assemblyai 4.2.2 → 4.3.0

package/CHANGELOG.md

@@ -1,5 +1,19 @@
 # Changelog
 
+## [4.3.0] - 2024-02-15
+
+### Added
+
+- Add `RealtimeTranscriber.configureEndUtteranceSilenceThreshold` function
+- Add `RealtimeTranscriber.forceEndUtterance` function
+- Add `end_utterance_silence_threshold` property to `CreateRealtimeTranscriberParams` and `RealtimeTranscriberParams` types.
+
+## [4.2.3] - 2024-02-13
+
+### Added
+
+- Add `speech_model` field to `TranscriptParams` and add `SpeechModel` type.
+
 ## [4.2.2] - 2024-01-29
 
 ### Changed

package/dist/assemblyai.umd.js

@@ -195,6 +195,8 @@
     }
 
     const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+    const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+    const terminateSessionMessage = `{"terminate_session":true}`;
     class RealtimeTranscriber {
         constructor(params) {
             var _a, _b;
@@ -203,6 +205,8 @@
             this.sampleRate = (_b = params.sampleRate) !== null && _b !== void 0 ? _b : 16000;
             this.wordBoost = params.wordBoost;
             this.encoding = params.encoding;
+            this.end_utterance_silence_threshold =
+                params.end_utterance_silence_threshold;
             if ("token" in params && params.token)
                 this.token = params.token;
             if ("apiKey" in params && params.apiKey)
@@ -249,6 +253,13 @@
                     });
                 }
                 this.socket.binaryType = "arraybuffer";
+                this.socket.onopen = () => {
+                    if (this.end_utterance_silence_threshold === undefined ||
+                        this.end_utterance_silence_threshold === null) {
+                        return;
+                    }
+                    this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+                };
                 this.socket.onclose = ({ code, reason }) => {
                     var _a, _b;
                     if (!reason) {
@@ -305,10 +316,7 @@
             });
         }
         sendAudio(audio) {
-            if (!this.socket || this.socket.readyState !== WebSocket$1.OPEN) {
-                throw new Error("Socket is not open for communication");
-            }
-            this.socket.send(audio);
+            this.send(audio);
         }
         stream() {
             return new WritableStream({
@@ -317,11 +325,30 @@
                 },
             });
         }
+        /**
+         * Manually end an utterance
+         */
+        forceEndUtterance() {
+            this.send(forceEndOfUtteranceMessage);
+        }
+        /**
+         * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+         * @param threshold The duration of the end utterance silence threshold in milliseconds
+         * @format integer
+         */
+        configureEndUtteranceSilenceThreshold(threshold) {
+            this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+        }
+        send(data) {
+            if (!this.socket || this.socket.readyState !== WebSocket$1.OPEN) {
+                throw new Error("Socket is not open for communication");
+            }
+            this.socket.send(data);
+        }
         close(waitForSessionTermination = true) {
             return __awaiter(this, void 0, void 0, function* () {
                 if (this.socket) {
                     if (this.socket.readyState === WebSocket$1.OPEN) {
-                        const terminateSessionMessage = `{"terminate_session": true}`;
                         if (waitForSessionTermination) {
                             const sessionTerminatedPromise = new Promise((resolve) => {
                                 this.sessionTerminatedResolve = resolve;

