted",message:`the conversation transcript exceeded ${U} items \u2014 the oldest item was evicted to keep the per-connection bound`})))}itemToProtocol(e){let t={id:e.id,object:"realtime.item"};if(e.type==="function_call")return{...t,type:"function_call",call_id:e.callId,name:e.name,arguments:e.arguments??"{}",status:"completed"};if(e.type==="function_call_output")return{...t,type:"function_call_output",call_id:e.callId,output:e.output};let i=(e.content??[]).map(n=>n.type==="input_text"||n.type==="output_text"?{type:n.type,text:n.text}:n.type==="input_audio"?{type:"input_audio",...n.transcript?{transcript:n.transcript}:{}}:{type:"output_audio",transcript:n.transcript??""});return{...t,type:"message",role:e.role,status:e.truncated?"incomplete":"completed",content:i}}send(e){if(!(this.closed||this.ws.readyState!==this.ws.OPEN)){if(this.ws.bufferedAmount>$){this.ws.close(ue,X("outbound buffer exceeded")),this.teardown("outbound buffer exceeded");return}this.ws.send(JSON.stringify(e))}}fail(e,t){this.send(m(e)),this.ws.close(de,X(t)),this.teardown(t)}teardown(e){if(!this.closed){this.closed=!0,this.turnLaneAbort?.abort(),this.turnLaneAbort=null,this.releaseVoiceBinding(),this.audioUnsubscribe?.(),this.audioUnsubscribe=null;try{this.ws.removeAllListeners("message")}catch{}this.onClosed?.(this),le(this.id,e)}}},V=[];function le(a,e){V.push({id:a,reason:e}),V.length>256&&V.shift()}var G="/v1/realtime",me=a=>a.slice(0,120);function _e(a,e,t){let i=e===401||e===403?"Unauthorized":e===502?"Upstream Fault":"Upgrade Refused";a.write(`HTTP/1.1 ${e} ${i}\r
-Connection: close\r
-Content-Type: text/plain\r
-\r
-${me(t)}\r
-`),a.destroy()}function J(a,e){let t=new ce({noServer:!0}),i=new Set;return a.on("upgrade",(n,s,p)=>{if(new URL(n.url??"/","http://localhost").pathname!==G)return;i.add(s),s.on("close",()=>{i.delete(s)});let l=c=>{let y=c,x=y.session!=null&&typeof y.session.sendResponseCreate=="function",C=y.clients!=null&&typeof y.clients.createResponse=="function";if(!x&&!C)throw new Error("resolver returned a binding without an upstream turn seam (session.sendResponseCreate or clients.createResponse required) \u2014 refusing the upgrade LOUD");return new Promise((N,h)=>{t.handleUpgrade(n,s,p,A=>{try{new R({ws:A,binding:c}),N()}catch(I){h(I)}})})};(typeof e.resolver=="function"?e.resolver:e.resolver.resolve.bind(e.resolver))(n).then(l).catch(c=>{let y=c&&typeof c=="object"&&typeof c.status=="number"?c.status:502,x=c instanceof Error?c.message:String(c);s.destroyed||_e(s,y,x)}).finally(()=>{i.delete(s)})}),async()=>{let n=Date.now()+5e3;for(;i.size>0&&Date.now()<n;)await new Promise(s=>setTimeout(s,25));await new Promise(s=>t.close(()=>s()))}}export{Z as GEMINI_LIVE_PATH,G as OPENAI_REALTIME_PATH,R as OpenAIRealtimeConnection,ee as ResumptionRegistry,K as SessionGoneError,H as UnknownModelError,Q as VIDEO_OUT_SETUP_KEY,te as attachGeminiLive,J as attachOpenAIRealtime,ie as createOpenAIProxy,ne as serveGeminiLiveSocket};
+import{Q as I,a as n,b as p,c as m,f as l,g as a,h as x,i as s,j as O,k as f}from"../chunk-NY23USZF.js";import{d as r,e as i}from"../chunk-3WAJD62J.js";import{a as o,h as t}from"../chunk-7M6LY6DR.js";import"../chunk-BSHT6RZZ.js";import{c as e}from"../chunk-5CCJUNH7.js";e();export{o as GEMINI_LIVE_PATH,a as OPENAI_REALTIME_PATH,l as OpenAIRealtimeConnection,n as ResumptionRegistry,i as SessionGoneError,r as UnknownModelError,t as VIDEO_OUT_SETUP_KEY,p as attachGeminiLive,x as attachOpenAIRealtime,I as createOpenAIProxy,f as fixedClientsResolver,O as realtimeBindingFor,s as realtimeModelOf,m as serveGeminiLiveSocket};