perso-interactive-sdk-web 1.3.4 → 1.4.0

package/README.md

@@ -2,6 +2,10 @@
 
 WebRTC-based real-time interactive AI avatar SDK for web applications.
 
+> **API server:** Use `https://platform.perso.ai` as the Perso Interactive API server URL.
+>
+> Legacy `https://live-api.perso.ai` remains backward-compatible.
+
 ## Installation
 
 ```bash
@@ -47,7 +51,7 @@ const { createSessionId } = require("perso-interactive-sdk-web/server");
 
 const app = express();
 
-const API_SERVER = "https://live-api.perso.ai";
+const API_SERVER = "https://platform.perso.ai";
 const API_KEY = process.env.PERSO_INTERACTIVE_API_KEY;
 
 app.post("/api/session", async (req, res) => {
@@ -60,6 +64,8 @@ app.post("/api/session", async (req, res) => {
       tts_type: "<tts_name>",
       stt_type: "<stt_name>",
       // text_normalization_config: "<textnormalizationconfig_id>", // optional
+      // stt_text_normalization_config: "<textnormalizationconfig_id>", // optional
+      // stt_text_normalization_locale: "ko", // optional
     });
     res.json({ sessionId });
   } catch (error) {
@@ -79,6 +85,19 @@ If you have pre-configured session templates, pass the template ID directly inst
 const sessionId = await createSessionId(API_SERVER, API_KEY, "<sessiontemplate_id>");
 ```
 
+#### Listing available resources from the server
+
+Use `getAllSettings` (or any individual `getXxx` helper) on the server to discover the LLM/TTS/STT/model-style options that your tenant has access to without exposing the API key in the browser:
+
+```javascript
+const { getAllSettings } = require("perso-interactive-sdk-web/server");
+
+app.get("/api/settings", async (req, res) => {
+  const settings = await getAllSettings(API_SERVER, API_KEY);
+  res.json(settings); // { llms, ttsTypes, sttTypes, modelStyles, ... }
+});
+```
+
 > ⚠️ **Security Warning**: Never use `createSessionId` on the client-side in production. Exposing your API key in browser code can lead to unauthorized access and quota abuse. Always create sessions on the server and pass only the `sessionId` to the client.
 
 #### Client-side Testing Only
@@ -91,7 +110,7 @@ import {
   createSession,
 } from "perso-interactive-sdk-web/client";
 
-const apiServer = "https://live-api.perso.ai";
+const apiServer = "https://platform.perso.ai";
 const apiKey = "YOUR_API_KEY"; // ⚠️ NEVER commit or expose this in production
 
 const sessionId = await createSessionId(apiServer, apiKey, {
@@ -102,6 +121,8 @@ const sessionId = await createSessionId(apiServer, apiKey, {
   tts_type: "<tts_name>",
   stt_type: "<stt_name>",
   // text_normalization_config: "<textnormalizationconfig_id>", // optional
+  // stt_text_normalization_config: "<textnormalizationconfig_id>", // optional
+  // stt_text_normalization_locale: "ko", // optional
 });
 
 const session = await createSession(apiServer, sessionId, 1920, 1080, []);
@@ -123,7 +144,7 @@ import {
   ChatState,
 } from "perso-interactive-sdk-web/client";
 
-const apiServer = "https://live-api.perso.ai";
+const apiServer = "https://platform.perso.ai";
 
 // Obtain sessionId from your server (see Express.js example above)
 const sessionId = await fetch("/api/session", { method: "POST" })
@@ -169,7 +190,7 @@ const audioBlob = await session.processTTS(llmResponse);
 
 // 3. Animate avatar with audio
 if (audioBlob) {
-  await session.processSTF(audioBlob, "wav", llmResponse);
+  await session.processSTF(audioBlob, audioBlob.type, llmResponse);
 }
 ```
 
@@ -243,7 +264,7 @@ For direct browser usage via `<script>` tag without a bundler. The SDK exposes a
 <script src="https://cdn.jsdelivr.net/npm/perso-interactive-sdk-web@latest/dist/client/index.iife.js"></script>
 <script>
   async function start() {
-    const apiServer = "https://live-api.perso.ai";
+    const apiServer = "https://platform.perso.ai";
 
     // Obtain sessionId from your server (see Express.js example above)
     const sessionId = await fetch("/api/session", { method: "POST" })
@@ -278,15 +299,28 @@ For direct browser usage via `<script>` tag without a bundler. The SDK exposes a
 
 ### Server Exports
 
-| Export                                         | Description                    |
-| ---------------------------------------------- | ------------------------------ |
-| `createSessionId(apiServer, apiKey, sessionTemplateId)` | Create a session ID from a SessionTemplate |
-| `createSessionId(apiServer, apiKey, params)`   | Create a new session ID        |
-| `getIntroMessage(apiServer, apiKey, promptId)` | Get intro message for a prompt |
-| `getSessionTemplates(apiServer, apiKey)`                                           | Get available session templates                            |
-| `getSessionTemplate(apiServer, apiKey, sessionTemplateId)`                         | Get a single session template by ID                        |
-| `PersoUtilServer`                              | Low-level API utilities        |
-| `ApiError`                                     | Error class for API errors     |
+| Export                                                     | Description                                          |
+| ---------------------------------------------------------- | ---------------------------------------------------- |
+| `createSessionId(apiServer, apiKey, sessionTemplateId)`    | Create a session ID from a SessionTemplate           |
+| `createSessionId(apiServer, apiKey, params)`               | Create a new session ID                              |
+| `getIntroMessage(apiServer, apiKey, promptId)`             | Get intro message for a prompt                       |
+| `getLLMs(apiServer, apiKey)`                               | Get available LLM providers                          |
+| `getTTSs(apiServer, apiKey)`                               | Get available TTS providers                          |
+| `getSTTs(apiServer, apiKey)`                               | Get available STT providers                          |
+| `getModelStyles(apiServer, apiKey)`                        | Get available avatar styles                          |
+| `getBackgroundImages(apiServer, apiKey)`                   | Get available backgrounds                            |
+| `getPrompts(apiServer, apiKey)`                            | Get available prompts                                |
+| `getDocuments(apiServer, apiKey)`                          | Get available documents                              |
+| `getMcpServers(apiServer, apiKey)`                         | Get available MCP servers                            |
+| `getTextNormalizations(apiServer, apiKey)`                 | Get available text normalization configs             |
+| `getTextNormalization(apiServer, apiKey, configId)`        | Download text normalization ruleset (pre-signed URL) |
+| `getAllSettings(apiServer, apiKey)`                        | Get all settings at once                             |
+| `getSessionTemplates(apiServer, apiKey)`                   | Get available session templates                      |
+| `getSessionTemplate(apiServer, apiKey, sessionTemplateId)` | Get a single session template by ID                  |
+| `getSessionInfo(apiServer, sessionId)`                     | Get session metadata                                 |
+| `makeTTS(apiServer, params)`                               | Generate TTS audio from text (standalone)            |
+| `PersoUtilServer`                                          | Low-level API utilities                              |
+| `ApiError`                                                 | Error class for API errors                           |
 
 ### Client Exports
 
@@ -333,7 +367,7 @@ For direct browser usage via `<script>` tag without a bundler. The SDK exposes a
 | `processLLM(options)`               | Stream LLM responses with full control         |
 | `processTTSTF(message)`             | Speak a message without LLM                    |
 | `processTTS(message, options?)`     | Generate TTS audio from text (returns Blob). Options: `resample`, `locale`, `output_format` |
-| `processSTF(file, format, message)` | Send audio/video to STF pipeline                |
+| `processSTF(file, format?, message?)` | Send audio to the STF pipeline. `format` accepts canonical (`'wav'` / `'mp3'`) or MIME (`'audio/wav'`, `'audio/mpeg'`, …); when omitted, derived from `file.type` |
 | `startProcessSTT(timeout?)`         | Start recording voice for STT                  |
 | `stopProcessSTT(language?)`         | Stop recording and get text                    |
 | `isSTTRecording()`                  | Check if STT recording is in progress          |

package/dist/client/index.cjs

@@ -1 +1 @@
-"use strict";var t,e,s=require("emoji-regex");class a extends Error{constructor(){super("WebRTC connection timeout")}}class r extends Error{errorCode;code;detail;attr;constructor(t,e,s,a){let r;r=null!=a?`${t}:${a}_${s}`:`${t}:${s}`,super(r),this.errorCode=t,this.code=e,this.detail=s,this.attr=a}}class n extends Error{underlyingError;constructor(t){super(),this.underlyingError=t}}class o extends Error{description;constructor(t){super(),this.description=t}}class i extends Error{underlyingError;constructor(t){super(`STT Error: ${t.detail}`),this.underlyingError=t}}class l extends Error{underlyingError;constructor(t){super(t.message),this.underlyingError=t}}class c extends Error{description;constructor(t){super(`TTS decode error: ${t}`),this.description=t}}!function(t){t.LLM="LLM",t.TTS="TTS",t.STT="STT",t.STF_ONPREMISE="STF_ONPREMISE",t.STF_WEBRTC="STF_WEBRTC"}(t||(t={})),function(t){t.SESSION_START="SESSION_START",t.SESSION_DURING="SESSION_DURING",t.SESSION_LOG="SESSION_LOG",t.SESSION_END="SESSION_END",t.SESSION_ERROR="SESSION_ERROR",t.SESSION_TTS="SESSION_TTS",t.SESSION_STT="SESSION_STT",t.SESSION_LLM="SESSION_LLM"}(e||(e={}));class h{static async getLLMs(t,e){const s=fetch(`${t}/api/v1/settings/llm_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getModelStyles(t,e){const s=fetch(`${t}/api/v1/settings/modelstyle/?platform_type=webrtc`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getBackgroundImages(t,e){const s=fetch(`${t}/api/v1/background_image/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getTTSs(t,e){const s=fetch(`${t}/api/v1/settings/tts_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getSTTs(t,e){const s=fetch(`${t}/api/v1/settings/stt_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async makeTTS(t,{sessionId:e,text:s,locale:a,output_format:r}){const n={text:s};a&&(n.locale=a),r&&(n.output_format=r);const o=await fetch(`${t}/api/v1/session/${e}/tts/`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify(n)});return await this.parseJson(o)}static async getPrompts(t,e){const s=fetch(`${t}/api/v1/prompt/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getDocuments(t,e){const s=fetch(`${t}/api/v1/document/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getTextNormalizations(t,e){const s=fetch(`${t}/api/v1/settings/text_normalization_config/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async downloadTextNormalization(t,e,s){const a=await fetch(`${t}/api/v1/settings/text_normalization_config/${s}/download/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(a)}static async getSessionTemplates(t,e){const s=await fetch(`${t}/api/v1/session_template/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(s)}static async getSessionTemplate(t,e,s){const a=await fetch(`${t}/api/v1/session_template/${s}/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(a)}static async getMcpServers(t,e){const s=fetch(`${t}/api/v1/settings/mcp_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getSessionInfo(t,e){const s=fetch(`${t}/api/v1/session/${e}/`,{method:"GET"}),a=await s;return await this.parseJson(a)}static async sessionEvent(t,e,s,a=""){const r=await fetch(`${t}/api/v1/session/${e}/event/create/`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify({detail:a,event:s})});await this.parseJson(r)}static async makeSTT(t,e,s,a){const r=new FormData;r.append("audio",s),a&&r.append("language",a);const n=await fetch(`${t}/api/v1/session/${e}/stt/`,{method:"POST",body:r});return await this.parseJson(n)}static async makeLLM(t,e,s,a){const n=await fetch(`${t}/api/v1/session/${e}/llm/v2/`,{body:JSON.stringify(s),headers:{"Content-Type":"application/json"},method:"POST",signal:a});if(!n.ok){const t=await n.json(),e=t.errors?.