speechstate 2.0.0-beta.9 → 2.0.0

package/dist/index.js~

@@ -0,0 +1,5 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.speechstate = void 0;
+var speechstate_1 = require("./speechstate");
+Object.defineProperty(exports, "speechstate", { enumerable: true, get: function () { return speechstate_1.speechstate; } });

package/dist/speechstate.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"speechstate.d.ts","sourceRoot":"","sources":["../src/speechstate.ts"],"names":[],"mappings":"AAIA,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,UAAU,EAAE,mBAAmB,EAAE,MAAM,SAAS,CAAC;AAC5E,UAAU,SAAS;IACjB,QAAQ,EAAE,QAAQ,CAAC;IACnB,YAAY,CAAC,EAAE,YAAY,CAAC;IAC5B,MAAM,CAAC,EAAE,GAAG,CAAC;IACb,MAAM,CAAC,EAAE,GAAG,CAAC;CACd;AAED,wDAAwD;AACxD,KAAK,YAAY,GACb;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAChB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GAChC;IAAE,IAAI,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,mBAAmB,CAAA;CAAE,CAAC;AAEnD,uCAAuC;AACvC,KAAK,aAAa,GACd;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,cAAc,CAAA;CAAE,GACxB;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,gBAAgB,CAAA;CAAE,GAC1B;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,KAAK,EAAE,UAAU,EAAE,CAAC;IAAC,QAAQ,CAAC,EAAE,GAAG,CAAA;CAAE,CAAC;AAEhE,KAAK,YAAY,GACb;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,GACrB;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,GACrB;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC;AAE1B,KAAK,OAAO,GAAG,YAAY,GAAG,YAAY,GAAG,aAAa,CAAC;AAE3D,QAAA,MAAM,WAAW,odAuQhB,CAAC;AAEF,OAAO,EAAE,WAAW,EAAE,CAAC"}
+{"version":3,"file":"speechstate.d.ts","sourceRoot":"","sources":["../src/speechstate.ts"],"names":[],"mappings":"AAUA,OAAO,EAAE,QAAQ,EAAE,MAAM,EAAE,UAAU,EAAE,mBAAmB,EAAE,MAAM,SAAS,CAAC;AAC5E,UAAU,SAAS;IACjB,QAAQ,EAAE,QAAQ,CAAC;IACnB,YAAY,CAAC,EAAE,YAAY,CAAC;IAC5B,MAAM,CAAC,EAAE,GAAG,CAAC;IACb,MAAM,CAAC,EAAE,GAAG,CAAC;CACd;AAED,wDAAwD;AACxD,KAAK,YAAY,GACb;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAChB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GAChC;IAAE,IAAI,EAAE,QAAQ,CAAC;IAAC,KAAK,EAAE,mBAAmB,CAAA;CAAE,CAAC;AAEnD,uCAAuC;AACvC,KAAK,aAAa,GACd;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,cAAc,CAAA;CAAE,GACxB;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,gBAAgB,CAAA;CAAE,GAC1B;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,KAAK,EAAE,UAAU,EAAE,CAAC;IAAC,QAAQ,CAAC,EAAE,GAAG,CAAA;CAAE,CAAC;AAEhE,KAAK,YAAY,GACb;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,GACrB;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,GACrB;IAAE,IAAI,EAAE,WAAW,CAAA;CAAE,CAAC;AAE1B,KAAK,OAAO,GAAG,YAAY,GAAG,YAAY,GAAG,aAAa,CAAC;AAE3D,QAAA,MAAM,WAAW,odAqThB,CAAC;AAEF,OAAO,EAAE,WAAW,EAAE,CAAC"}