package/dist/assemblyai.umd.min.js

@@ -1 +1 @@
-!function(e,t){"object"==typeof exports&&"undefined"!=typeof module?t(exports):"function"==typeof define&&define.amd?define(["exports"],t):t((e="undefined"!=typeof globalThis?globalThis:e||self).assemblyai={})}(this,(function(e){"use strict";function t(e,t,s,i){return new(s||(s=Promise))((function(o,n){function r(e){try{c(i.next(e))}catch(e){n(e)}}function a(e){try{c(i.throw(e))}catch(e){n(e)}}function c(e){var t;e.done?o(e.value):(t=e.value,t instanceof s?t:new s((function(e){e(t)}))).then(r,a)}c((i=i.apply(e,t||[])).next())}))}"function"==typeof SuppressedError&&SuppressedError;class s{constructor(e){this.params=e}fetch(e,s){var i;return t(this,void 0,void 0,(function*(){(s=null!=s?s:{}).headers=null!==(i=s.headers)&&void 0!==i?i:{},s.headers=Object.assign({Authorization:this.params.apiKey,"Content-Type":"application/json"},s.headers),e.startsWith("http")||(e=this.params.baseUrl+e);const t=yield fetch(e,s);if(t.status>=400){let e;const s=yield t.text();if(s){try{e=JSON.parse(s)}catch(e){}if(null==e?void 0:e.error)throw new Error(e.error);throw new Error(s)}throw new Error(`HTTP Error: ${t.status} ${t.statusText}`)}return t}))}fetchJson(e,s){return t(this,void 0,void 0,(function*(){return(yield this.fetch(e,s)).json()}))}}class i extends s{summary(e){return this.fetchJson("/lemur/v3/generate/summary",{method:"POST",body:JSON.stringify(e)})}questionAnswer(e){return this.fetchJson("/lemur/v3/generate/question-answer",{method:"POST",body:JSON.stringify(e)})}actionItems(e){return this.fetchJson("/lemur/v3/generate/action-items",{method:"POST",body:JSON.stringify(e)})}task(e){return this.fetchJson("/lemur/v3/generate/task",{method:"POST",body:JSON.stringify(e)})}purgeRequestData(e){return this.fetchJson(`/lemur/v3/${e}`,{method:"DELETE"})}}const{WritableStream:o}="undefined"!=typeof window?window:"undefined"!=typeof global?global:globalThis;var n=null;"undefined"!=typeof WebSocket?n=WebSocket:"undefined"!=typeof MozWebSocket?n=MozWebSocket:"undefined"!=typeof global?n=global.WebSocket||global.MozWebSocket:"undefined"!=typeof window?n=window.WebSocket||window.MozWebSocket:"undefined"!=typeof self&&(n=self.WebSocket||self.MozWebSocket);var r,a=n;!function(e){e[e.BadSampleRate=4e3]="BadSampleRate",e[e.AuthFailed=4001]="AuthFailed",e[e.InsufficientFundsOrFreeAccount=4002]="InsufficientFundsOrFreeAccount",e[e.NonexistentSessionId=4004]="NonexistentSessionId",e[e.SessionExpired=4008]="SessionExpired",e[e.ClosedSession=4010]="ClosedSession",e[e.RateLimited=4029]="RateLimited",e[e.UniqueSessionViolation=4030]="UniqueSessionViolation",e[e.SessionTimeout=4031]="SessionTimeout",e[e.AudioTooShort=4032]="AudioTooShort",e[e.AudioTooLong=4033]="AudioTooLong",e[e.BadJson=4100]="BadJson",e[e.BadSchema=4101]="BadSchema",e[e.TooManyStreams=4102]="TooManyStreams",e[e.Reconnected=4103]="Reconnected",e[e.ReconnectAttemptsExhausted=1013]="ReconnectAttemptsExhausted"}(r||(r={}));const c={[r.BadSampleRate]:"Sample rate must be a positive integer",[r.AuthFailed]:"Not Authorized",[r.InsufficientFundsOrFreeAccount]:"Insufficient funds or you are using a free account. This feature is paid-only and requires you to add a credit card. Please visit https://assemblyai.com/dashboard/ to add a credit card to your account.",[r.NonexistentSessionId]:"Session ID does not exist",[r.SessionExpired]:"Session has expired",[r.ClosedSession]:"Session is closed",[r.RateLimited]:"Rate limited",[r.UniqueSessionViolation]:"Unique session violation",[r.SessionTimeout]:"Session Timeout",[r.AudioTooShort]:"Audio too short",[r.AudioTooLong]:"Audio too long",[r.BadJson]:"Bad JSON",[r.BadSchema]:"Bad schema",[r.TooManyStreams]:"Too many streams",[r.Reconnected]:"Reconnected",[r.ReconnectAttemptsExhausted]:"Reconnect attempts exhausted"};class l extends Error{}class d{constructor(e){var t,s;if(this.listeners={},this.realtimeUrl=null!==(t=e.realtimeUrl)&&void 0!==t?t:"wss://api.assemblyai.com/v2/realtime/ws",this.sampleRate=null!==(s=e.sampleRate)&&void 0!==s?s:16e3,this.wordBoost=e.wordBoost,this.encoding=e.encoding,"token"in e&&e.token&&(this.token=e.token),"apiKey"in e&&e.apiKey&&(this.apiKey=e.apiKey),!this.token&&!this.apiKey)throw new Error("API key or temporary token is required.")}connectionUrl(){const e=new URL(this.realtimeUrl);if("wss:"!==e.protocol)throw new Error("Invalid protocol, must be wss");const t=new URLSearchParams;return this.token&&t.set("token",this.token),t.set("sample_rate",this.sampleRate.toString()),this.wordBoost&&this.wordBoost.length>0&&t.set("word_boost",JSON.stringify(this.wordBoost)),this.encoding&&t.set("encoding",this.encoding),e.search=t.toString(),e}on(e,t){this.listeners[e]=t}connect(){return new Promise((e=>{if(this.socket)throw new Error("Already connected");const t=this.connectionUrl();this.token?this.socket=new a(t.toString()):this.socket=new a(t.toString(),{headers:{Authorization:this.apiKey}}),this.socket.binaryType="arraybuffer",this.socket.onclose=({code:e,reason:t})=>{var s,i;t||e in r&&(t=c[e]),null===(i=(s=this.listeners).close)||void 0===i||i.call(s,e,t)},this.socket.onerror=e=>{var t,s,i,o;e.error?null===(s=(t=this.listeners).error)||void 0===s||s.call(t,e.error):null===(o=(i=this.listeners).error)||void 0===o||o.call(i,new Error(e.message))},this.socket.onmessage=({data:t})=>{var s,i,o,n,r,a,c,d,u,h,f,p,m;const y=JSON.parse(t.toString());if("error"in y)null===(i=(s=this.listeners).error)||void 0===i||i.call(s,new l(y.error));else switch(y.message_type){case"SessionBegins":{const t={sessionId:y.session_id,expiresAt:new Date(y.expires_at)};e(t),null===(n=(o=this.listeners).open)||void 0===n||n.call(o,t);break}case"PartialTranscript":y.created=new Date(y.created),null===(a=(r=this.listeners).transcript)||void 0===a||a.call(r,y),null===(d=(c=this.listeners)["transcript.partial"])||void 0===d||d.call(c,y);break;case"FinalTranscript":y.created=new Date(y.created),null===(h=(u=this.listeners).transcript)||void 0===h||h.call(u,y),null===(p=(f=this.listeners)["transcript.final"])||void 0===p||p.call(f,y);break;case"SessionTerminated":null===(m=this.sessionTerminatedResolve)||void 0===m||m.call(this)}}}))}sendAudio(e){if(!this.socket||this.socket.readyState!==a.OPEN)throw new Error("Socket is not open for communication");this.socket.send(e)}stream(){return new o({write:e=>{this.sendAudio(e)}})}close(e=!0){return t(this,void 0,void 0,(function*(){if(this.socket){if(this.socket.readyState===a.OPEN){const t='{"terminate_session": true}';if(e){const e=new Promise((e=>{this.sessionTerminatedResolve=e}));this.socket.send(t),yield e}else this.socket.send(t)}"removeAllListeners"in this.socket&&this.socket.removeAllListeners(),this.socket.close()}this.listeners={},this.socket=void 0}))}}class u extends s{constructor(e){super(e),this.rtFactoryParams=e}createService(e){return this.transcriber(e)}transcriber(e){const t=Object.assign({},e);return t.token||t.apiKey||(t.apiKey=this.rtFactoryParams.apiKey),new d(t)}createTemporaryToken(e){return t(this,void 0,void 0,(function*(){return(yield this.fetchJson("/v2/realtime/token",{method:"POST",body:JSON.stringify(e)})).token}))}}function h(e){return e.startsWith("http")||e.startsWith("https")?null:e.startsWith("file://")?e.substring(7):e.startsWith("file:")?e.substring(5):e}class f extends s{constructor(e,t){super(e),this.files=t}transcribe(e,s){return t(this,void 0,void 0,(function*(){const t=yield this.submit(e);return yield this.waitUntilReady(t.id,s)}))}submit(e){return t(this,void 0,void 0,(function*(){const{audio:t}=e,s=function(e,t){var s={};for(var i in e)Object.prototype.hasOwnProperty.call(e,i)&&t.indexOf(i)<0&&(s[i]=e[i]);if(null!