[0]??{code:"UNKNOWN_ERROR",detail:`Server returned status ${n.status} with no error details`,attr:null};throw new r(n.status,e.code,e.detail,e.attr)}return n.body.getReader()}static async getIceServers(t,e){const s=await fetch(`${t}/api/v1/session/${e}/ice-servers/`,{method:"GET"});return(await this.parseJson(s)).ice_servers}static async exchangeSDP(t,e,s){const a=await fetch(`${t}/api/v1/session/${e}/exchange/`,{body:JSON.stringify({client_sdp:s}),headers:{"Content-Type":"application/json"},method:"POST"});return(await this.parseJson(a)).server_sdp}static async parseJson(t){const e=await t.json();if(t.ok)return e;{const s=e.errors?.[0]??{code:"UNKNOWN_ERROR",detail:`Server returned status ${t.status} with no error details`,attr:null};throw new r(t.status,s.code,s.detail,s.attr)}}}class d extends Error{constructor(t){super(`WAV parse error: ${t}`),this.name="WavParseError"}}function u(t,e,s=1){const a=2*t.length,r=new ArrayBuffer(44+a),n=new DataView(r);g(n,0,"RIFF"),n.setUint32(4,36+a,!0),g(n,8,"WAVE"),g(n,12,"fmt "),n.setUint32(16,16,!0),n.setUint16(20,1,!0),n.setUint16(22,s,!0),n.setUint32(24,e,!0),n.setUint32(28,e*s*2,!0),n.setUint16(32,2*s,!0),n.setUint16(34,16,!0),g(n,36,"data"),n.setUint32(40,a,!0);let o=44;for(let e=0;e<t.length;e++){const s=Math.max(-1,Math.min(1,t[e]));n.setInt16(o,s<0?32768*s:32767*s,!0),o+=2}return r}function p(t,e,s){let a="";for(let r=0;r<s;r++)a+=String.fromCharCode(t.getUint8(e+r));return a}function g(t,e,s){for(let a=0;a<s.length;a++)t.setUint8(e+a,s.charCodeAt(a))}function m(t,e,s){switch(s){case 8:return(t.getUint8(e)-128)/128;case 16:return t.getInt16(e,!0)/32768;case 24:{const s=t.getUint8(e),a=t.getUint8(e+1),r=t.getUint8(e+2)<<16|a<<8|s;return(r>8388607?r-16777216:r)/8388608}case 32:return t.getInt32(e,!0)/2147483648;default:return 0}}class S extends Error{constructor(t){super(`Audio resample error: ${t}`),this.name="AudioResampleError"}}async function f(t,e,s,a=1){if(0===t.length)throw new S("Cannot resample empty audio data");if(e<=0||s<=0)throw new S(`Invalid sample rate: original=${e}, target=${s}`);if(e===s)return t;try{const r=t.length/e,n=Math.ceil(r*s),o=new OfflineAudioContext(a,t.length,e).createBuffer(a,t.length,e);o.getChannelData(0).set(t);const i=new OfflineAudioContext(a,n,s),l=i.createBufferSource();l.buffer=o,l.connect(i.destination),l.start(0);return(await i.startRendering()).getChannelData(0)}catch(t){const a=t instanceof Error?t.message:String(t);throw new S(`Failed to resample audio from ${e}Hz to ${s}Hz: ${a}`)}}const y=16e3;async function w(t,e=!0){let s;try{const e=atob(t),a=new Array(e.length);for(let t=0;t<e.length;t++)a[t]=e.charCodeAt(t);s=new Uint8Array(a).buffer}catch{throw new c("Invalid Base64 audio data")}const a=function(t){const e=new Uint8Array(t);if(e.length>=4){const t=String.fromCharCode(e[0],e[1],e[2],e[3]);if("RIFF"===t)return"audio/wav";if(t.startsWith("ID3"))return"audio/mpeg";if(255===e[0]&&!(224&~e[1]))return"audio/mpeg"}return"audio/wav"}(s);if(!e)return new Blob([s],{type:a});try{const t=await async function(t,e){if("audio/wav"===e){const e=function(t){const e=new DataView(t);if(t.byteLength<44)throw new d("File too small to be a valid WAV");if("RIFF"!==p(e,0,4))throw new d("Missing RIFF header");if("WAVE"!==p(e,8,4))throw new d("Missing WAVE format identifier");let s=12,a=!1,r=0,n=0,o=0,i=0;for(;s<t.byteLength-8;){const l=p(e,s,4),c=e.getUint32(s+4,!0);if("fmt "===l){if(s+24>t.byteLength)throw new d("fmt chunk extends beyond file");r=e.getUint16(s+8,!0),n=e.getUint16(s+10,!0),o=e.getUint32(s+12,!0),i=e.getUint16(s+22,!0),a=!0,s+=8+c;break}const h=s+8+c;if(h<=s)break;s=h}if(!a)throw new d("Missing fmt chunk");if(1!==r)throw new d(`Unsupported audio format: ${r} (only PCM format 1 is supported)`);if(0===n||n>8)throw new d(`Invalid channel count: ${n}`);if(0===o||o>192e3)throw new d(`Invalid sample rate: ${o}`);if(![8,16,24,32].includes(i))throw new d(`Unsupported bits per sample: ${i}`);let l=-1,c=0;for(;s<t.byteLength-8;){const t=p(e,s,4),a=e.getUint32(s+4,!0);if("data"===t){l=s+8,c=a;break}const r=s+8+a;if(r<=s)break;s=r}if(-1===l)throw new d("Missing data chunk");const h=t.byteLength-l,u=Math.min(c,h),g=i/8,S=Math.floor(u/(g*n)),f=new Float32Array(S*n);let y=0;for(let s=0;s<S*n;s++){const a=l+s*g;if(a+g>t.byteLength)break;f[y++]=Math.max(-1,Math.min(1,m(e,a,i)))}if(2===n){const t=new Float32Array(S);for(let e=0;e<S;e++)t[e]=(f[2*e]+f[2*e+1])/2;return{sampleRate:o,channels:1,bitsPerSample:i,samples:t}}return{sampleRate:o,channels:n,bitsPerSample:i,samples:f.slice(0,y)}}(t);if(e.sampleRate===y)return{samples:e.samples,sampleRate:e.sampleRate};return{samples:await f(e.samples,e.sampleRate,y,e.channels),sampleRate:y}}const s=new AudioContext({sampleRate:y});try{const e=await s.decodeAudioData(t.slice(0));if(e.sampleRate===y)return{samples:e.getChannelData(0),sampleRate:y};return{samples:await f(e.getChannelData(0),e.sampleRate,y,1),sampleRate:y}}finally{await s.close()}}(s,a),e=u(t.samples,y,1);return new Blob([e],{type:"audio/wav"})}catch{return new Blob([s],{type:a})}}const T=s();function C(t){return t.replace(T,"")}var E;exports.ChatState=void 0,(E=exports.ChatState||(exports.ChatState={})).RECORDING="RECORDING",E.LLM="LLM",E.ANALYZING="ANALYZING",E.SPEAKING="SPEAKING",E.TTS="TTS";class v{config;messageHistory=[];constructor(t){this.config=t}async*processLLM(t){if(0===t.message.length)throw new Error("Message cannot be empty");const e=t.tools??this.config.clientTools,s=e.map(t=>({type:"function",function:{description:t.description,name:t.name,parameters:t.parameters}})),a={newMessageHistory:[{role:"user",content:t.message}],allChunks:[],message:"",lastYieldedChunkCount:0,pendingToolCallsMessage:null,aborted:!1,streamingError:null};let i=0,l=[...this.messageHistory,...a.newMessageHistory];this.config.callbacks.onChatStateChange(exports.ChatState.LLM,null);try{for(;;){if(t.signal?.aborted)return void(a.allChunks.length>0&&(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0}));let c;try{c=await h.makeLLM(this.config.apiServer,this.config.sessionId,{messages:l,tools:s},t.signal)}catch(t){if(t instanceof r)return void(yield{type:"error",error:new n(t)});throw t}if(a.streamingError=null,yield*this.parseSSEStream(c,a,t),a.streamingError)return;if(a.aborted)return void(a.allChunks.length>0&&(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0}));if(null!=a.pendingToolCallsMessage){yield*this.executeToolCalls(a,e);const t=a.lastToolCallResults,s=t.length>0&&a.pendingToolCallsMessage.tool_calls.length!==t.length,r=t.some(t=>!t.chatTool.executeOnly);if(s||r){if(i++,i>=10)return void(yield{type:"error",error:new n(new o("Tool follow-up loop exceeded maximum rounds (10)"))});l=[...this.messageHistory,...a.newMessageHistory],a.pendingToolCallsMessage=null;continue}}return this.messageHistory.push(...a.newMessageHistory),void(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0})}}finally{this.config.callbacks.onChatStateChange(null,exports.ChatState.LLM)}}async*parseSSEStream(t,e,s){const a=new TextDecoder("utf-8");let r="",i="";e.pendingToolCallsMessage=null;const l=()=>e.allChunks.length>e.lastYieldedChunkCount?(e.lastYieldedChunkCount=e.allChunks.length,{type:"message",chunks:[...e.allChunks],message:e.message,finish:!1}):null;for(;;){const{done:c,value:h}=await t.read();if(c)break;let d;for(r+=a.decode(h,{stream:!0});-1!==(d=r.indexOf("\n"));){if(s.signal?.aborted)return void(e.aborted=!0);const t=r.slice(0,d).trim();if(r=r.slice(d+1),!t.startsWith("data: {"))return e.streamingError=new n(new o("Failed to parse SSE response")),void(yield{type:"error",error:e.streamingError});let a;try{a=JSON.parse(t.slice(6).trim())}catch{return e.streamingError=new n(new o("Failed to parse SSE JSON")),void(yield{type:"error",error:e.streamingError})}if("success"!==a.status)return e.streamingError=new n(new o(a.reason)),void(yield{type:"error",error:e.streamingError});if(i.length>0&&"message"!=a.type){e.newMessageHistory.push({role:"assistant",type:"message",content:i}),i="";const t=l();t&&(yield t)}if("message"===a.type){const t=C(a.content);i+=t,e.message+=t,e.allChunks.push(t);continue}"tool_call"!==a.type||null==a.tool_calls?"tool"!==a.role||"tool_call"===a.type&&e.newMessageHistory.push({role:a.role,type:a.type,content:a.content,tool_call_id:a.tool_call_id}):(e.newMessageHistory.push({role:"assistant",type:a.type,content:a.content,tool_calls:a.tool_calls}),e.pendingToolCallsMessage=a,yield{type:"tool_call",tool_calls:a.tool_calls})}const u=l();u&&(yield u)}const c=r.trim();if(c.length>0){if(!c.startsWith("data: {"))return e.streamingError=new n(new o("Failed to parse SSE response")),void(yield{type:"error",error:e.streamingError});let t;try{t=JSON.parse(c.slice(6).trim())}catch{return e.streamingError=new n(new o("Failed to parse SSE JSON")),void(yield{type:"error",error:e.streamingError})}if("success"!==t.status)return e.streamingError=new n(new o(t.reason)),void(yield{type:"error",error:e.streamingError});if("message"===t.type){const s=C(t.content);i+=s,e.message+=s,e.allChunks.push(s)}else if("tool_call"===t.type&&null!=t.tool_calls){if(i.length>0){e.newMessageHistory.push({role:"assistant",type:"message",content:i}),i="";const t=l();t&&(yield t)}e.newMessageHistory.push({role:"assistant",type:t.type,content:t.content,tool_calls:t.tool_calls}),e.pendingToolCallsMessage=t,yield{type:"tool_call",tool_calls:t.tool_calls}}}i.length>0&&e.newMessageHistory.push({role:"assistant",type:"message",content:i})}async*executeToolCalls(t,e){const s=t=>{for(const s of e)if(s.name===t)return s;return null},a=[];for(const e of t.pendingToolCallsMessage.tool_calls){const t=s(e.function.name);null!