package/dist/speechstate.js

@@ -14,6 +14,9 @@ const speechstate = (0, xstate_1.createMachine)({
     states: {
         AsrTtsSpawner: {
             initial: "Idle",
+            on: {
+                STOP: { target: ".Stopped", actions: (0, xstate_1.assign)({ audioContext: null }) },
+            },
             states: {
                 Idle: { on: { PREPARE: "CreateAudioContext" } },
                 CreateAudioContext: {
@@ -31,6 +34,7 @@ const speechstate = (0, xstate_1.createMachine)({
                         (0, xstate_1.assign)({
                             ttsRef: ({ context, spawn }) => {
                                 return spawn(tts_1.ttsMachine, {
+                                    id: "ttsRef",
                                     input: {
                                         ttsDefaultVoice: context.settings.ttsDefaultVoice,
                                         audioContext: context.audioContext,
@@ -42,6 +46,7 @@ const speechstate = (0, xstate_1.createMachine)({
                         (0, xstate_1.assign)({
                             asrRef: ({ context, spawn }) => {
                                 return spawn(asr_1.asrMachine, {
+                                    id: "asrRef",
                                     input: {
                                         asrDefaultCompleteTimeout: context.settings.asrDefaultCompleteTimeout,
                                         asrDefaultNoInputTimeout: context.settings.asrDefaultNoInputTimeout,
@@ -55,12 +60,19 @@ const speechstate = (0, xstate_1.createMachine)({
                             },
                         }),
                     ],
+                    exit: [
+                        (0, xstate_1.stopChild)("ttsRef"),
+                        (0, xstate_1.stopChild)("asrRef"),
+                        (0, xstate_1.assign)({ ttsRef: undefined, asrRef: undefined }),
+                    ],
                     after: {
                         300000: {
                             target: "Spawn",
+                            reenter: true,
                         },
                     },
                 },
+                Stopped: {},
             },
         },
         AsrTtsManager: {
@@ -86,10 +98,14 @@ const speechstate = (0, xstate_1.createMachine)({
                     ],
                     target: ".Ready",
                 },
+                STOP: ".Stopped",
             },
             states: {
-                Initialize: {},
+                Initialize: {
+                    meta: { view: "not-ready" },
+                },
                 PreReady: {
+                    meta: { view: "not-ready" },
                     on: {
                         TTS_READY: {
                             actions: () => console.debug("[TTS→SpSt] TTS_READY"),
@@ -109,6 +125,7 @@ const speechstate = (0, xstate_1.createMachine)({
                     ],
                     states: {
                         Idle: {
+                            meta: { view: "idle" },
                             on: {
                                 LISTEN: { target: "WaitForRecogniser" },
                                 SPEAK: [
@@ -119,6 +136,7 @@ const speechstate = (0, xstate_1.createMachine)({
                             },
                         },
                         Speaking: {
+                            initial: "Proceed",
                             entry: [
                                 ({ event }) => console.debug("[SpSt→TTS] SPEAK", event.value),
                                 ({ context, event }) => context.ttsRef.send({
@@ -127,15 +145,8 @@ const speechstate = (0, xstate_1.createMachine)({
                                 }),
                             ],
                             on: {
-                                CONTROL: {
-                                    actions: [
-                                        () => console.debug("[SpSt→TTS] CONTROL"),
-                                        ({ context }) => context.ttsRef.send({
-                                            type: "CONTROL",
-                                        }),
-                                    ],
-                                },
                                 STOP: {
+                                    target: "#speechstate.AsrTtsManager.Stopped",
                                     actions: [
                                         () => console.debug("[SpSt→TTS] STOP"),
                                         ({ context }) => context.ttsRef.send({
@@ -157,8 +168,39 @@ const speechstate = (0, xstate_1.createMachine)({
                                     ],
                                 },
                             },
+                            states: {
+                                Proceed: {
+                                    meta: { view: "speaking" },
+                                    on: {
+                                        CONTROL: {
+                                            target: "Paused",
+                                            actions: [
+                                                () => console.debug("[SpSt→TTS] CONTROL"),
+                                                ({ context }) => context.ttsRef.send({
+                                                    type: "CONTROL",
+                                                }),
+                                            ],
+                                        },
+                                    },
+                                },
+                                Paused: {
+                                    meta: { view: "speaking-paused" },
+                                    on: {
+                                        CONTROL: {
+                                            target: "Proceed",
+                                            actions: [
+                                                () => console.debug("[SpSt→TTS] CONTROL"),
+                                                ({ context }) => context.ttsRef.send({
+                                                    type: "CONTROL",
+                                                }),
+                                            ],
+                                        },
+                                    },
+                                },
+                            },
                         },
                         WaitForRecogniser: {
+                            meta: { view: "idle" },
                             entry: [
                                 ({ event }) => console.debug("[SpSt→ASR] START", event.value),
                                 ({ context, event }) => context.asrRef.send({
@@ -177,8 +219,10 @@ const speechstate = (0, xstate_1.createMachine)({
                             },
                         },
                         Recognising: {
+                            meta: { view: "recognising" },
                             on: {
                                 CONTROL: {
+                                    /** TODO go to paused state? */
                                     actions: [
                                         () => console.debug("[SpSt→ASR] CONTROL"),
                                         ({ context }) => context.asrRef.send({
@@ -187,7 +231,7 @@ const speechstate = (0, xstate_1.createMachine)({
                                     ],
                                 },
                                 STOP: {
-                                    target: "Idle",
+                                    target: "#speechstate.AsrTtsManager.Stopped",
                                     actions: [
                                         () => console.debug("[SpSt→ASR] STOP"),
                                         ({ context }) => context.asrRef.send({
@@ -210,7 +254,8 @@ const speechstate = (0, xstate_1.createMachine)({
                         },
                     },
                 },
-                Fail: {},
+                Fail: { meta: { view: "error" } },
+                Stopped: { meta: { view: "stopped" } },
             },
         },
     },