=e&&"function"==typeof Object.getOwnPropertySymbols){var o=0;for(i=Object.getOwnPropertySymbols(e);o<i.length;o++)t.indexOf(i[o])<0&&Object.prototype.propertyIsEnumerable.call(e,i[o])&&(s[i[o]]=e[i[o]])}return s}(e,["audio"]);let i;if("string"==typeof t){const e=h(t);i=null!==e?yield this.files.upload(e):t}else i=yield this.files.upload(t);return yield this.fetchJson("/v2/transcript",{method:"POST",body:JSON.stringify(Object.assign(Object.assign({},s),{audio_url:i}))})}))}create(e,s){var i;return t(this,void 0,void 0,(function*(){const t=h(e.audio_url);if(null!==t){const s=yield this.files.upload(t);e.audio_url=s}const o=yield this.fetchJson("/v2/transcript",{method:"POST",body:JSON.stringify(e)});return null===(i=null==s?void 0:s.poll)||void 0===i||i?yield this.waitUntilReady(o.id,s):o}))}waitUntilReady(e,s){var i,o;return t(this,void 0,void 0,(function*(){const t=null!==(i=null==s?void 0:s.pollingInterval)&&void 0!==i?i:3e3,n=null!==(o=null==s?void 0:s.pollingTimeout)&&void 0!==o?o:-1,r=Date.now();for(;;){const s=yield this.get(e);if("completed"===s.status||"error"===s.status)return s;if(n>0&&Date.now()-r>n)throw new Error("Polling timeout");yield new Promise((e=>setTimeout(e,t)))}}))}get(e){return this.fetchJson(`/v2/transcript/${e}`)}list(e){return t(this,void 0,void 0,(function*(){let t="/v2/transcript";"string"==typeof e?t=e:e&&(t=`${t}?${new URLSearchParams(Object.keys(e).map((t=>{var s;return[t,(null===(s=e[t])||void 0===s?void 0:s.toString())||""]})))}`);const s=yield this.fetchJson(t);for(const e of s.transcripts)e.created=new Date(e.created),e.completed&&(e.completed=new Date(e.completed));return s}))}delete(e){return this.fetchJson(`/v2/transcript/${e}`,{method:"DELETE"})}wordSearch(e,t){const s=new URLSearchParams({words:t.join(",")});return this.fetchJson(`/v2/transcript/${e}/word-search?${s.toString()}`)}sentences(e){return this.fetchJson(`/v2/transcript/${e}/sentences`)}paragraphs(e){return this.fetchJson(`/v2/transcript/${e}/paragraphs`)}subtitles(e,s="srt",i){return t(this,void 0,void 0,(function*(){let t=`/v2/transcript/${e}/${s}`;if(i){const e=new URLSearchParams;e.set("chars_per_caption",i.toString()),t+=`?${e.toString()}`}const o=yield this.fetch(t);return yield o.text()}))}redactions(e){return this.fetchJson(`/v2/transcript/${e}/redacted-audio`)}}class p extends s{upload(e){return t(this,void 0,void 0,(function*(){let s;s="string"==typeof e?yield function(e){return t(this,void 0,void 0,(function*(){throw new Error("Interacting with the file system is not supported in this environment.")}))}():e;return(yield this.fetchJson("/v2/upload",{method:"POST",body:s,headers:{"Content-Type":"application/octet-stream"},duplex:"half"})).upload_url}))}}e.AssemblyAI=class{constructor(e){e.baseUrl=e.baseUrl||"https://api.assemblyai.com",e.baseUrl&&e.baseUrl.endsWith("/")&&(e.baseUrl=e.baseUrl.slice(0,-1)),this.files=new p(e),this.transcripts=new f(e,this.files),this.lemur=new i(e),this.realtime=new u(e)}},e.FileService=p,e.LemurService=i,e.RealtimeService=class extends d{},e.RealtimeServiceFactory=class extends u{},e.RealtimeTranscriber=d,e.RealtimeTranscriberFactory=u,e.TranscriptService=f}));
+!function(e,t){"object"==typeof exports&&"undefined"!=typeof module?t(exports):"function"==typeof define&&define.amd?define(["exports"],t):t((e="undefined"!=typeof globalThis?globalThis:e||self).assemblyai={})}(this,(function(e){"use strict";function t(e,t,s,i){return new(s||(s=Promise))((function(n,o){function r(e){try{c(i.next(e))}catch(e){o(e)}}function a(e){try{c(i.throw(e))}catch(e){o(e)}}function c(e){var t;e.done?n(e.value):(t=e.value,t instanceof s?t:new s((function(e){e(t)}))).then(r,a)}c((i=i.apply(e,t||[])).next())}))}"function"==typeof SuppressedError&&SuppressedError;class s{constructor(e){this.params=e}fetch(e,s){var i;return t(this,void 0,void 0,(function*(){(s=null!=s?s:{}).headers=null!==(i=s.headers)&&void 0!==i?i:{},s.headers=Object.assign({Authorization:this.params.apiKey,"Content-Type":"application/json"},s.headers),e.startsWith("http")||(e=this.params.baseUrl+e);const t=yield fetch(e,s);if(t.status>=400){let e;const s=yield t.text();if(s){try{e=JSON.parse(s)}catch(e){}if(null==e?void 0:e.error)throw new Error(e.error);throw new Error(s)}throw new Error(`HTTP Error: ${t.status} ${t.statusText}`)}return t}))}fetchJson(e,s){return t(this,void 0,void 0,(function*(){return(yield this.fetch(e,s)).json()}))}}class i extends s{summary(e){return this.fetchJson("/lemur/v3/generate/summary",{method:"POST",body:JSON.stringify(e)})}questionAnswer(e){return this.fetchJson("/lemur/v3/generate/question-answer",{method:"POST",body:JSON.stringify(e)})}actionItems(e){return this.fetchJson("/lemur/v3/generate/action-items",{method:"POST",body:JSON.stringify(e)})}task(e){return this.fetchJson("/lemur/v3/generate/task",{method:"POST",body:JSON.stringify(e)})}purgeRequestData(e){return this.fetchJson(`/lemur/v3/${e}`,{method:"DELETE"})}}const{WritableStream:n}="undefined"!=typeof window?window:"undefined"!=typeof global?global:globalThis;var o=null;"undefined"!=typeof WebSocket?o=WebSocket:"undefined"!=typeof MozWebSocket?o=MozWebSocket:"undefined"!=typeof global?o=global.WebSocket||global.MozWebSocket:"undefined"!=typeof window?o=window.WebSocket||window.MozWebSocket:"undefined"!=typeof self&&(o=self.WebSocket||self.MozWebSocket);var r,a=o;!function(e){e[e.BadSampleRate=4e3]="BadSampleRate",e[e.AuthFailed=4001]="AuthFailed",e[e.InsufficientFundsOrFreeAccount=4002]="InsufficientFundsOrFreeAccount",e[e.NonexistentSessionId=4004]="NonexistentSessionId",e[e.SessionExpired=4008]="SessionExpired",e[e.ClosedSession=4010]="ClosedSession",e[e.RateLimited=4029]="RateLimited",e[e.UniqueSessionViolation=4030]="UniqueSessionViolation",e[e.SessionTimeout=4031]="SessionTimeout",e[e.AudioTooShort=4032]="AudioTooShort",e[e.AudioTooLong=4033]="AudioTooLong",e[e.BadJson=4100]="BadJson",e[e.BadSchema=4101]="BadSchema",e[e.TooManyStreams=4102]="TooManyStreams",e[e.Reconnected=4103]="Reconnected",e[e.ReconnectAttemptsExhausted=1013]="ReconnectAttemptsExhausted"}(r||(r={}));const c={[r.BadSampleRate]:"Sample rate must be a positive integer",[r.AuthFailed]:"Not Authorized",[r.InsufficientFundsOrFreeAccount]:"Insufficient funds or you are using a free account. This feature is paid-only and requires you to add a credit card. Please visit https://assemblyai.com/dashboard/ to add a credit card to your account.",[r.NonexistentSessionId]:"Session ID does not exist",[r.SessionExpired]:"Session has expired",[r.ClosedSession]:"Session is closed",[r.RateLimited]:"Rate limited",[r.UniqueSessionViolation]:"Unique session violation",[r.SessionTimeout]:"Session Timeout",[r.AudioTooShort]:"Audio too short",[r.AudioTooLong]:"Audio too long",[r.BadJson]:"Bad JSON",[r.BadSchema]:"Bad schema",[r.TooManyStreams]:"Too many streams",[r.Reconnected]:"Reconnected",[r.ReconnectAttemptsExhausted]:"Reconnect attempts exhausted"};class l extends Error{}const d='{"terminate_session":true}';class h{constructor(e){var t,s;if(this.listeners={},this.realtimeUrl=null!