=t&&a.push((async()=>{try{const s=await t.call(JSON.parse(e.function.arguments));return{toolCallId:e.id,chatTool:t,chatToolResult:s}}catch(s){return{toolCallId:e.id,chatTool:t,chatToolResult:{error:s.message}}}})())}const r=await Promise.all(a);t.lastToolCallResults=r;for(const e of r)t.newMessageHistory.push({role:"tool",content:JSON.stringify(e.chatToolResult),tool_call_id:e.toolCallId}),yield{type:"tool_result",tool_call_id:e.toolCallId,result:e.chatToolResult}}addToHistory(t){this.messageHistory.push(t)}getHistory(){return this.messageHistory}}const I=`data:application/javascript,${encodeURIComponent("\nclass RecorderProcessor extends AudioWorkletProcessor {\n  constructor() {\n    super();\n    this.isRecording = true;\n    \n    // Listen for stop message from main thread\n    this.port.onmessage = (event) => {\n      if (event.data.type === 'stop') {\n        this.isRecording = false;\n        // Send confirmation back to main thread\n        this.port.postMessage({ type: 'stopped' });\n      }\n    };\n  }\n\n  process(inputs, outputs, parameters) {\n    const input = inputs[0];\n    if (input && input.length > 0 && this.isRecording) {\n      // Clone the audio data and send to main thread\n      const channelData = new Float32Array(input[0]);\n      this.port.postMessage({ type: 'audio', data: channelData });\n    }\n    // Return true to keep the processor alive until stopped\n    return this.isRecording;\n  }\n}\n\nregisterProcessor('recorder-processor', RecorderProcessor);\n")}`;class R{audioContext=null;mediaStream=null;workletNode=null;sourceNode=null;audioChunks=[];isRecordingState=!1;channels;targetSampleRate;constructor(t={}){this.channels=t.channels||1,this.targetSampleRate=t.targetSampleRate}async start(){if(this.isRecordingState)throw new Error("WavRecorder is already recording");if(this.mediaStream=await navigator.mediaDevices.getUserMedia({audio:!0}),this.audioContext=new AudioContext,"running"!==this.audioContext.state)try{await this.audioContext.resume()}catch(t){console.warn("WavRecorder: Failed to resume AudioContext:",t)}await this.audioContext.audioWorklet.addModule(I),this.sourceNode=this.audioContext.createMediaStreamSource(this.mediaStream),this.workletNode=new AudioWorkletNode(this.audioContext,"recorder-processor"),this.audioChunks=[],this.workletNode.port.onmessage=t=>{"audio"===t.data.type&&this.audioChunks.push(t.data.data)},this.sourceNode.connect(this.workletNode),this.workletNode.connect(this.audioContext.destination),this.isRecordingState=!0}async stop(){if(!this.isRecordingState)throw new Error("WavRecorder is not recording");this.isRecordingState=!1,await new Promise(t=>{this.workletNode.port.onmessage=e=>{"stopped"===e.data.type?t():"audio"===e.data.type&&this.audioChunks.push(e.data.data)},this.workletNode.port.postMessage({type:"stop"})}),this.workletNode?.disconnect(),this.sourceNode?.disconnect(),this.mediaStream?.getTracks().forEach(t=>t.stop());const t=this.audioChunks,e=this.audioContext.sampleRate;try{const s=t.reduce((t,e)=>t+e.length,0),a=new Float32Array(s);let r,n,o=0;for(const e of t)a.set(e,o),o+=e.length;this.targetSampleRate&&this.targetSampleRate!==e?(r=await f(a,e,this.targetSampleRate,this.channels),n=this.targetSampleRate):(r=a,n=e);const i=u(r,n,this.channels),l=new Blob([i],{type:"audio/wav"});return new File([l],"recording.wav",{type:"audio/wav"})}finally{await(this.audioContext?.close()),this.audioContext=null,this.mediaStream=null,this.workletNode=null,this.sourceNode=null,this.audioChunks=[]}}isRecording(){return this.isRecordingState}}class _ extends EventTarget{pc;dc;streams=[];pingTime;pingIntervalId=null;constructor(t,e){super(),this.pc=t,this.dc=e,this.pingTime=Date.now()+3e3,this.pc.addEventListener("track",t=>{this.streams=this.streams.concat(t.streams)}),this.pc.addEventListener("connectionstatechange",()=>{"disconnected"!==this.pc.connectionState&&"failed"!==this.pc.connectionState||this.close()}),this.dc.onopen=()=>{this.pingIntervalId=setInterval(()=>{this.ping(),Date.now()-this.pingTime>5e3&&this.close()},1e3)},this.dc.onclose=()=>{null!=this.pingIntervalId&&clearInterval(this.pingIntervalId)},this.#t({live:!0,code:200,reason:"OK"}),this.setMessageCallback("ping",()=>{this.pingTime=Date.now()})}static async create(s,a,r,n,o){const i=await h.getSessionInfo(s,a);if(!(Array.isArray(i.capability)&&i.capability.some(e=>e.name===t.STF_ONPREMISE||e.name===t.STF_WEBRTC)))return await h.sessionEvent(s,a,e.SESSION_START),null;const l=await h.getIceServers(s,a);let c=await _.createPeerConnection(l),d=c.createDataChannel("message",{protocol:"message"}),u=new _(c,d);o?o.getTracks().forEach(function(t){c.addTrack(t,o)}):c.addTransceiver("audio",{direction:"recvonly"});const p=c.addTransceiver("video",{direction:"recvonly"}),g=RTCRtpReceiver.getCapabilities("video");null!=g&&p.setCodecPreferences(g.codecs);const m=await c.createOffer();await c.setLocalDescription(m);const S=await h.exchangeSDP(s,a,m);return await c.setRemoteDescription(S),await _.waitFor(()=>u.isReady(),100,50),u.changeSize(r,n),u}static async createPeerConnection(t){return new RTCPeerConnection({sdpSemantics:"unified-plan",iceServers:t})}static async waitFor(t,e,s){let r=0;if(await new Promise(a=>{const n=setInterval(()=>{r+=1,r>=s&&(clearInterval(n),a("bad")),t()&&(clearInterval(n),a("good"))},e)}),r>=s)throw new a}isReady(){return this.streams.length>0&&"open"===this.dc.readyState}#t(t){this.dispatchEvent(new CustomEvent("status",{detail:t}))}subscribeStatus(t){return this.addEventListener("status",t),()=>{this.removeEventListener("status",t)}}getStream(){return this.streams[0]}sendMessage(t,e){this.dc.send(JSON.stringify({type:t,data:e}))}ttstf(t){this.sendMessage("ttstf",{message:t})}static BACKPRESSURE_THRESHOLD=524288;static FILE_TRANSFER_TIMEOUT=3e4;sendFile(t,e=65536){return new Promise((s,a)=>{let r=!1;const n=t=>{r||(r=!0,clearTimeout(i),t())},o=this.pc.createDataChannel("file",{protocol:"file"}),i=setTimeout(()=>{n(()=>{o.close(),a(new Error("File transfer timed out"))})},_.FILE_TRANSFER_TIMEOUT);o.onclose=()=>{n(()=>{a(new Error("File channel closed before transfer completed"))})},o.onerror=t=>{n(()=>{o.close(),a(new Error(`File channel error: ${t}`))})},o.addEventListener("message",async r=>{try{if(0===r.data.length){const s=new Uint8Array(await t.arrayBuffer());let r=0;const i=()=>{for(;r<s.length;){if(o.bufferedAmount>_.BACKPRESSURE_THRESHOLD)return o.bufferedAmountLowThreshold=_.BACKPRESSURE_THRESHOLD/2,o.onbufferedamountlow=()=>{o.onbufferedamountlow=null,o.onclose=null,i()},void(o.onclose=()=>{o.onbufferedamountlow=null,n(()=>{a(new Error("File channel closed during transfer"))})});o.send(s.slice(r,r+e)),r+=e}o.send(new Uint8Array(0))};i()}else n(()=>{o.close(),s(r.data)})}catch(t){n(()=>{o.close(),a(t instanceof Error?t:new Error(String(t)))})}})})}async stf(t,e,s){const a=await this.sendFile(t);return this.sendMessage("stf",{message:s,file_ref:a,format:e}),a}recordStart(){this.sendMessage("record-start",{})}recordEndStt(t){this.sendMessage("record-end-stt",{language:t})}recordEndTranslate(t,e){this.sendMessage("record-end-translate",{src_lang:t,dst_lang:e})}changeSize(t,e){this.sendMessage("change-size",{width:t,height:e})}setTemplate(t,e){this.sendMessage("set-template",{model:t,dress:e})}clearBuffer(){this.sendMessage("clear-buffer",{})}ping(){this.sendMessage("ping",{})}setMessageCallback(t,e){const s=s=>{const a=JSON.parse(s.data);a.type===t&&e(a.data)};return this.dc.addEventListener("message",s),()=>{this.dc.removeEventListener("message",s)}}async tts(t,e=!0){return w(t,e)}close(){this.dc.close(),this.pc.close(),this.#t({live:!1,code:408,reason:"Request Timeout"})}closeSelf(){this.dc.close(),this.pc.close(),this.#t({live:!1,code:200,reason:"OK"})}}class x{apiServer;sessionId;perso;clientTools;chatStatesHandler=new EventTarget;chatLogHandler=new EventTarget;sttEventHandler=null;errorHandler=new EventTarget;lastStfTimeoutHandle=null;stfTotalDuration=0;stfTimeoutStartTime=0;messageHistory=[];chatLog=[];llmProcessor;chatStateMap=new Map([[exports.ChatState.RECORDING,0],[exports.ChatState.LLM,0],[exports.ChatState.ANALYZING,0],[exports.ChatState.SPEAKING,0],[exports.ChatState.TTS,0]]);sttRecorder=null;sttTimeoutHandle=null;sttTimeoutAudioFile=null;heartbeatIntervalId=null;legacyVoiceChatMode;stream;constructor(t,e,s,a,r){this.apiServer=t,this.sessionId=e,this.perso=s,this.clientTools=a,this.legacyVoiceChatMode=r?.legacyVoiceChatMode??!1,this.stream=r?.stream??null,this.resetChatState(),this.llmProcessor=new v({apiServer:t,sessionId:e,clientTools:a,callbacks:{onChatStateChange:(t,e)=>this.setChatState(t,e),onError:t=>this.setError(t),onChatLog:(t,e)=>this.addMessageToChatLog(t,e),onTTSTF:t=>this.processTTSTFInternal(t)}}),s?(s.subscribeStatus(t=>{!1===t.detail?.live&&this.stopHeartbeat()}),s.setMessageCallback("stf",t=>{if(this.chatStateMap.get(exports.ChatState.ANALYZING)||this.chatStateMap.get(exports.ChatState.SPEAKING))if(this.setChatState(exports.ChatState.SPEAKING,exports.ChatState.ANALYZING),null!==this.lastStfTimeoutHandle){clearTimeout(this.lastStfTimeoutHandle);let e=Date.now();this.stfTotalDuration+=t.duration+1e3-(e-this.stfTimeoutStartTime),this.stfTimeoutStartTime=e,this.lastStfTimeoutHandle=setTimeout(()=>{this.lastStfTimeoutHandle=null,this.stfTimeoutStartTime=0,this.stfTotalDuration=0,this.setChatState(null,exports.ChatState.SPEAKING)},this.stfTotalDuration)}else this.stfTimeoutStartTime=Date.now(),this.stfTotalDuration=t.duration+2e3,this.lastStfTimeoutHandle=setTimeout(()=>{this.lastStfTimeoutHandle=null,this.stfTimeoutStartTime=0,this.stfTotalDuration=0,this.setChatState(null,exports.ChatState.SPEAKING)},this.stfTotalDuration)}),s.setMessageCallback("stt",t=>{if(this.setChatState(null,exports.ChatState.ANALYZING),null!=this.sttEventHandler)this.sttEventHandler.dispatchEvent(new CustomEvent("stt",{detail:t.text}));else{if(""===t.text)return;this.processChat(t.text)}}),s.setMessageCallback("stt-error",t=>{this.setChatState(null,exports.ChatState.ANALYZING)})):this.startHeartbeat()}llmJob=null;async processChat(t){0!==t.trim().length&&(this.pipelineSuppressed=!1,this.addMessageToChatLog(t,!0),this.llmJob=this.processChatInternal(t))}processLLM(t){return this.pipelineSuppressed=!1,this.llmProcessor.processLLM(t)}getMessageHistory(){return this.llmProcessor.getHistory()}processCustomChat(t){0!==t.trim().length&&this.processTTSTFInternal(t)}processTTSTF(t){0!==t.trim().length&&(this.pipelineSuppressed=!1,this.messageHistory.push({role:"assistant",type:"message",content:t}),this.addMessageToChatLog(t,!1),this.processTTSTFInternal(t))}async transcribeAudio(t,e){const s=t instanceof File?t:new File([t],"audio.wav",{type:t.type});try{return(await h.makeSTT(this.apiServer,this.sessionId,s,e)).text}catch(t){if(t instanceof r)throw new i(t);throw t}}async processSTF(t,e,s){if(!this.perso)throw new Error("processSTF requires WebRTC (STF mode)");this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.ANALYZING);try{const a=await this.perso.stf(t,e,s);return this.pipelineSuppressed?(this.setChatState(null,exports.ChatState.ANALYZING),a):a}catch(t){throw this.setChatState(null,exports.ChatState.ANALYZING),t}}async processTTS(t,e={}){const{resample:s=!1,locale:a,output_format:n}=e,o=C(t).trim();if(0===o.length)return;this.pipelineSuppressed=!1;const i=/[.?!]