==(t=e.realtimeUrl)&&void 0!==t?t:"wss://api.assemblyai.com/v2/realtime/ws",this.sampleRate=null!==(s=e.sampleRate)&&void 0!==s?s:16e3,this.wordBoost=e.wordBoost,this.encoding=e.encoding,this.end_utterance_silence_threshold=e.end_utterance_silence_threshold,"token"in e&&e.token&&(this.token=e.token),"apiKey"in e&&e.apiKey&&(this.apiKey=e.apiKey),!this.token&&!this.apiKey)throw new Error("API key or temporary token is required.")}connectionUrl(){const e=new URL(this.realtimeUrl);if("wss:"!==e.protocol)throw new Error("Invalid protocol, must be wss");const t=new URLSearchParams;return this.token&&t.set("token",this.token),t.set("sample_rate",this.sampleRate.toString()),this.wordBoost&&this.wordBoost.length>0&&t.set("word_boost",JSON.stringify(this.wordBoost)),this.encoding&&t.set("encoding",this.encoding),e.search=t.toString(),e}on(e,t){this.listeners[e]=t}connect(){return new Promise((e=>{if(this.socket)throw new Error("Already connected");const t=this.connectionUrl();this.token?this.socket=new a(t.toString()):this.socket=new a(t.toString(),{headers:{Authorization:this.apiKey}}),this.socket.binaryType="arraybuffer",this.socket.onopen=()=>{void 0!==this.end_utterance_silence_threshold&&null!==this.end_utterance_silence_threshold&&this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold)},this.socket.onclose=({code:e,reason:t})=>{var s,i;t||e in r&&(t=c[e]),null===(i=(s=this.listeners).close)||void 0===i||i.call(s,e,t)},this.socket.onerror=e=>{var t,s,i,n;e.error?null===(s=(t=this.listeners).error)||void 0===s||s.call(t,e.error):null===(n=(i=this.listeners).error)||void 0===n||n.call(i,new Error(e.message))},this.socket.onmessage=({data:t})=>{var s,i,n,o,r,a,c,d,h,u,f,p,m;const y=JSON.parse(t.toString());if("error"in y)null===(i=(s=this.listeners).error)||void 0===i||i.call(s,new l(y.error));else switch(y.message_type){case"SessionBegins":{const t={sessionId:y.session_id,expiresAt:new Date(y.expires_at)};e(t),null===(o=(n=this.listeners).open)||void 0===o||o.call(n,t);break}case"PartialTranscript":y.created=new Date(y.created),null===(a=(r=this.listeners).transcript)||void 0===a||a.call(r,y),null===(d=(c=this.listeners)["transcript.partial"])||void 0===d||d.call(c,y);break;case"FinalTranscript":y.created=new Date(y.created),null===(u=(h=this.listeners).transcript)||void 0===u||u.call(h,y),null===(p=(f=this.listeners)["transcript.final"])||void 0===p||p.call(f,y);break;case"SessionTerminated":null===(m=this.sessionTerminatedResolve)||void 0===m||m.call(this)}}}))}sendAudio(e){this.send(e)}stream(){return new n({write:e=>{this.sendAudio(e)}})}forceEndUtterance(){this.send('{"force_end_utterance":true}')}configureEndUtteranceSilenceThreshold(e){this.send(`{"end_utterance_silence_threshold":${e}}`)}send(e){if(!this.socket||this.socket.readyState!==a.OPEN)throw new Error("Socket is not open for communication");this.socket.send(e)}close(e=!0){return t(this,void 0,void 0,(function*(){if(this.socket){if(this.socket.readyState===a.OPEN)if(e){const e=new Promise((e=>{this.sessionTerminatedResolve=e}));this.socket.send(d),yield e}else this.socket.send(d);"removeAllListeners"in this.socket&&this.socket.removeAllListeners(),this.socket.close()}this.listeners={},this.socket=void 0}))}}class u extends s{constructor(e){super(e),this.rtFactoryParams=e}createService(e){return this.transcriber(e)}transcriber(e){const t=Object.assign({},e);return t.token||t.apiKey||(t.apiKey=this.rtFactoryParams.apiKey),new h(t)}createTemporaryToken(e){return t(this,void 0,void 0,(function*(){return(yield this.fetchJson("/v2/realtime/token",{method:"POST",body:JSON.stringify(e)})).token}))}}function f(e){return e.startsWith("http")||e.startsWith("https")?null:e.startsWith("file://")?e.substring(7):e.startsWith("file:")?e.substring(5):e}class p extends s{constructor(e,t){super(e),this.files=t}transcribe(e,s){return t(this,void 0,void 0,(function*(){const t=yield this.submit(e);return yield this.waitUntilReady(t.id,s)}))}submit(e){return t(this,void 0,void 0,(function*(){const{audio:t}=e,s=function(e,t){var s={};for(var i in e)Object.prototype.hasOwnProperty.call(e,i)&&t.indexOf(i)<0&&(s[i]=e[i]);if(null!=e&&"function"==typeof Object.getOwnPropertySymbols){var n=0;for(i=Object.getOwnPropertySymbols(e);n<i.length;n++)t.indexOf(i[n])<0&&Object.prototype.propertyIsEnumerable.call(e,i[n])&&(s[i[n]]=e[i[n]])}return s}(e,["audio"]);let i;if("string"==typeof t){const e=f(t);i=null!==e?yield this.files.upload(e):t}else i=yield this.files.upload(t);return yield this.fetchJson("/v2/transcript",{method:"POST",body:JSON.stringify(Object.assign(Object.assign({},s),{audio_url:i}))})}))}create(e,s){var i;return t(this,void 0,void 0,(function*(){const t=f(e.audio_url);if(null!==t){const s=yield this.files.upload(t);e.audio_url=s}const n=yield this.fetchJson("/v2/transcript",{method:"POST",body:JSON.stringify(e)});return null===(i=null==s?void 0:s.poll)||void 0===i||i?yield this.waitUntilReady(n.id,s):n}))}waitUntilReady(e,s){var i,n;return t(this,void 0,void 0,(function*(){const t=null!==(i=null==s?void 0:s.pollingInterval)&&void 0!==i?i:3e3,o=null!==(n=null==s?void 0:s.pollingTimeout)&&void 0!==n?n:-1,r=Date.now();for(;;){const s=yield this.get(e);if("completed"===s.status||"error"===s.status)return s;if(o>0&&Date.now()-r>o)throw new Error("Polling timeout");yield new Promise((e=>setTimeout(e,t)))}}))}get(e){return this.fetchJson(`/v2/transcript/${e}`)}list(e){return t(this,void 0,void 0,(function*(){let t="/v2/transcript";"string"==typeof e?t=e:e&&(t=`${t}?${new URLSearchParams(Object.keys(e).map((t=>{var s;return[t,(null===(s=e[t])||void 0===s?void 0:s.toString())||""]})))}`);const s=yield this.fetchJson(t);for(const e of s.transcripts)e.created=new Date(e.created),e.completed&&(e.completed=new Date(e.completed));return s}))}delete(e){return this.fetchJson(`/v2/transcript/${e}`,{method:"DELETE"})}wordSearch(e,t){const s=new URLSearchParams({words:t.join(",")});return this.fetchJson(`/v2/transcript/${e}/word-search?${s.toString()}`)}sentences(e){return this.fetchJson(`/v2/transcript/${e}/sentences`)}paragraphs(e){return this.fetchJson(`/v2/transcript/${e}/paragraphs`)}subtitles(e,s="srt",i){return t(this,void 0,void 0,(function*(){let t=`/v2/transcript/${e}/${s}`;if(i){const e=new URLSearchParams;e.set("chars_per_caption",i.toString()),t+=`?${e.toString()}`}const n=yield this.fetch(t);return yield n.text()}))}redactions(e){return this.fetchJson(`/v2/transcript/${e}/redacted-audio`)}}class m extends s{upload(e){return t(this,void 0,void 0,(function*(){let s;s="string"==typeof e?yield function(e){return t(this,void 0,void 0,(function*(){throw new Error("Interacting with the file system is not supported in this environment.")}))}():e;return(yield this.fetchJson("/v2/upload",{method:"POST",body:s,headers:{"Content-Type":"application/octet-stream"},duplex:"half"})).upload_url}))}}e.AssemblyAI=class{constructor(e){e.baseUrl=e.baseUrl||"https://api.assemblyai.com",e.baseUrl&&e.baseUrl.endsWith("/")&&(e.baseUrl=e.baseUrl.slice(0,-1)),this.files=new m(e),this.transcripts=new p(e,this.files),this.lemur=new i(e),this.realtime=new u(e)}},e.FileService=m,e.LemurService=i,e.RealtimeService=class extends h{},e.RealtimeServiceFactory=class extends u{},e.RealtimeTranscriber=h,e.RealtimeTranscriberFactory=u,e.TranscriptService=p}));

package/dist/bun.mjs

@@ -130,6 +130,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         this.listeners = {};
@@ -137,6 +139,8 @@ class RealtimeTranscriber {
         this.sampleRate = params.sampleRate ?? 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -183,6 +187,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 if (!reason) {
                     if (code in RealtimeErrorType) {
@@ -236,10 +247,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new WritableStream({
@@ -248,10 +256,29 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     async close(waitForSessionTermination = true) {
         if (this.socket) {
             if (this.socket.readyState === WebSocket.OPEN) {
-                const terminateSessionMessage = `{"terminate_session": true}`;
                 if (waitForSessionTermination) {
                     const sessionTerminatedPromise = new Promise((resolve) => {
                         this.sessionTerminatedResolve = resolve;

package/dist/deno.mjs

@@ -130,6 +130,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         this.listeners = {};
@@ -137,6 +139,8 @@ class RealtimeTranscriber {
         this.sampleRate = params.sampleRate ?? 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -183,6 +187,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 if (!reason) {
                     if (code in RealtimeErrorType) {
@@ -236,10 +247,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new WritableStream({
@@ -248,10 +256,29 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     async close(waitForSessionTermination = true) {
         if (this.socket) {
             if (this.socket.readyState === WebSocket.OPEN) {
-                const terminateSessionMessage = `{"terminate_session": true}`;
                 if (waitForSessionTermination) {
                     const sessionTerminatedPromise = new Promise((resolve) => {
                         this.sessionTerminatedResolve = resolve;

package/dist/index.cjs

@@ -177,6 +177,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         var _a, _b;
@@ -185,6 +187,8 @@ class RealtimeTranscriber {
         this.sampleRate = (_b = params.sampleRate) !== null && _b !== void 0 ? _b : 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -231,6 +235,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 var _a, _b;
                 if (!reason) {
@@ -287,10 +298,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new WritableStream({
@@ -299,11 +307,30 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     close(waitForSessionTermination = true) {
         return __awaiter(this, void 0, void 0, function* () {
             if (this.socket) {
                 if (this.socket.readyState === WebSocket.OPEN) {
-                    const terminateSessionMessage = `{"terminate_session": true}`;
                     if (waitForSessionTermination) {
                         const sessionTerminatedPromise = new Promise((resolve) => {
                             this.sessionTerminatedResolve = resolve;