$/.test(o)?o:o+".";this.setChatState(exports.ChatState.TTS,null);try{const t={sessionId:this.sessionId,text:i,...a&&{locale:a},...n&&{output_format:n}},{audio:e}=await h.makeTTS(this.apiServer,t);if(this.pipelineSuppressed)return;return await w(e,s)}catch(t){t instanceof r||t instanceof c?this.setError(new l(t)):this.setError(t instanceof Error?t:new Error(String(t)))}finally{this.setChatState(null,exports.ChatState.TTS)}}startVoiceChat(){if(!this.perso)throw new Error("startVoiceChat requires WebRTC (STF mode)");return this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.RECORDING),this.perso.recordStart()}stopVoiceChat(){if(!this.perso)throw new Error("stopVoiceChat requires WebRTC (STF mode)");this.setChatState(exports.ChatState.ANALYZING,exports.ChatState.RECORDING),this.perso.recordEndStt()}async startProcessSTT(t){if(this.sttRecorder?.isRecording())throw new Error("STT recording is already in progress");this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.RECORDING);try{this.sttRecorder=new R({targetSampleRate:16e3}),await this.sttRecorder.start(),t&&t>0&&(this.sttTimeoutHandle=setTimeout(async()=>{if(this.sttTimeoutHandle=null,this.sttRecorder?.isRecording()){try{this.sttTimeoutAudioFile=await this.sttRecorder.stop()}catch{this.sttTimeoutAudioFile=null,this.setChatState(null,exports.ChatState.RECORDING)}this.sttRecorder=null}},t))}catch(t){throw this.setChatState(null,exports.ChatState.RECORDING),this.sttRecorder=null,t}}lastRecordedAudioFile=null;async stopProcessSTT(t){let e;if(this.sttTimeoutHandle&&(clearTimeout(this.sttTimeoutHandle),this.sttTimeoutHandle=null),this.setChatState(null,exports.ChatState.RECORDING),this.sttTimeoutAudioFile)e=this.sttTimeoutAudioFile,this.sttTimeoutAudioFile=null;else{if(!this.sttRecorder?.isRecording())throw this.sttRecorder?(this.sttRecorder=null,new Error("STT recording is not in progress")):new Error("STT recording has not been started");e=await this.sttRecorder.stop(),this.sttRecorder=null}this.lastRecordedAudioFile=e;try{return(await h.makeSTT(this.apiServer,this.sessionId,e,t)).text}catch(t){if(t instanceof r)throw new i(t);throw t}}isSTTRecording(){return(this.sttRecorder?.isRecording()??!1)||null!==this.sttTimeoutAudioFile}changeSize(t,e){this.perso?.changeSize(t,e)}async clearBuffer(){this.perso?.clearBuffer(),await this.clearLLMJob(),null!==this.lastStfTimeoutHandle&&(clearTimeout(this.lastStfTimeoutHandle),this.lastStfTimeoutHandle=null),this.pipelineSuppressed=!0,this.resetChatState()}setSrc(t){t.srcObject=this.getRemoteStream()??null}getRemoteStream(){return this.perso?.getStream()}getLocalStream(){return this.stream}stopSession(){this.close()}onClose(t){return this.perso?this.perso.subscribeStatus(e=>{null!=e.detail&&!1===e.detail.live&&t(200===e.detail.code)}):()=>{}}subscribeChatStates(t){const e=e=>{t(e.detail.status)};return this.chatStatesHandler.addEventListener("status",e),()=>{this.chatStatesHandler.removeEventListener("status",e)}}subscribeChatLog(t){const e=e=>{t(e.detail.chatLog)};return this.chatLogHandler.addEventListener("chatLog",e),()=>{this.chatLogHandler.removeEventListener("chatLog",e)}}setSttResultCallback(t){const e=e=>{t(e.detail)};return this.sttEventHandler=new EventTarget,this.sttEventHandler.addEventListener("stt",e),()=>{this.sttEventHandler?.removeEventListener("stt",e),this.sttEventHandler=null}}setErrorHandler(t){const e=e=>{t(e.detail.error)};return this.errorHandler.addEventListener("error",e),()=>{this.errorHandler.removeEventListener("error",e)}}getSessionId(){return this.sessionId}async processChatInternal(t){this.setChatState(exports.ChatState.LLM);const e=this.clientTools.map(t=>({type:"function",function:{description:t.description,name:t.name,parameters:t.parameters}})),s=new Array;null===t||(t instanceof Array?s.push(...t):"string"==typeof t&&s.push({role:"user",content:t}));const a=await fetch(`${this.apiServer}/api/v1/session/${this.sessionId}/llm/v2/`,{body:JSON.stringify({messages:[...this.messageHistory,...s],tools:e}),headers:{"Content-Type":"application/json"},method:"POST"});if(!a.ok){const t=await a.json(),e=new n(new r(a.status,t.errors[0].code,t.errors[0].detail,t.errors[0].attr));return this.setError(e),void this.setChatState(null,exports.ChatState.LLM)}const i=a.body?.getReader(),l=new TextDecoder("utf-8");let c="",h=null,d="";for(;;){const{done:t,value:e}=await i.read();if(t)break;let a;for(d+=l.decode(e,{stream:!0});-1!==(a=d.indexOf("\n"));){if(this.llmCancel)return c.length>0&&this.addMessageToChatLog(c,!1),void this.setChatState(null,exports.ChatState.LLM);const t=d.slice(0,a).trim();if(d=d.slice(a+1),!t.startsWith("data: {")){const t=new n(new o("Failed to parse SSE response"));return this.setError(t),void this.setChatState(null,exports.ChatState.LLM)}const e=JSON.parse(t.slice(6).trim());if("success"!==e.status){const t=new n(new o(e.reason));return this.setError(t),void this.setChatState(null,exports.ChatState.LLM)}c.length>0&&"message"!=e.type&&(s.push({role:"assistant",type:"message",content:c}),this.addMessageToChatLog(c,!1),c=""),"message"!==e.type?"tool_call"!==e.type||null==e.tool_calls?"tool"!==e.role||"tool_call"===e.type&&s.push({role:e.role,type:e.type,content:e.content,tool_call_id:e.tool_call_id}):(s.push({role:"assistant",type:e.type,content:e.content,tool_calls:e.tool_calls}),h=e):(c+=C(e.content),this.processTTSTFInternal(e.content))}}if(this.llmCancel)this.setChatState(null,exports.ChatState.LLM);else{if(null!=h){const t=[];for(const e of h.tool_calls){const s=this.getChatTool(this.clientTools,e.function.name);null!=s&&t.push(new Promise(async t=>{try{const a=await s.call(JSON.parse(e.function.arguments));t({toolCallId:e.id,chatTool:s,chatToolResult:a})}catch(a){t({toolCallId:e.id,chatTool:s,chatToolResult:{result:"error!"}})}}))}const e=await Promise.all(t);for(const t of e)s.push({role:"tool",content:JSON.stringify(t.chatToolResult),tool_call_id:t.toolCallId});const a=e.length>0&&h.tool_calls.length!==e.length,r=e.some(t=>!t.chatTool.executeOnly);a||r?await this.processChatInternal(s):this.messageHistory.push(...s)}else this.messageHistory.push(...s);this.setChatState(null,exports.ChatState.LLM)}}getChatTool(t,e){for(const s of t)if(s.name===e)return s;return null}llmCancel=!1;pipelineSuppressed=!1;async clearLLMJob(){null!=this.llmJob&&(this.llmCancel=!0,await this.llmJob,this.llmCancel=!1)}processTTSTFInternal(t){const e=C(t).trim();0!==e.length&&this.perso&&(this.setChatState(exports.ChatState.ANALYZING),this.perso.ttstf(e))}addMessageToChatLog(t,e){this.chatLog=[{text:t,isUser:e,timestamp:new Date},...this.chatLog],this.chatLogHandler.dispatchEvent(new CustomEvent("chatLog",{detail:{chatLog:this.chatLog}}))}setChatState(t=null,e=null){const s=new Map(this.chatStateMap);function a(t){t===exports.ChatState.ANALYZING?s.set(t,(s.get(t)||0)+1):s.set(t,1)}function r(t){t===exports.ChatState.ANALYZING?s.set(t,Math.max((s.get(t)||0)-1,0)):s.set(t,0)}if(null!=t)if(t instanceof Array)for(let e of t)a(e);else a(t);if(null!=e)if(e instanceof Array)for(let t of e)r(t);else r(e);const n=this.exchangeChatStateMapToSet(this.chatStateMap),o=this.exchangeChatStateMapToSet(s);this.chatStateMap=s,this.isEqualChatStateMap(n,o)||this.dispatchChatState(o)}resetChatState(){this.chatStateMap=new Map([[exports.ChatState.RECORDING,0],[exports.ChatState.LLM,0],[exports.ChatState.ANALYZING,0],[exports.ChatState.SPEAKING,0],[exports.ChatState.TTS,0]]),this.dispatchChatState(this.exchangeChatStateMapToSet(this.chatStateMap))}exchangeChatStateMapToSet(t){const e=new Set;for(const s of t)s[1]>0&&e.add(s[0]);return e}dispatchChatState(t){this.chatStatesHandler.dispatchEvent(new CustomEvent("status",{detail:{status:t}}))}isEqualChatStateMap(t,e){if(t.size!==e.size)return!1;for(const s of t)if(t.has(s)!==e.has(s))return!1;return!0}async logSessionEvent(t){const s="object"==typeof t?JSON.stringify(t):t;await h.sessionEvent(this.apiServer,this.sessionId,e.SESSION_LOG,s)}setError(t){this.errorHandler.dispatchEvent(new CustomEvent("error",{detail:{error:t}}))}close(){this.stopHeartbeat(),this.perso?.closeSelf()}startHeartbeat(){const t=async()=>{try{await h.sessionEvent(this.apiServer,this.sessionId,e.SESSION_DURING),null!==this.heartbeatIntervalId&&(this.heartbeatIntervalId=setTimeout(t,1e4))}catch(t){t instanceof r?this.setError(t):this.setError(t instanceof Error?t:new Error(String(t))),this.close()}};this.heartbeatIntervalId=setTimeout(t,1e4)}stopHeartbeat(){null!==this.heartbeatIntervalId&&(clearTimeout(this.heartbeatIntervalId),this.heartbeatIntervalId=null)}}async function L(t,e,s,a,r,n){if("boolean"!=typeof r){const n=await _.create(t,e,s,a);return new x(t,e,n,r)}const o=n??[];let i,l;if(r)i=await navigator.mediaDevices.getUserMedia({audio:!0,video:!1}),l=()=>{};else{const t=new AudioContext,e=t.createOscillator();e.frequency.value=0;const s=t.createMediaStreamDestination();e.connect(s),e.start(),i=s.stream,l=()=>{e.stop(),e.disconnect(s),t.close()}}const c=await _.create(t,e,s,a,i);if(!c)return l(),new x(t,e,null,o);const h=new x(t,e,c,o,{stream:i,legacyVoiceChatMode:!0});return h.onClose(()=>{l()}),h}async function M(t,e){return await h.getLLMs(t,e)}async function b(t,e){return await h.getTTSs(t,e)}async function N(t,e){return await h.getSTTs(t,e)}async function A(t,e){return await h.getModelStyles(t,e)}async function k(t,e){return await h.getBackgroundImages(t,e)}async function O(t,e){return await h.getPrompts(t,e)}async function P(t,e){return await h.getDocuments(t,e)}async function H(t,e){return await h.getTextNormalizations(t,e)}async function F(t,e){return await h.getMcpServers(t,e)}exports.ApiError=r,exports.ChatTool=class{name;description;parameters;call;executeOnly;constructor(t,e,s,a,r=!1){this.name=t,this.description=e,this.parameters=s,this.call=a,this.executeOnly=r}},exports.LLMError=n,exports.LLMStreamingResponseError=o,exports.LlmProcessor=v,exports.STTError=i,exports.Session=x,exports.TTSDecodeError=c,exports.TTSError=l,exports.TTS_TARGET_SAMPLE_RATE=y,exports.WavRecorder=R,exports.createSession=async function(t,e,s,a,r,n){return"boolean"==typeof r?await L(t,e,s,a,r,n??[]):await L(t,e,s,a,r)},exports.createSessionId=async function(e,s,a){let r;if("undefined"!=typeof window&&console.warn("[perso-interactive-sdk-web] WARNING: createSessionId is being called from the browser. This exposes your API key and is not recommended for production. Use server-side session creation with 'perso-interactive-sdk-web/server' instead. See: https://github.com/perso-ai/perso-interactive-sdk-web#server-side"),"string"==typeof a){const n=await h.getSessionTemplate(e,s,a);if("webrtc"!==n.model_style.platform_type)throw new Error(`SessionTemplate "${a}" uses platform_type "${n.model_style.platform_type}", but only "webrtc" is supported`);r=function(e){const s=t=>e.capability.some(e=>e.name===t);return{using_stf_webrtc:s(t.STF_WEBRTC),model_style:e.model_style.name,prompt:e.prompt.prompt_id,document:e.document?.document_id,background_image:e.background_image?.backgroundimage_id,mcp_servers:e.mcp_servers?.length?e.mcp_servers.map(t=>t.mcpserver_id):[],llm_type:s(t.LLM)?e.llm_type.name:void 0,tts_type:s(t.TTS)?e.tts_type.name:void 0,stt_type:s(t.STT)?e.stt_type.name:void 0,text_normalization_config:e.text_normalization_config?.textnormalizationconfig_id,text_normalization_locale:e.text_normalization_locale,padding_left:e.padding_left??void 0,padding_top:e.padding_top??void 0,padding_height:e.padding_height??void 0}}(n)}else r=a;const n={capability:[],...r};r.using_stf_webrtc&&n.capability.push(t.STF_WEBRTC),r?.llm_type&&(n.capability.push(t.LLM),n.llm_type=r.llm_type),r?.tts_type&&(n.capability.push(t.TTS),n.tts_type=r.tts_type),r?.stt_type&&(n.capability.push(t.STT),n.stt_type=r.stt_type);const o=await fetch(`${e}/api/v1/session/`,{body:JSON.stringify(n),headers:{"PersoLive-APIKey":s,"Content-Type":"application/json"},method:"POST"});return(await h.parseJson(o)).session_id},exports.createWavRecorder=function(t){return new R(t)},exports.getAllSettings=async function(t,e){const[s,a,r,n,o,i,l,c,h]=await Promise.all([M(t,e),b(t,e),N(t,e),A(t,e),k(t,e),O(t,e),P(t,e),F(t,e),H(t,e).catch(()=>[])]);return{llms:s,ttsTypes:a,sttTypes:r,modelStyles:n,backgroundImages:o,prompts:i,documents:l,mcpServers:c,textNormalizations:h}},exports.getBackgroundImages=k,exports.getDocuments=P,exports.getLLMs=M,exports.getMcpServers=F,exports.getModelStyles=A,exports.getPrompts=O,exports.getSTTs=N,exports.getSessionInfo=async function(t,e){return await h.getSessionInfo(t,e)},exports.getSessionTemplates=async function(t,e){return await h.getSessionTemplates(t,e)},exports.getTTSs=b,exports.getTextNormalization=async function(t,e,s){return await h.downloadTextNormalization(t,e,s)},exports.getTextNormalizations=H,exports.getWavSampleRate=function(t){const e=new DataView(t);if(t.byteLength<28)throw new d("File too small to be a valid WAV");if("RIFF"!==p(e,0,4))throw new d("Missing RIFF header");let s=12;for(;s<t.byteLength-8;){const a=p(e,s,4),r=e.getUint32(s+4,!0);if("fmt "===a){if(s+16>t.byteLength)throw new d("fmt chunk extends beyond file");return e.getUint32(s+12,!0)}const n=s+8+r;if(n<=s)break;s=n}throw new d("Missing fmt chunk")},exports.makeTTS=async function(t,e){return await h.makeTTS(t,e)};
+"use strict";var t,e,s,a=require("emoji-regex");exports.ChatState=void 0,(t=exports.ChatState||(exports.ChatState={})).RECORDING="RECORDING",t.LLM="LLM",t.ANALYZING="ANALYZING",t.SPEAKING="SPEAKING",t.TTS="TTS";class r extends Error{constructor(){super("WebRTC connection timeout")}}class n extends Error{errorCode;code;detail;attr;constructor(t,e,s,a){let r;r=null!=a?`${t}:${a}_${s}`:`${t}:${s}`,super(r),this.errorCode=t,this.code=e,this.detail=s,this.attr=a}}class o extends Error{underlyingError;constructor(t){super(),this.underlyingError=t}}class i extends Error{description;constructor(t){super(),this.description=t}}class l extends Error{underlyingError;constructor(t){super(`STT Error: ${t.detail}`),this.underlyingError=t}}class c extends Error{underlyingError;constructor(t){super(t.message),this.underlyingError=t}}class h extends Error{description;constructor(t){super(`TTS decode error: ${t}`),this.description=t}}!function(t){t.LLM="LLM",t.TTS="TTS",t.STT="STT",t.STF_ONPREMISE="STF_ONPREMISE",t.STF_WEBRTC="STF_WEBRTC"}(e||(e={})),function(t){t.SESSION_START="SESSION_START",t.SESSION_DURING="SESSION_DURING",t.SESSION_LOG="SESSION_LOG",t.SESSION_END="SESSION_END",t.SESSION_ERROR="SESSION_ERROR",t.SESSION_TTS="SESSION_TTS",t.SESSION_STT="SESSION_STT",t.SESSION_LLM="SESSION_LLM"}(s||(s={}));class d{static async getLLMs(t,e){const s=fetch(`${t}/api/v1/settings/llm_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getModelStyles(t,e){const s=fetch(`${t}/api/v1/settings/modelstyle/?platform_type=webrtc`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getBackgroundImages(t,e){const s=fetch(`${t}/api/v1/background_image/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getTTSs(t,e){const s=fetch(`${t}/api/v1/settings/tts_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getSTTs(t,e){const s=fetch(`${t}/api/v1/settings/stt_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async makeTTS(t,{sessionId:e,text:s,locale:a,output_format:r}){const n={text:s};a&&(n.locale=a),r&&(n.output_format=r);const o=await fetch(`${t}/api/v1/session/${e}/tts/`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify(n)});return await this.parseJson(o)}static async getPrompts(t,e){const s=fetch(`${t}/api/v1/prompt/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getDocuments(t,e){const s=fetch(`${t}/api/v1/document/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getTextNormalizations(t,e){const s=fetch(`${t}/api/v1/settings/text_normalization_config/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async downloadTextNormalization(t,e,s){const a=await fetch(`${t}/api/v1/settings/text_normalization_config/${s}/download/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(a)}static async getSessionTemplates(t,e){const s=await fetch(`${t}/api/v1/session_template/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(s)}static async getSessionTemplate(t,e,s){const a=await fetch(`${t}/api/v1/session_template/${s}/`,{headers:{"PersoLive-APIKey":e},method:"GET"});return await this.parseJson(a)}static async getMcpServers(t,e){const s=fetch(`${t}/api/v1/settings/mcp_type/`,{headers:{"PersoLive-APIKey":e},method:"GET"}),a=await s;return await this.parseJson(a)}static async getSessionInfo(t,e){const s=fetch(`${t}/api/v1/session/${e}/`,{method:"GET"}),a=await s;return await this.parseJson(a)}static async sessionEvent(t,e,s,a=""){const r=await fetch(`${t}/api/v1/session/${e}/event/create/`,{method:"POST",headers:{"Content-Type":"application/json"},body:JSON.stringify({detail:a,event:s})});await this.parseJson(r)}static async makeSTT(t,e,s,a){const r=new FormData;r.append("audio",s),a&&r.append("language",a);const n=await fetch(`${t}/api/v1/session/${e}/stt/`,{method:"POST",body:r});return await this.parseJson(n)}static async makeLLM(t,e,s,a){const r=await fetch(`${t}/api/v1/session/${e}/llm/v2/`,{body:JSON.stringify(s),headers:{"Content-Type":"application/json"},method:"POST",signal:a});if(!r.ok){const t=await r.json(),e=t.errors?.[0]??{code:"UNKNOWN_ERROR",detail:`Server returned status ${r.status} with no error details`,attr:null};throw new n(r.status,e.code,e.detail,e.attr)}return r.body.getReader()}static async getIceServers(t,e){const s=await fetch(`${t}/api/v1/session/${e}/ice-servers/`,{method:"GET"});return(await this.parseJson(s)).ice_servers}static async exchangeSDP(t,e,s){const a=await fetch(`${t}/api/v1/session/${e}/exchange/`,{body:JSON.stringify({client_sdp:s}),headers:{"Content-Type":"application/json"},method:"POST"});return(await this.parseJson(a)).server_sdp}static async parseJson(t){const e=await t.json();if(t.ok)return e;{const s=e.errors?.[0]??{code:"UNKNOWN_ERROR",detail:`Server returned status ${t.status} with no error details`,attr:null};throw new n(t.status,s.code,s.detail,s.attr)}}}async function u(t,e){return await d.getLLMs(t,e)}async function p(t,e){return await d.getTTSs(t,e)}async function g(t,e){return await d.getSTTs(t,e)}async function m(t,e){return await d.getModelStyles(t,e)}async function S(t,e){return await d.getBackgroundImages(t,e)}async function f(t,e){return await d.getPrompts(t,e)}async function y(t,e){return await d.getDocuments(t,e)}async function w(t,e){return await d.getMcpServers(t,e)}async function T(t,e){return await d.getTextNormalizations(t,e)}class C extends Error{constructor(t){super(`WAV parse error: ${t}`),this.name="WavParseError"}}function E(t,e,s=1){const a=2*t.length,r=new ArrayBuffer(44+a),n=new DataView(r);_(n,0,"RIFF"),n.setUint32(4,36+a,!0),_(n,8,"WAVE"),_(n,12,"fmt "),n.setUint32(16,16,!0),n.setUint16(20,1,!0),n.setUint16(22,s,!0),n.setUint32(24,e,!0),n.setUint32(28,e*s*2,!0),n.setUint16(32,2*s,!0),n.setUint16(34,16,!0),_(n,36,"data"),n.setUint32(40,a,!0);let o=44;for(let e=0;e<t.length;e++){const s=Math.max(-1,Math.min(1,t[e]));n.setInt16(o,s<0?32768*s:32767*s,!0),o+=2}return r}function v(t,e,s){let a="";for(let r=0;r<s;r++)a+=String.fromCharCode(t.getUint8(e+r));return a}function _(t,e,s){for(let a=0;a<s.length;a++)t.setUint8(e+a,s.charCodeAt(a))}function I(t,e,s){switch(s){case 8:return(t.getUint8(e)-128)/128;case 16:return t.getInt16(e,!0)/32768;case 24:{const s=t.getUint8(e),a=t.getUint8(e+1),r=t.getUint8(e+2)<<16|a<<8|s;return(r>8388607?r-16777216:r)/8388608}case 32:return t.getInt32(e,!0)/2147483648;default:return 0}}class R extends Error{constructor(t){super(`Audio resample error: ${t}`),this.name="AudioResampleError"}}async function x(t,e,s,a=1){if(0===t.length)throw new R("Cannot resample empty audio data");if(e<=0||s<=0)throw new R(`Invalid sample rate: original=${e}, target=${s}`);if(e===s)return t;try{const r=t.length/e,n=Math.ceil(r*s),o=new OfflineAudioContext(a,t.length,e).createBuffer(a,t.length,e);o.getChannelData(0).set(t);const i=new OfflineAudioContext(a,n,s),l=i.createBufferSource();l.buffer=o,l.connect(i.destination),l.start(0);return(await i.startRendering()).getChannelData(0)}catch(t){const a=t instanceof Error?t.message:String(t);throw new R(`Failed to resample audio from ${e}Hz to ${s}Hz: ${a}`)}}const L=16e3;async function M(t,e=!0){let s;try{const e=atob(t),a=new Array(e.length);for(let t=0;t<e.length;t++)a[t]=e.charCodeAt(t);s=new Uint8Array(a).buffer}catch{throw new h("Invalid Base64 audio data")}const a=function(t){const e=new Uint8Array(t);if(e.length>=4){const t=String.fromCharCode(e[0],e[1],e[2],e[3]);if("RIFF"===t)return"audio/wav";if(t.startsWith("ID3"))return"audio/mpeg";if(255===e[0]&&!(224&~e[1]))return"audio/mpeg"}return"audio/wav"}(s);if(!e)return new Blob([s],{type:a});try{const t=await async function(t,e){if("audio/wav"===e){const e=function(t){const e=new DataView(t);if(t.byteLength<44)throw new C("File too small to be a valid WAV");if("RIFF"!==v(e,0,4))throw new C("Missing RIFF header");if("WAVE"!==v(e,8,4))throw new C("Missing WAVE format identifier");let s=12,a=!1,r=0,n=0,o=0,i=0;for(;s<t.byteLength-8;){const l=v(e,s,4),c=e.getUint32(s+4,!0);if("fmt "===l){if(s+24>t.byteLength)throw new C("fmt chunk extends beyond file");r=e.getUint16(s+8,!0),n=e.getUint16(s+10,!0),o=e.getUint32(s+12,!0),i=e.getUint16(s+22,!0),a=!0,s+=8+c;break}const h=s+8+c;if(h<=s)break;s=h}if(!a)throw new C("Missing fmt chunk");if(1!==r)throw new C(`Unsupported audio format: ${r} (only PCM format 1 is supported)`);if(0===n||n>8)throw new C(`Invalid channel count: ${n}`);if(0===o||o>192e3)throw new C(`Invalid sample rate: ${o}`);if(![8,16,24,32].includes(i))throw new C(`Unsupported bits per sample: ${i}`);let l=-1,c=0;for(;s<t.byteLength-8;){const t=v(e,s,4),a=e.getUint32(s+4,!0);if("data"===t){l=s+8,c=a;break}const r=s+8+a;if(r<=s)break;s=r}if(-1===l)throw new C("Missing data chunk");const h=t.byteLength-l,d=Math.min(c,h),u=i/8,p=Math.floor(d/(u*n)),g=new Float32Array(p*n);let m=0;for(let s=0;s<p*n;s++){const a=l+s*u;if(a+u>t.byteLength)break;g[m++]=Math.max(-1,Math.min(1,I(e,a,i)))}if(2===n){const t=new Float32Array(p);for(let e=0;e<p;e++)t[e]=(g[2*e]+g[2*e+1])/2;return{sampleRate:o,channels:1,bitsPerSample:i,samples:t}}return{sampleRate:o,channels:n,bitsPerSample:i,samples:g.slice(0,m)}}(t);if(e.sampleRate===L)return{samples:e.samples,sampleRate:e.sampleRate};return{samples:await x(e.samples,e.sampleRate,L,e.channels),sampleRate:L}}const s=new AudioContext({sampleRate:L});try{const e=await s.decodeAudioData(t.slice(0));if(e.sampleRate===L)return{samples:e.getChannelData(0),sampleRate:L};return{samples:await x(e.getChannelData(0),e.sampleRate,L,1),sampleRate:L}}finally{await s.close()}}(s,a),e=E(t.samples,L,1);return new Blob([e],{type:"audio/wav"})}catch{return new Blob([s],{type:a})}}const b=a();function N(t){return t.replace(b,"")}class A{config;messageHistory=[];constructor(t){this.config=t}async*processLLM(t){if(0===t.message.length)throw new Error("Message cannot be empty");const e=t.tools??this.config.clientTools,s=e.map(t=>({type:"function",function:{description:t.description,name:t.name,parameters:t.parameters}})),a={newMessageHistory:[{role:"user",content:t.message}],allChunks:[],message:"",lastYieldedChunkCount:0,pendingToolCallsMessage:null,aborted:!1,streamingError:null};let r=0,l=[...this.messageHistory,...a.newMessageHistory];this.config.callbacks.onChatStateChange(exports.ChatState.LLM,null);try{for(;;){if(t.signal?.aborted)return void(a.allChunks.length>0&&(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0}));let c;try{c=await d.makeLLM(this.config.apiServer,this.config.sessionId,{messages:l,tools:s},t.signal)}catch(t){if(t instanceof n)return void(yield{type:"error",error:new o(t)});throw t}if(a.streamingError=null,yield*this.parseSSEStream(c,a,t),a.streamingError)return;if(a.aborted)return void(a.allChunks.length>0&&(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0}));if(null!