package/dist/index.mjs

@@ -175,6 +175,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         var _a, _b;
@@ -183,6 +185,8 @@ class RealtimeTranscriber {
         this.sampleRate = (_b = params.sampleRate) !== null && _b !== void 0 ? _b : 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -229,6 +233,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 var _a, _b;
                 if (!reason) {
@@ -285,10 +296,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new WritableStream({
@@ -297,11 +305,30 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     close(waitForSessionTermination = true) {
         return __awaiter(this, void 0, void 0, function* () {
             if (this.socket) {
                 if (this.socket.readyState === WebSocket.OPEN) {
-                    const terminateSessionMessage = `{"terminate_session": true}`;
                     if (waitForSessionTermination) {
                         const sessionTerminatedPromise = new Promise((resolve) => {
                             this.sessionTerminatedResolve = resolve;

package/dist/node.cjs

@@ -129,6 +129,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         this.listeners = {};
@@ -136,6 +138,8 @@ class RealtimeTranscriber {
         this.sampleRate = params.sampleRate ?? 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -182,6 +186,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 if (!reason) {
                     if (code in RealtimeErrorType) {
@@ -235,10 +246,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new web.WritableStream({
@@ -247,10 +255,29 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     async close(waitForSessionTermination = true) {
         if (this.socket) {
             if (this.socket.readyState === WebSocket.OPEN) {
-                const terminateSessionMessage = `{"terminate_session": true}`;
                 if (waitForSessionTermination) {
                     const sessionTerminatedPromise = new Promise((resolve) => {
                         this.sessionTerminatedResolve = resolve;

package/dist/node.mjs

@@ -127,6 +127,8 @@ class RealtimeError extends Error {
 }
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
 class RealtimeTranscriber {
     constructor(params) {
         this.listeners = {};
@@ -134,6 +136,8 @@ class RealtimeTranscriber {
         this.sampleRate = params.sampleRate ?? 16000;
         this.wordBoost = params.wordBoost;
         this.encoding = params.encoding;
+        this.end_utterance_silence_threshold =
+            params.end_utterance_silence_threshold;
         if ("token" in params && params.token)
             this.token = params.token;
         if ("apiKey" in params && params.apiKey)
@@ -180,6 +184,13 @@ class RealtimeTranscriber {
                 });
             }
             this.socket.binaryType = "arraybuffer";
+            this.socket.onopen = () => {
+                if (this.end_utterance_silence_threshold === undefined ||
+                    this.end_utterance_silence_threshold === null) {
+                    return;
+                }
+                this.configureEndUtteranceSilenceThreshold(this.end_utterance_silence_threshold);
+            };
             this.socket.onclose = ({ code, reason }) => {
                 if (!reason) {
                     if (code in RealtimeErrorType) {
@@ -233,10 +244,7 @@ class RealtimeTranscriber {
         });
     }
     sendAudio(audio) {
-        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-            throw new Error("Socket is not open for communication");
-        }
-        this.socket.send(audio);
+        this.send(audio);
     }
     stream() {
         return new WritableStream({
@@ -245,10 +253,29 @@ class RealtimeTranscriber {
             },
         });
     }
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance() {
+        this.send(forceEndOfUtteranceMessage);
+    }
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold) {
+        this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+    }
+    send(data) {
+        if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+            throw new Error("Socket is not open for communication");
+        }
+        this.socket.send(data);
+    }
     async close(waitForSessionTermination = true) {
         if (this.socket) {
             if (this.socket.readyState === WebSocket.OPEN) {
-                const terminateSessionMessage = `{"terminate_session": true}`;
                 if (waitForSessionTermination) {
                     const sessionTerminatedPromise = new Promise((resolve) => {
                         this.sessionTerminatedResolve = resolve;

package/dist/services/realtime/service.d.ts

@@ -6,6 +6,7 @@ export declare class RealtimeTranscriber {
     private encoding?;
     private apiKey?;
     private token?;
+    private end_utterance_silence_threshold?;
     private socket?;
     private listeners;
     private sessionTerminatedResolve?;
@@ -20,6 +21,17 @@ export declare class RealtimeTranscriber {
     connect(): Promise<SessionBeginsEventData>;
     sendAudio(audio: AudioData): void;
     stream(): WritableStream<AudioData>;
+    /**
+     * Manually end an utterance
+     */
+    forceEndUtterance(): void;
+    /**
+     * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+     * @param threshold The duration of the end utterance silence threshold in milliseconds
+     * @format integer
+     */
+    configureEndUtteranceSilenceThreshold(threshold: number): void;
+    private send;
     close(waitForSessionTermination?: boolean): Promise<void>;
 }
 /**

package/dist/types/asyncapi.generated.d.ts

@@ -9,6 +9,11 @@ export type AudioData = ArrayBufferLike;
  * @enum {string}
  */
 export type AudioEncoding = "pcm_s16le" | "pcm_mulaw";
+/** @description Configure the threshold for how long to wait before ending an utterance. Default is 700ms. */
+export type ConfigureEndUtteranceSilenceThreshold = {
+    /** @description The duration threshold in milliseconds */
+    end_utterance_silence_threshold: number;
+};
 export type FinalTranscript = RealtimeBaseTranscript & {
     /**
      * @description Describes the type of message
@@ -20,6 +25,11 @@ export type FinalTranscript = RealtimeBaseTranscript & {
     /** @description Whether the text is formatted, for example Dollar -> $ */
     text_formatted: boolean;
 };
+/** @description Manually end an utterance */
+export type ForceEndUtterance = {
+    /** @description A boolean value to communicate that you wish to force the end of the utterance */
+    force_end_utterance: boolean;
+};
 /** @enum {string} */
 export type MessageType = "SessionBegins" | "PartialTranscript" | "FinalTranscript" | "SessionTerminated";
 export type PartialTranscript = RealtimeBaseTranscript & {

package/dist/types/openapi.generated.d.ts

@@ -991,6 +991,11 @@ export type SeverityScoreSummary = {
     /** Format: double */
     medium: number;
 };
+/**
+ * @description The speech model to use for the transcription.
+ * @enum {string}
+ */
+export type SpeechModel = "nano";
 /**
  * @description The replacement logic for detected PII, can be "entity_type" or "hash". See [PII redaction](https://www.assemblyai.com/docs/models/pii-redaction) for more details.
  * @enum {string}
@@ -1187,6 +1192,7 @@ export type TopicDetectionResult = {
  * @description A transcript object
  * @example {
  *   "id": "9ea68fd3-f953-42c1-9742-976c447fb463",
+ *   "speech_model": null,
  *   "language_model": "assemblyai_default",
  *   "acoustic_model": "assemblyai_default",
  *   "language_code": "en_us",
@@ -1989,6 +1995,11 @@ export type Transcript = {
     speaker_labels?: boolean | null;
     /** @description Tell the speaker label model how many speakers it should attempt to identify, up to 10. See [Speaker diarization](https://www.assemblyai.com/docs/models/speaker-diarization) for more details. */
     speakers_expected?: number | null;
+    /**
+     * @description The speech model used for the transcription. When `null`, the default model is used.
+     * @default null
+     */
+    speech_model: SpeechModel | null;
     /**
      * Format: float
      * @description Defaults to null. Reject audio files that contain less than this fraction of speech.
@@ -2123,6 +2134,7 @@ export type TranscriptListItem = {
 /**
  * @description The parameters for creating a transcript
  * @example {
+ *   "speech_model": null,
  *   "language_code": "en_us",
  *   "punctuate": true,
  *   "format_text": true,
@@ -2228,6 +2240,11 @@ export type TranscriptOptionalParams = {
      * @default null
      */
     speakers_expected?: number | null;
+    /**
+     * @description The speech model to use for the transcription. When `null`, the default model is used.
+     * @default null
+     */
+    speech_model?: SpeechModel | null;
     /**
      * Format: float
      * @description Reject audio files that contain less than this fraction of speech.

package/dist/types/realtime/index.d.ts

@@ -1,12 +1,35 @@
 import { AudioEncoding, FinalTranscript, PartialTranscript, RealtimeTranscript, RealtimeTranscriptType } from "../asyncapi.generated";
 type CreateRealtimeTranscriberParams = {
+    /**
+     * The WebSocket URL that the RealtimeTranscriber connects to
+     */
     realtimeUrl?: string;
+    /**
+     * The sample rate of the streamed audio
+     */
     sampleRate?: number;
+    /**
+     * Add up to 2500 characters of custom vocabulary
+     */
     wordBoost?: string[];
+    /**
+     * The encoding of the audio data
+     */
     encoding?: AudioEncoding;
+    /**
+     * The duration of the end utterance silence threshold in milliseconds
+     */
+    end_utterance_silence_threshold?: number;
 } & ({
+    /**
+     * The API key used to authenticate the RealtimeTranscriber
+     * Using an API key to authenticate the RealtimeTranscriber is not supported in the browser.
+     */
     apiKey?: string;
 } | {
+    /**
+     * The temporary token used to authenticate the RealtimeTranscriber
+     */
     token: string;
 });
 /**
@@ -14,13 +37,36 @@ type CreateRealtimeTranscriberParams = {
  */
 type CreateRealtimeServiceParams = CreateRealtimeTranscriberParams;
 type RealtimeTranscriberParams = {
+    /**
+     * The WebSocket URL that the RealtimeTranscriber connects to
+     */
     realtimeUrl?: string;
+    /**
+     * The sample rate of the streamed audio
+     */
     sampleRate?: number;
+    /**
+     * Add up to 2500 characters of custom vocabulary
+     */
     wordBoost?: string[];
+    /**
+     * The encoding of the audio data
+     */
     encoding?: AudioEncoding;
+    /**
+     * The duration of the end utterance silence threshold in milliseconds
+     */
+    end_utterance_silence_threshold?: number;
 } & ({
+    /**
+     * The API key used to authenticate the RealtimeTranscriber.
+     * Using an API key to authenticate the RealtimeTranscriber is not supported in the browser.
+     */
     apiKey: string;
 } | {
+    /**
+     * The temporary token used to authenticate the RealtimeTranscriber
+     */
     token: string;
 });
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "assemblyai",
-  "version": "4.2.2",
+  "version": "4.3.0",
   "description": "The AssemblyAI JavaScript SDK provides an easy-to-use interface for interacting with the AssemblyAI API, which supports async and real-time transcription, as well as the latest LeMUR models.",
   "engines": {
     "node": ">=18"

package/src/services/realtime/service.ts

@@ -20,6 +20,26 @@ import {
 } from "../../utils/errors";
 
 const defaultRealtimeUrl = "wss://api.assemblyai.com/v2/realtime/ws";
+const forceEndOfUtteranceMessage = `{"force_end_utterance":true}`;
+const terminateSessionMessage = `{"terminate_session":true}`;
+
+type BufferLike =
+  | string
+  | Buffer
+  | DataView
+  | number
+  | ArrayBufferView
+  | Uint8Array
+  | ArrayBuffer
+  | SharedArrayBuffer
+  | ReadonlyArray<unknown>
+  | ReadonlyArray<number>
+  | { valueOf(): ArrayBuffer }
+  | { valueOf(): SharedArrayBuffer }
+  | { valueOf(): Uint8Array }
+  | { valueOf(): ReadonlyArray<number> }
+  | { valueOf(): string }
+  | { [Symbol.toPrimitive](hint: string): string };
 
 export class RealtimeTranscriber {
   private realtimeUrl: string;
@@ -28,6 +48,7 @@ export class RealtimeTranscriber {
   private encoding?: AudioEncoding;
   private apiKey?: string;
   private token?: string;
+  private end_utterance_silence_threshold?: number;
   private socket?: WebSocket;
   private listeners: RealtimeListeners = {};
   private sessionTerminatedResolve?: () => void;
@@ -37,6 +58,8 @@ export class RealtimeTranscriber {
     this.sampleRate = params.sampleRate ?? 16_000;
     this.wordBoost = params.wordBoost;
     this.encoding = params.encoding;
+    this.end_utterance_silence_threshold =
+      params.end_utterance_silence_threshold;
     if ("token" in params && params.token) this.token = params.token;
     if ("apiKey" in params && params.apiKey) this.apiKey = params.apiKey;
 
@@ -105,6 +128,18 @@ export class RealtimeTranscriber {
       }
       this.socket.binaryType = "arraybuffer";
 
+      this.socket.onopen = () => {
+        if (
+          this.end_utterance_silence_threshold === undefined ||
+          this.end_utterance_silence_threshold === null
+        ) {
+          return;
+        }
+        this.configureEndUtteranceSilenceThreshold(
+          this.end_utterance_silence_threshold
+        );
+      };
+
       this.socket.onclose = ({ code, reason }: CloseEvent) => {
         if (!reason) {
           if (code in RealtimeErrorType) {
@@ -159,10 +194,7 @@ export class RealtimeTranscriber {
   }
 
   sendAudio(audio: AudioData) {
-    if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
-      throw new Error("Socket is not open for communication");
-    }
-    this.socket.send(audio);
+    this.send(audio);
   }
 
   stream(): WritableStream<AudioData> {
@@ -173,10 +205,32 @@ export class RealtimeTranscriber {
     });
   }
 