=a.pendingToolCallsMessage){yield*this.executeToolCalls(a,e);const t=a.lastToolCallResults,s=t.length>0&&a.pendingToolCallsMessage.tool_calls.length!==t.length,n=t.some(t=>!t.chatTool.executeOnly);if(s||n){if(r++,r>=10)return void(yield{type:"error",error:new o(new i("Tool follow-up loop exceeded maximum rounds (10)"))});l=[...this.messageHistory,...a.newMessageHistory],a.pendingToolCallsMessage=null;continue}}return this.messageHistory.push(...a.newMessageHistory),void(yield{type:"message",chunks:[...a.allChunks],message:a.message,finish:!0})}}finally{this.config.callbacks.onChatStateChange(null,exports.ChatState.LLM)}}async*parseSSEStream(t,e,s){const a=new TextDecoder("utf-8");let r="",n="";e.pendingToolCallsMessage=null;const l=()=>e.allChunks.length>e.lastYieldedChunkCount?(e.lastYieldedChunkCount=e.allChunks.length,{type:"message",chunks:[...e.allChunks],message:e.message,finish:!1}):null;for(;;){const{done:c,value:h}=await t.read();if(c)break;let d;for(r+=a.decode(h,{stream:!0});-1!==(d=r.indexOf("\n"));){if(s.signal?.aborted)return void(e.aborted=!0);const t=r.slice(0,d).trim();if(r=r.slice(d+1),!t.startsWith("data: {"))return e.streamingError=new o(new i("Failed to parse SSE response")),void(yield{type:"error",error:e.streamingError});let a;try{a=JSON.parse(t.slice(6).trim())}catch{return e.streamingError=new o(new i("Failed to parse SSE JSON")),void(yield{type:"error",error:e.streamingError})}if("success"!==a.status)return e.streamingError=new o(new i(a.reason)),void(yield{type:"error",error:e.streamingError});if(n.length>0&&"message"!=a.type){e.newMessageHistory.push({role:"assistant",type:"message",content:n}),n="";const t=l();t&&(yield t)}if("message"===a.type){const t=N(a.content);n+=t,e.message+=t,e.allChunks.push(t);continue}"tool_call"!==a.type||null==a.tool_calls?"tool"!==a.role||"tool_call"===a.type&&e.newMessageHistory.push({role:a.role,type:a.type,content:a.content,tool_call_id:a.tool_call_id}):(e.newMessageHistory.push({role:"assistant",type:a.type,content:a.content,tool_calls:a.tool_calls}),e.pendingToolCallsMessage=a,yield{type:"tool_call",tool_calls:a.tool_calls})}const u=l();u&&(yield u)}const c=r.trim();if(c.length>0){if(!c.startsWith("data: {"))return e.streamingError=new o(new i("Failed to parse SSE response")),void(yield{type:"error",error:e.streamingError});let t;try{t=JSON.parse(c.slice(6).trim())}catch{return e.streamingError=new o(new i("Failed to parse SSE JSON")),void(yield{type:"error",error:e.streamingError})}if("success"!==t.status)return e.streamingError=new o(new i(t.reason)),void(yield{type:"error",error:e.streamingError});if("message"===t.type){const s=N(t.content);n+=s,e.message+=s,e.allChunks.push(s)}else if("tool_call"===t.type&&null!=t.tool_calls){if(n.length>0){e.newMessageHistory.push({role:"assistant",type:"message",content:n}),n="";const t=l();t&&(yield t)}e.newMessageHistory.push({role:"assistant",type:t.type,content:t.content,tool_calls:t.tool_calls}),e.pendingToolCallsMessage=t,yield{type:"tool_call",tool_calls:t.tool_calls}}}n.length>0&&e.newMessageHistory.push({role:"assistant",type:"message",content:n})}async*executeToolCalls(t,e){const s=t=>{for(const s of e)if(s.name===t)return s;return null},a=[];for(const e of t.pendingToolCallsMessage.tool_calls){const t=s(e.function.name);null!=t&&a.push((async()=>{try{const s=await t.call(JSON.parse(e.function.arguments));return{toolCallId:e.id,chatTool:t,chatToolResult:s}}catch(s){return{toolCallId:e.id,chatTool:t,chatToolResult:{error:s.message}}}})())}const r=await Promise.all(a);t.lastToolCallResults=r;for(const e of r)t.newMessageHistory.push({role:"tool",content:JSON.stringify(e.chatToolResult),tool_call_id:e.toolCallId}),yield{type:"tool_result",tool_call_id:e.toolCallId,result:e.chatToolResult}}addToHistory(t){this.messageHistory.push(t)}getHistory(){return this.messageHistory}}const k=`data:application/javascript,${encodeURIComponent("\nclass RecorderProcessor extends AudioWorkletProcessor {\n  constructor() {\n    super();\n    this.isRecording = true;\n    \n    // Listen for stop message from main thread\n    this.port.onmessage = (event) => {\n      if (event.data.type === 'stop') {\n        this.isRecording = false;\n        // Send confirmation back to main thread\n        this.port.postMessage({ type: 'stopped' });\n      }\n    };\n  }\n\n  process(inputs, outputs, parameters) {\n    const input = inputs[0];\n    if (input && input.length > 0 && this.isRecording) {\n      // Clone the audio data and send to main thread\n      const channelData = new Float32Array(input[0]);\n      this.port.postMessage({ type: 'audio', data: channelData });\n    }\n    // Return true to keep the processor alive until stopped\n    return this.isRecording;\n  }\n}\n\nregisterProcessor('recorder-processor', RecorderProcessor);\n")}`;class O{audioContext=null;mediaStream=null;workletNode=null;sourceNode=null;audioChunks=[];isRecordingState=!1;channels;targetSampleRate;constructor(t={}){this.channels=t.channels||1,this.targetSampleRate=t.targetSampleRate}async start(){if(this.isRecordingState)throw new Error("WavRecorder is already recording");if(this.mediaStream=await navigator.mediaDevices.getUserMedia({audio:!0}),this.audioContext=new AudioContext,"running"!==this.audioContext.state)try{await this.audioContext.resume()}catch(t){console.warn("WavRecorder: Failed to resume AudioContext:",t)}await this.audioContext.audioWorklet.addModule(k),this.sourceNode=this.audioContext.createMediaStreamSource(this.mediaStream),this.workletNode=new AudioWorkletNode(this.audioContext,"recorder-processor"),this.audioChunks=[],this.workletNode.port.onmessage=t=>{"audio"===t.data.type&&this.audioChunks.push(t.data.data)},this.sourceNode.connect(this.workletNode),this.workletNode.connect(this.audioContext.destination),this.isRecordingState=!0}async stop(){if(!this.isRecordingState)throw new Error("WavRecorder is not recording");this.isRecordingState=!1,await new Promise(t=>{this.workletNode.port.onmessage=e=>{"stopped"===e.data.type?t():"audio"===e.data.type&&this.audioChunks.push(e.data.data)},this.workletNode.port.postMessage({type:"stop"})}),this.workletNode?.disconnect(),this.sourceNode?.disconnect(),this.mediaStream?.getTracks().forEach(t=>t.stop());const t=this.audioChunks,e=this.audioContext.sampleRate;try{const s=t.reduce((t,e)=>t+e.length,0),a=new Float32Array(s);let r,n,o=0;for(const e of t)a.set(e,o),o+=e.length;this.targetSampleRate&&this.targetSampleRate!==e?(r=await x(a,e,this.targetSampleRate,this.channels),n=this.targetSampleRate):(r=a,n=e);const i=E(r,n,this.channels),l=new Blob([i],{type:"audio/wav"});return new File([l],"recording.wav",{type:"audio/wav"})}finally{await(this.audioContext?.close()),this.audioContext=null,this.mediaStream=null,this.workletNode=null,this.sourceNode=null,this.audioChunks=[]}}isRecording(){return this.isRecordingState}}class P extends EventTarget{pc;dc;streams=[];pingTime;pingIntervalId=null;constructor(t,e){super(),this.pc=t,this.dc=e,this.pingTime=Date.now()+3e3,this.pc.addEventListener("track",t=>{this.streams=this.streams.concat(t.streams)}),this.pc.addEventListener("connectionstatechange",()=>{"disconnected"!==this.pc.connectionState&&"failed"!==this.pc.connectionState||this.close()}),this.dc.onopen=()=>{this.pingIntervalId=setInterval(()=>{this.ping(),Date.now()-this.pingTime>3e4&&this.close()},1e3)},this.dc.onclose=()=>{null!=this.pingIntervalId&&clearInterval(this.pingIntervalId)},this.#t({live:!0,code:200,reason:"OK"}),this.setMessageCallback("ping",()=>{this.pingTime=Date.now()})}static async create(t,a,r,n,o){const i=await d.getSessionInfo(t,a);if(!(Array.isArray(i.capability)&&i.capability.some(t=>t.name===e.STF_ONPREMISE||t.name===e.STF_WEBRTC)))return await d.sessionEvent(t,a,s.SESSION_START),null;const l=await d.getIceServers(t,a);let c=await P.createPeerConnection(l),h=c.createDataChannel("message",{protocol:"message"}),u=new P(c,h);o?o.getTracks().forEach(function(t){c.addTrack(t,o)}):c.addTransceiver("audio",{direction:"recvonly"});const p=c.addTransceiver("video",{direction:"recvonly"}),g=RTCRtpReceiver.getCapabilities("video");null!=g&&p.setCodecPreferences(g.codecs);const m=await c.createOffer();await c.setLocalDescription(m);const S=await d.exchangeSDP(t,a,m);return await c.setRemoteDescription(S),await P.waitFor(()=>u.isReady(),100,50),u.changeSize(r,n),u}static async createPeerConnection(t){return new RTCPeerConnection({sdpSemantics:"unified-plan",iceServers:t})}static async waitFor(t,e,s){let a=0;if(await new Promise(r=>{const n=setInterval(()=>{a+=1,a>=s&&(clearInterval(n),r("bad")),t()&&(clearInterval(n),r("good"))},e)}),a>=s)throw new r}isReady(){return this.streams.length>0&&"open"===this.dc.readyState}#t(t){this.dispatchEvent(new CustomEvent("status",{detail:t}))}subscribeStatus(t){return this.addEventListener("status",t),()=>{this.removeEventListener("status",t)}}getStream(){return this.streams[0]}sendMessage(t,e){this.dc.send(JSON.stringify({type:t,data:e}))}ttstf(t){this.sendMessage("ttstf",{message:t})}static BACKPRESSURE_THRESHOLD=524288;static FILE_TRANSFER_TIMEOUT=3e4;sendFile(t,e=65536){return new Promise((s,a)=>{let r=!1;const n=t=>{r||(r=!0,clearTimeout(i),t())},o=this.pc.createDataChannel("file",{protocol:"file"}),i=setTimeout(()=>{n(()=>{o.close(),a(new Error("File transfer timed out"))})},P.FILE_TRANSFER_TIMEOUT);o.onclose=()=>{n(()=>{a(new Error("File channel closed before transfer completed"))})},o.onerror=t=>{n(()=>{o.close(),a(new Error(`File channel error: ${t}`))})},o.addEventListener("message",async r=>{try{if(0===r.data.length){const s=new Uint8Array(await t.arrayBuffer());let r=0;const i=()=>{for(;r<s.length;){if(o.bufferedAmount>P.BACKPRESSURE_THRESHOLD)return o.bufferedAmountLowThreshold=P.BACKPRESSURE_THRESHOLD/2,o.onbufferedamountlow=()=>{o.onbufferedamountlow=null,o.onclose=null,i()},void(o.onclose=()=>{o.onbufferedamountlow=null,n(()=>{a(new Error("File channel closed during transfer"))})});o.send(s.slice(r,r+e)),r+=e}o.send(new Uint8Array(0))};i()}else n(()=>{o.close(),s(r.data)})}catch(t){n(()=>{o.close(),a(t instanceof Error?t:new Error(String(t)))})}})})}async stf(t,e,s){const a=await this.sendFile(t);return this.sendMessage("stf",{message:s,file_ref:a,format:e}),a}recordStart(){this.sendMessage("record-start",{})}recordEndStt(t){this.sendMessage("record-end-stt",{language:t})}recordEndTranslate(t,e){this.sendMessage("record-end-translate",{src_lang:t,dst_lang:e})}changeSize(t,e){this.sendMessage("change-size",{width:t,height:e})}setTemplate(t,e){this.sendMessage("set-template",{model:t,dress:e})}clearBuffer(){this.sendMessage("clear-buffer",{})}ping(){this.sendMessage("ping",{})}setMessageCallback(t,e){const s=s=>{const a=JSON.parse(s.data);a.type===t&&e(a.data)};return this.dc.addEventListener("message",s),()=>{this.dc.removeEventListener("message",s)}}async tts(t,e=!0){return M(t,e)}close(){this.dc.close(),this.pc.close(),this.