+  /**
+   * Manually end an utterance
+   */
+  forceEndUtterance() {
+    this.send(forceEndOfUtteranceMessage);
+  }
+
+  /**
+   * Configure the threshold for how long to wait before ending an utterance. Default is 700ms.
+   * @param threshold The duration of the end utterance silence threshold in milliseconds
+   * @format integer
+   */
+  configureEndUtteranceSilenceThreshold(threshold: number) {
+    this.send(`{"end_utterance_silence_threshold":${threshold}}`);
+  }
+
+  private send(data: BufferLike) {
+    if (!this.socket || this.socket.readyState !== WebSocket.OPEN) {
+      throw new Error("Socket is not open for communication");
+    }
+    this.socket.send(data);
+  }
+
   async close(waitForSessionTermination = true) {
     if (this.socket) {
       if (this.socket.readyState === WebSocket.OPEN) {
-        const terminateSessionMessage = `{"terminate_session": true}`;
         if (waitForSessionTermination) {
           const sessionTerminatedPromise = new Promise<void>((resolve) => {
             this.sessionTerminatedResolve = resolve;

package/src/types/asyncapi.generated.ts

@@ -28,6 +28,12 @@ export type AudioData = ArrayBufferLike;
  */
 export type AudioEncoding = "pcm_s16le" | "pcm_mulaw";
 
+/** @description Configure the threshold for how long to wait before ending an utterance. Default is 700ms. */
+export type ConfigureEndUtteranceSilenceThreshold = {
+  /** @description The duration threshold in milliseconds */
+  end_utterance_silence_threshold: number;
+};
+
 export type FinalTranscript = RealtimeBaseTranscript & {
   /**
    * @description Describes the type of message
@@ -40,6 +46,12 @@ export type FinalTranscript = RealtimeBaseTranscript & {
   text_formatted: boolean;
 };
 
+/** @description Manually end an utterance */
+export type ForceEndUtterance = {
+  /** @description A boolean value to communicate that you wish to force the end of the utterance */
+  force_end_utterance: boolean;
+};
+
 /** @enum {string} */
 export type MessageType =
   | "SessionBegins"

package/src/types/openapi.generated.ts

@@ -1102,6 +1102,12 @@ export type SeverityScoreSummary = {
   medium: number;
 };
 
+/**
+ * @description The speech model to use for the transcription.
+ * @enum {string}
+ */
+export type SpeechModel = "nano";
+
 /**
  * @description The replacement logic for detected PII, can be "entity_type" or "hash". See [PII redaction](https://www.assemblyai.com/docs/models/pii-redaction) for more details.
  * @enum {string}
@@ -1310,6 +1316,7 @@ export type TopicDetectionResult = {
  * @description A transcript object
  * @example {
  *   "id": "9ea68fd3-f953-42c1-9742-976c447fb463",
+ *   "speech_model": null,
  *   "language_model": "assemblyai_default",
  *   "acoustic_model": "assemblyai_default",
  *   "language_code": "en_us",
@@ -2112,6 +2119,11 @@ export type Transcript = {
   speaker_labels?: boolean | null;
   /** @description Tell the speaker label model how many speakers it should attempt to identify, up to 10. See [Speaker diarization](https://www.assemblyai.com/docs/models/speaker-diarization) for more details. */
   speakers_expected?: number | null;
+  /**
+   * @description The speech model used for the transcription. When `null`, the default model is used.
+   * @default null
+   */
+  speech_model: SpeechModel | null;
   /**
    * Format: float
    * @description Defaults to null. Reject audio files that contain less than this fraction of speech.
@@ -2272,6 +2284,7 @@ export type TranscriptListItem = {
 /**
  * @description The parameters for creating a transcript
  * @example {
+ *   "speech_model": null,
  *   "language_code": "en_us",
  *   "punctuate": true,
  *   "format_text": true,
@@ -2377,6 +2390,11 @@ export type TranscriptOptionalParams = {
    * @default null
    */
   speakers_expected?: number | null;
+  /**
+   * @description The speech model to use for the transcription. When `null`, the default model is used.
+   * @default null
+   */
+  speech_model?: SpeechModel | null;
   /**
    * Format: float
    * @description Reject audio files that contain less than this fraction of speech.

package/src/types/realtime/index.ts

@@ -7,15 +7,38 @@ import {
 } from "../asyncapi.generated";
 
 type CreateRealtimeTranscriberParams = {
+  /**
+   * The WebSocket URL that the RealtimeTranscriber connects to
+   */
   realtimeUrl?: string;
+  /**
+   * The sample rate of the streamed audio
+   */
   sampleRate?: number;
+  /**
+   * Add up to 2500 characters of custom vocabulary
+   */
   wordBoost?: string[];
+  /**
+   * The encoding of the audio data
+   */
   encoding?: AudioEncoding;
+  /**
+   * The duration of the end utterance silence threshold in milliseconds
+   */
+  end_utterance_silence_threshold?: number;
 } & (
   | {
+      /**
+       * The API key used to authenticate the RealtimeTranscriber
+       * Using an API key to authenticate the RealtimeTranscriber is not supported in the browser.
+       */
       apiKey?: string;
     }
   | {
+      /**
+       * The temporary token used to authenticate the RealtimeTranscriber
+       */
       token: string;
     }
 );
@@ -26,15 +49,38 @@ type CreateRealtimeTranscriberParams = {
 type CreateRealtimeServiceParams = CreateRealtimeTranscriberParams;
 
 type RealtimeTranscriberParams = {
+  /**
+   * The WebSocket URL that the RealtimeTranscriber connects to
+   */
   realtimeUrl?: string;
+  /**
+   * The sample rate of the streamed audio
+   */
   sampleRate?: number;
+  /**
+   * Add up to 2500 characters of custom vocabulary
+   */
   wordBoost?: string[];
+  /**
+   * The encoding of the audio data
+   */
   encoding?: AudioEncoding;
+  /**
+   * The duration of the end utterance silence threshold in milliseconds
+   */
+  end_utterance_silence_threshold?: number;
 } & (
   | {
+      /**
+       * The API key used to authenticate the RealtimeTranscriber.
+       * Using an API key to authenticate the RealtimeTranscriber is not supported in the browser.
+       */
       apiKey: string;
     }
   | {
+      /**
+       * The temporary token used to authenticate the RealtimeTranscriber
+       */
       token: string;
     }
 );