#t({live:!1,code:408,reason:"Request Timeout"})}closeSelf(){this.dc.close(),this.pc.close(),this.#t({live:!1,code:200,reason:"OK"})}}class H{apiServer;sessionId;perso;clientTools;chatStatesHandler=new EventTarget;chatLogHandler=new EventTarget;sttEventHandler=null;errorHandler=new EventTarget;lastStfTimeoutHandle=null;stfTotalDuration=0;stfTimeoutStartTime=0;messageHistory=[];chatLog=[];llmProcessor;chatStateMap=new Map([[exports.ChatState.RECORDING,0],[exports.ChatState.LLM,0],[exports.ChatState.ANALYZING,0],[exports.ChatState.SPEAKING,0],[exports.ChatState.TTS,0]]);sttRecorder=null;sttTimeoutHandle=null;sttTimeoutAudioFile=null;heartbeatIntervalId=null;legacyVoiceChatMode;stream;constructor(t,e,s,a,r){this.apiServer=t,this.sessionId=e,this.perso=s,this.clientTools=a,this.legacyVoiceChatMode=r?.legacyVoiceChatMode??!1,this.stream=r?.stream??null,this.resetChatState(),this.llmProcessor=new A({apiServer:t,sessionId:e,clientTools:a,callbacks:{onChatStateChange:(t,e)=>this.setChatState(t,e),onError:t=>this.setError(t),onChatLog:(t,e)=>this.addMessageToChatLog(t,e),onTTSTF:t=>this.processTTSTFInternal(t)}}),s?(s.subscribeStatus(t=>{!1===t.detail?.live&&this.stopHeartbeat()}),s.setMessageCallback("stf",t=>{if(this.chatStateMap.get(exports.ChatState.ANALYZING)||this.chatStateMap.get(exports.ChatState.SPEAKING))if(this.setChatState(exports.ChatState.SPEAKING,exports.ChatState.ANALYZING),null!==this.lastStfTimeoutHandle){clearTimeout(this.lastStfTimeoutHandle);let e=Date.now();this.stfTotalDuration+=t.duration+1e3-(e-this.stfTimeoutStartTime),this.stfTimeoutStartTime=e,this.lastStfTimeoutHandle=setTimeout(()=>{this.lastStfTimeoutHandle=null,this.stfTimeoutStartTime=0,this.stfTotalDuration=0,this.setChatState(null,exports.ChatState.SPEAKING)},this.stfTotalDuration)}else this.stfTimeoutStartTime=Date.now(),this.stfTotalDuration=t.duration+2e3,this.lastStfTimeoutHandle=setTimeout(()=>{this.lastStfTimeoutHandle=null,this.stfTimeoutStartTime=0,this.stfTotalDuration=0,this.setChatState(null,exports.ChatState.SPEAKING)},this.stfTotalDuration)}),s.setMessageCallback("stt",t=>{if(this.setChatState(null,exports.ChatState.ANALYZING),null!=this.sttEventHandler)this.sttEventHandler.dispatchEvent(new CustomEvent("stt",{detail:t.text}));else{if(""===t.text)return;this.processChat(t.text)}}),s.setMessageCallback("stt-error",t=>{this.setChatState(null,exports.ChatState.ANALYZING)})):this.startHeartbeat()}llmJob=null;async processChat(t){0!==t.trim().length&&(this.pipelineSuppressed=!1,this.addMessageToChatLog(t,!0),this.llmJob=this.processChatInternal(t))}processLLM(t){return this.pipelineSuppressed=!1,this.llmProcessor.processLLM(t)}getMessageHistory(){return this.llmProcessor.getHistory()}processCustomChat(t){0!==t.trim().length&&this.processTTSTFInternal(t)}processTTSTF(t){0!==t.trim().length&&(this.pipelineSuppressed=!1,this.messageHistory.push({role:"assistant",type:"message",content:t}),this.addMessageToChatLog(t,!1),this.processTTSTFInternal(t))}async transcribeAudio(t,e){const s=t instanceof File?t:new File([t],"audio.wav",{type:t.type});try{return(await d.makeSTT(this.apiServer,this.sessionId,s,e)).text}catch(t){if(t instanceof n)throw new l(t);throw t}}async processSTF(t,e,s=""){if(!this.perso)throw new Error("processSTF requires WebRTC (STF mode)");this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.ANALYZING);try{const a=function(t,e){const s=t=>{if(!t)return null;const e=t.toLowerCase().split(";")[0].trim();return"mp3"===e||"audio/mpeg"===e||"audio/mp3"===e?"mp3":"wav"===e||"audio/wav"===e||"audio/x-wav"===e||"audio/wave"===e?"wav":null};return s(t)??s(e.type)??"wav"}(e,t),r=await this.perso.stf(t,a,s);return this.pipelineSuppressed?(this.setChatState(null,exports.ChatState.ANALYZING),r):r}catch(t){throw this.setChatState(null,exports.ChatState.ANALYZING),t}}async processTTS(t,e={}){const{resample:s=!1,locale:a,output_format:r}=e,o=N(t).trim();if(0===o.length)return;this.pipelineSuppressed=!1;const i=/[.?!]$/.test(o)?o:o+".";this.setChatState(exports.ChatState.TTS,null);try{const t={sessionId:this.sessionId,text:i,...a&&{locale:a},...r&&{output_format:r}},{audio:e}=await d.makeTTS(this.apiServer,t);if(this.pipelineSuppressed)return;return await M(e,s)}catch(t){t instanceof n||t instanceof h?this.setError(new c(t)):this.setError(t instanceof Error?t:new Error(String(t)))}finally{this.setChatState(null,exports.ChatState.TTS)}}startVoiceChat(){if(!this.perso)throw new Error("startVoiceChat requires WebRTC (STF mode)");return this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.RECORDING),this.perso.recordStart()}stopVoiceChat(){if(!this.perso)throw new Error("stopVoiceChat requires WebRTC (STF mode)");this.setChatState(exports.ChatState.ANALYZING,exports.ChatState.RECORDING),this.perso.recordEndStt()}async startProcessSTT(t){if(this.sttRecorder?.isRecording())throw new Error("STT recording is already in progress");this.pipelineSuppressed=!1,this.setChatState(exports.ChatState.RECORDING);try{this.sttRecorder=new O({targetSampleRate:16e3}),await this.sttRecorder.start(),t&&t>0&&(this.sttTimeoutHandle=setTimeout(async()=>{if(this.sttTimeoutHandle=null,this.sttRecorder?.isRecording()){try{this.sttTimeoutAudioFile=await this.sttRecorder.stop()}catch{this.sttTimeoutAudioFile=null,this.setChatState(null,exports.ChatState.RECORDING)}this.sttRecorder=null}},t))}catch(t){throw this.setChatState(null,exports.ChatState.RECORDING),this.sttRecorder=null,t}}lastRecordedAudioFile=null;async stopProcessSTT(t){let e;if(this.sttTimeoutHandle&&(clearTimeout(this.sttTimeoutHandle),this.sttTimeoutHandle=null),this.setChatState(null,exports.ChatState.RECORDING),this.sttTimeoutAudioFile)e=this.sttTimeoutAudioFile,this.sttTimeoutAudioFile=null;else{if(!this.sttRecorder?.isRecording())throw this.sttRecorder?(this.sttRecorder=null,new Error("STT recording is not in progress")):new Error("STT recording has not been started");e=await this.sttRecorder.stop(),this.sttRecorder=null}this.lastRecordedAudioFile=e;try{return(await d.makeSTT(this.apiServer,this.sessionId,e,t)).text}catch(t){if(t instanceof n)throw new l(t);throw t}}isSTTRecording(){return(this.sttRecorder?.isRecording()??!1)||null!==this.sttTimeoutAudioFile}changeSize(t,e){this.perso?.changeSize(t,e)}async clearBuffer(){this.perso?.clearBuffer(),await this.clearLLMJob(),null!==this.lastStfTimeoutHandle&&(clearTimeout(this.lastStfTimeoutHandle),this.lastStfTimeoutHandle=null),this.pipelineSuppressed=!0,this.resetChatState()}setSrc(t){t.srcObject=this.getRemoteStream()??null}getRemoteStream(){return this.perso?.getStream()}getLocalStream(){return this.stream}stopSession(){this.close()}onClose(t){return this.perso?this.perso.subscribeStatus(e=>{null!=e.detail&&!1===e.detail.live&&t(200===e.detail.code)}):()=>{}}subscribeChatStates(t){const e=e=>{t(e.detail.status)};return this.chatStatesHandler.addEventListener("status",e),()=>{this.chatStatesHandler.removeEventListener("status",e)}}subscribeChatLog(t){const e=e=>{t(e.detail.chatLog)};return this.chatLogHandler.addEventListener("chatLog",e),()=>{this.chatLogHandler.removeEventListener("chatLog",e)}}setSttResultCallback(t){const e=e=>{t(e.detail)};return this.sttEventHandler=new EventTarget,this.sttEventHandler.addEventListener("stt",e),()=>{this.sttEventHandler?.removeEventListener("stt",e),this.sttEventHandler=null}}setErrorHandler(t){const e=e=>{t(e.detail.error)};return this.errorHandler.addEventListener("error",e),()=>{this.errorHandler.removeEventListener("error",e)}}getSessionId(){return this.sessionId}async processChatInternal(t){this.setChatState(exports.ChatState.LLM);const e=this.clientTools.map(t=>({type:"function",function:{description:t.description,name:t.name,parameters:t.parameters}})),s=new Array;null===t||(t instanceof Array?s.push(...t):"string"==typeof t&&s.push({role:"user",content:t}));const a=await fetch(`${this.apiServer}/api/v1/session/${this.sessionId}/llm/v2/`,{body:JSON.stringify({messages:[...this.messageHistory,...s],tools:e}),headers:{"Content-Type":"application/json"},method:"POST"});if(!a.ok){const t=await a.json(),e=new o(new n(a.status,t.errors[0].code,t.errors[0].detail,t.errors[0].attr));return this.setError(e),void this.setChatState(null,exports.ChatState.LLM)}const r=a.body?.getReader(),l=new TextDecoder("utf-8");let c="",h=null,d="";for(;;){const{done:t,value:e}=await r.read();if(t)break;let a;for(d+=l.decode(e,{stream:!0});-1!==(a=d.indexOf("\n"));){if(this.llmCancel)return c.length>0&&this.addMessageToChatLog(c,!1),void this.setChatState(null,exports.ChatState.LLM);const t=d.slice(0,a).trim();if(d=d.slice(a+1),!t.startsWith("data: {")){const t=new o(new i("Failed to parse SSE response"));return this.setError(t),void this.setChatState(null,exports.ChatState.LLM)}const e=JSON.parse(t.slice(6).trim());if("success"!==e.status){const t=new o(new i(e.reason));return this.setError(t),void this.setChatState(null,exports.ChatState.LLM)}c.length>0&&"message"!=e.type&&(s.push({role:"assistant",type:"message",content:c}),this.addMessageToChatLog(c,!1),c=""),"message"!==e.type?"tool_call"!==e.type||null==e.tool_calls?"tool"!==e.role||"tool_call"===e.type&&s.push({role:e.role,type:e.type,content:e.content,tool_call_id:e.tool_call_id}):(s.push({role:"assistant",type:e.type,content:e.content,tool_calls:e.tool_calls}),h=e):(c+=N(e.content),this.processTTSTFInternal(e.content))}}if(this.llmCancel)this.setChatState(null,exports.ChatState.LLM);else{if(null!=h){const t=[];for(const e of h.tool_calls){const s=this.getChatTool(this.clientTools,e.function.name);null!=s&&t.push(new Promise(async t=>{try{const a=await s.call(JSON.parse(e.function.arguments));t({toolCallId:e.id,chatTool:s,chatToolResult:a})}catch(a){t({toolCallId:e.id,chatTool:s,chatToolResult:{result:"error!"}})}}))}const e=await Promise.all(t);for(const t of e)s.push({role:"tool",content:JSON.stringify(t.chatToolResult),tool_call_id:t.toolCallId});const a=e.length>0&&h.tool_calls.length!==e.length,r=e.some(t=>!t.chatTool.executeOnly);a||r?await this.processChatInternal(s):this.messageHistory.push(...s)}else this.messageHistory.push(...s);this.setChatState(null,exports.ChatState.LLM)}}getChatTool(t,e){for(const s of t)if(s.name===e)return s;return null}llmCancel=!1;pipelineSuppressed=!1;async clearLLMJob(){null!=this.llmJob&&(this.llmCancel=!0,await this.llmJob,this.llmCancel=!1)}processTTSTFInternal(t){const e=N(t).trim();0!==e.length&&this.perso&&(this.setChatState(exports.ChatState.ANALYZING),this.perso.ttstf(e))}addMessageToChatLog(t,e){this.chatLog=[{text:t,isUser:e,timestamp:new Date},...this.chatLog],this.chatLogHandler.dispatchEvent(new CustomEvent("chatLog",{detail:{chatLog:this.chatLog}}))}setChatState(t=null,e=null){const s=new Map(this.chatStateMap);function a(t){t===exports.ChatState.ANALYZING?s.set(t,(s.get(t)||0)+1):s.set(t,1)}function r(t){t===exports.ChatState.ANALYZING?s.set(t,Math.max((s.get(t)||0)-1,0)):s.set(t,0)}if(null!=t)if(t instanceof Array)for(let e of t)a(e);else a(t);if(null!=e)if(e instanceof Array)for(let t of e)r(t);else r(e);const n=this.exchangeChatStateMapToSet(this.chatStateMap),o=this.exchangeChatStateMapToSet(s);this.chatStateMap=s,this.isEqualChatStateMap(n,o)||this.dispatchChatState(o)}resetChatState(){this.chatStateMap=new Map([[exports.ChatState.RECORDING,0],[exports.ChatState.LLM,0],[exports.ChatState.ANALYZING,0],[exports.ChatState.SPEAKING,0],[exports.ChatState.TTS,0]]),this.dispatchChatState(this.exchangeChatStateMapToSet(this.chatStateMap))}exchangeChatStateMapToSet(t){const e=new Set;for(const s of t)s[1]>0&&e.add(s[0]);return e}dispatchChatState(t){this.chatStatesHandler.dispatchEvent(new CustomEvent("status",{detail:{status:t}}))}isEqualChatStateMap(t,e){if(t.size!==e.size)return!1;for(const s of t)if(t.has(s)!==e.has(s))return!1;return!0}async logSessionEvent(t){const e="object"==typeof t?JSON.stringify(t):t;await d.sessionEvent(this.apiServer,this.sessionId,s.SESSION_LOG,e)}setError(t){this.errorHandler.dispatchEvent(new CustomEvent("error",{detail:{error:t}}))}close(){this.stopHeartbeat(),this.perso?.closeSelf()}startHeartbeat(){const t=async()=>{try{await d.sessionEvent(this.apiServer,this.sessionId,s.SESSION_DURING),null!==this.heartbeatIntervalId&&(this.heartbeatIntervalId=setTimeout(t,1e4))}catch(t){t instanceof n?this.setError(t):this.setError(t instanceof Error?t:new Error(String(t))),this.close()}};this.heartbeatIntervalId=setTimeout(t,1e4)}stopHeartbeat(){null!==this.heartbeatIntervalId&&(clearTimeout(this.heartbeatIntervalId),this.heartbeatIntervalId=null)}}async function F(t,e,s,a,r,n){if("boolean"!=typeof r){const n=await P.create(t,e,s,a);return new H(t,e,n,r)}const o=n??[];let i,l;if(r)i=await navigator.mediaDevices.getUserMedia({audio:!0,video:!1}),l=()=>{};else{const t=new AudioContext,e=t.createOscillator();e.frequency.value=0;const s=t.createMediaStreamDestination();e.connect(s),e.start(),i=s.stream,l=()=>{e.stop(),e.disconnect(s),t.close()}}const c=await P.create(t,e,s,a,i);if(!c)return l(),new H(t,e,null,o);const h=new H(t,e,c,o,{stream:i,legacyVoiceChatMode:!0});return h.onClose(()=>{l()}),h}exports.ApiError=n,exports.ChatTool=class{name;description;parameters;call;executeOnly;constructor(t,e,s,a,r=!1){this.name=t,this.description=e,this.parameters=s,this.call=a,this.executeOnly=r}},exports.LLMError=o,exports.LLMStreamingResponseError=i,exports.LlmProcessor=A,exports.STTError=l,exports.Session=H,exports.TTSDecodeError=h,exports.TTSError=c,exports.TTS_TARGET_SAMPLE_RATE=L,exports.WavRecorder=O,exports.createSession=async function(t,e,s,a,r,n){return"boolean"==typeof r?await F(t,e,s,a,r,n??[]):await F(t,e,s,a,r)},exports.createSessionId=async function(t,s,a){let r;if("undefined"!=typeof window&&console.warn("[perso-interactive-sdk-web] WARNING: createSessionId is being called from the browser. This exposes your API key and is not recommended for production. Use server-side session creation with 'perso-interactive-sdk-web/server' instead. See: https://github.com/perso-ai/perso-interactive-sdk-web#server-side"),"string"==typeof a){const n=await d.getSessionTemplate(t,s,a);if("webrtc"!==n.model_style.platform_type)throw new Error(`SessionTemplate "${a}" uses platform_type "${n.model_style.platform_type}", but only "webrtc" is supported`);r=function(t){const s=e=>t.capability.some(t=>t.name===e);return{using_stf_webrtc:s(e.STF_WEBRTC),model_style:t.model_style.name,prompt:t.prompt.prompt_id,document:t.document?.document_id,background_image:t.background_image?.backgroundimage_id,mcp_servers:t.mcp_servers?.length?t.mcp_servers.map(t=>t.mcpserver_id):[],llm_type:s(e.LLM)?t.llm_type.name:void 0,tts_type:s(e.TTS)?t.tts_type.name:void 0,stt_type:s(e.STT)?t.stt_type.name:void 0,text_normalization_config:t.text_normalization_config?.textnormalizationconfig_id,text_normalization_locale:t.text_normalization_locale,stt_text_normalization_config:t.stt_text_normalization_config?.textnormalizationconfig_id,stt_text_normalization_locale:t.stt_text_normalization_locale,padding_left:t.padding_left??void 0,padding_top:t.padding_top??void 0,padding_height:t.padding_height??void 0}}(n)}else r=a;const n={capability:[],...r};r.using_stf_webrtc&&n.capability.push(e.STF_WEBRTC),r?.llm_type&&(n.capability.push(e.LLM),n.llm_type=r.llm_type),r?.tts_type&&(n.capability.push(e.TTS),n.tts_type=r.tts_type),r?.stt_type&&(n.capability.push(e.STT),n.stt_type=r.stt_type);const o=await fetch(`${t}/api/v1/session/`,{body:JSON.stringify(n),headers:{"PersoLive-APIKey":s,"Content-Type":"application/json"},method:"POST"});return(await d.parseJson(o)).session_id},exports.createWavRecorder=function(t){return new O(t)},exports.getAllSettings=async function(t,e){const[s,a,r,n,o,i,l,c,h]=await Promise.all([u(t,e),p(t,e),g(t,e),m(t,e),S(t,e),f(t,e),y(t,e),w(t,e),T(t,e).catch(()=>[])]);return{llms:s,ttsTypes:a,sttTypes:r,modelStyles:n,backgroundImages:o,prompts:i,documents:l,mcpServers:c,textNormalizations:h}},exports.getBackgroundImages=S,exports.getDocuments=y,exports.getLLMs=u,exports.getMcpServers=w,exports.getModelStyles=m,exports.getPrompts=f,exports.getSTTs=g,exports.getSessionInfo=async function(t,e){return await d.getSessionInfo(t,e)},exports.getSessionTemplates=async function(t,e){return await d.getSessionTemplates(t,e)},exports.getTTSs=p,exports.getTextNormalization=async function(t,e,s){return await d.downloadTextNormalization(t,e,s)},exports.getTextNormalizations=T,exports.getWavSampleRate=function(t){const e=new DataView(t);if(t.byteLength<28)throw new C("File too small to be a valid WAV");if("RIFF"!==v(e,0,4))throw new C("Missing RIFF header");let s=12;for(;s<t.byteLength-8;){const a=v(e,s,4),r=e.getUint32(s+4,!0);if("fmt "===a){if(s+16>t.byteLength)throw new C("fmt chunk extends beyond file");return e.getUint32(s+12,!0)}const n=s+8+r;if(n<=s)break;s=n}throw new C("Missing fmt chunk")},exports.makeTTS=async function(t,e){return await d.makeTTS(t,e)};

package/dist/client/index.d.ts

@@ -92,6 +92,8 @@ interface SessionTemplate {
     stt_type: STTType;
     text_normalization_config?: TextNormalizationConfig | null;
     text_normalization_locale?: string | null;
+    stt_text_normalization_config?: TextNormalizationConfig | null;
+    stt_text_normalization_locale?: string | null;
     model_style: ModelStyle;
     background_image: BackgroundImage | null;
     agent: string | null;
@@ -380,7 +382,20 @@ declare class Session {
      */
     processTTSTF(message: string): void;
     transcribeAudio(audio: Blob | File, language?: string): Promise<string>;
-    processSTF(file: Blob, format: string, message: string): Promise<string>;
+    /**
+     * Sends an audio file for Speech-to-Face processing.
+     *
+     * `format` is optional — when omitted, undefined, or supplied as a MIME type
+     * (`audio/wav` / `audio/mpeg`), it's normalized via `normalizeAudioFormat`,
+     * which falls back to `file.type`. Pass `'wav'` or `'mp3'` directly when the
+     * source format is already known and authoritative.
+     *
+     * @param file Audio blob to lip-sync.
+     * @param format Optional format hint. Canonical (`'wav'`/`'mp3'`), MIME, or omitted.
+     * @param message Optional text caption.
+     * @returns File reference returned by the server.
+     */
+    processSTF(file: Blob, format?: string, message?: string): Promise<string>;
     processTTS(message: string, options?: {
         resample?: boolean;
         locale?: string;
@@ -727,6 +742,7 @@ declare function getDocuments(apiServer: string, apiKey: string): Promise<any>;
  * @param apiServer Perso API server URL.
  * @param apiKey API key used for authentication.
  */
+declare function getMcpServers(apiServer: string, apiKey: string): Promise<any>;
 /**
  * Retrieves available text normalization options.
  * @param apiServer Perso API server URL.
@@ -741,7 +757,6 @@ declare function getTextNormalizations(apiServer: string, apiKey: string): Promi
  * @param configId Text Normalization Config ID.
  */
 declare function getTextNormalization(apiServer: string, apiKey: string, configId: string): Promise<TextNormalizationDownload>;
-declare function getMcpServers(apiServer: string, apiKey: string): Promise<any>;
 /**
  * Retrieves the list of session templates.
  * @param apiServer Perso API server URL.
@@ -766,21 +781,6 @@ declare function getAllSettings(apiServer: string, apiKey: string): Promise<{
     mcpServers: any;
     textNormalizations: any;
 }>;
-/**
- * Creates a Session with REST-based STT/TTS (current mode).
- */
-declare function createSession(apiServer: string, sessionId: string, width: number, height: number, clientTools: Array<ChatTool>): Promise<Session>;
-/**
- * Creates a Session with bidirectional WebRTC audio (legacy mode).
- * @deprecated Legacy voice chat mode will be removed in a future version.
- *   Use the 5-argument overload with REST-based STT/TTS instead.
- */
-declare function createSession(apiServer: string, sessionId: string, width: number, height: number, enableVoiceChat: boolean, clientTools: Array<ChatTool>): Promise<Session>;
-/**
- * Retrieves metadata for an existing session.
- * @param apiServer Perso API server URL.
- * @param sessionId Session id to inspect.
- */
 /**
  * Sends text to the TTS API and returns Base64-encoded audio.
  * @param apiServer Perso API server URL.
@@ -795,8 +795,24 @@ declare function makeTTS(apiServer: string, params: {
 }): Promise<{
     audio: string;
 }>;
+/**
+ * Retrieves metadata for an existing session.
+ * @param apiServer Perso API server URL.
+ * @param sessionId Session id to inspect.
+ */
 declare function getSessionInfo(apiServer: string, sessionId: string): Promise<any>;
 
+/**
+ * Creates a Session with REST-based STT/TTS (current mode).
+ */
+declare function createSession(apiServer: string, sessionId: string, width: number, height: number, clientTools: Array<ChatTool>): Promise<Session>;
+/**
+ * Creates a Session with bidirectional WebRTC audio (legacy mode).
+ * @deprecated Legacy voice chat mode will be removed in a future version.
+ *   Use the 5-argument overload with REST-based STT/TTS instead.
+ */
+declare function createSession(apiServer: string, sessionId: string, width: number, height: number, enableVoiceChat: boolean, clientTools: Array<ChatTool>): Promise<Session>;
+
 type CreateSessionIdBody = {
     using_stf_webrtc: boolean;
     model_style: string;
@@ -812,6 +828,8 @@ type CreateSessionIdBody = {
     stt_type?: string;
     text_normalization_config?: string;
     text_normalization_locale?: string | null;
+    stt_text_normalization_config?: string;
+    stt_text_normalization_locale?: string | null;
 };
 /**
  * Requests a new session creation ID by POSTing the desired runtime options to