package/dist/tts.d.ts

@@ -1,10 +1,10 @@
-import { AzureCredentials, Agenda } from "./types";
+import { AzureSpeechCredentials, Agenda } from "./types";
 interface MySpeechSynthesisUtterance extends SpeechSynthesisUtterance {
     new (s: string): any;
 }
 interface TTSInit {
     audioContext: AudioContext;
-    azureCredentials: string | AzureCredentials;
+    azureCredentials: string | AzureSpeechCredentials;
     ttsDefaultVoice: string;
 }
 interface TTSContext extends TTSInit {
@@ -14,6 +14,8 @@ interface TTSContext extends TTSInit {
     wsaVoice?: SpeechSynthesisVoice;
     wsaUtt?: MySpeechSynthesisUtterance;
     agenda?: Agenda;
+    buffer?: string;
+    utteranceFromStream?: string;
 }
 type TTSEvent = {
     type: "PREPARE";
@@ -34,9 +36,85 @@ type TTSEvent = {
     value: Agenda;
 } | {
     type: "TTS_STARTED";
+} | {
+    type: "STREAMING_CHUNK";
+    value: string;
+} | {
+    type: "STREAMING_DONE";
 } | {
     type: "SPEAK_COMPLETE";
 };
-export declare const ttsMachine: import("xstate").StateMachine<TTSContext, TTSEvent, Record<string, import("xstate").AnyActorRef>, import("xstate").ProvidedActor, import("xstate").ParameterizedObject, import("xstate").ParameterizedObject, string, import("xstate").StateValue, string, TTSInit, {}, import("xstate").ResolveTypegenMeta<import("xstate").TypegenDisabled, TTSEvent, import("xstate").ProvidedActor, import("xstate").ParameterizedObject, import("xstate").ParameterizedObject, string, string>>;
+export declare const ttsMachine: import("xstate").StateMachine<TTSContext, TTSEvent, {
+    [x: string]: import("xstate").ActorRef<import("xstate").CallbackSnapshot<{}>, import("xstate").EventObject> | import("xstate").ActorRef<import("xstate").PromiseSnapshot<string, any>, {
+        [k: string]: unknown;
+        type: string;
+    }> | import("xstate").ActorRef<import("xstate").CallbackSnapshot<Agenda>, import("xstate").EventObject>;
+}, {
+    src: "getToken";
+    logic: import("xstate").PromiseActorLogic<string, any>;
+    id: string;
+} | {
+    src: "createEventsFromStream";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, Agenda>;
+    id: string;
+} | {
+    src: "ponyfill";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, {}>;
+    id: string;
+} | {
+    src: "start";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, {}>;
+    id: string;
+}, {
+    type: "ttsStop";
+    params: unknown;
+} | {
+    type: "addFiller";
+    params: {};
+}, {
+    type: "bufferContainsUtterancePartReadyToBeSpoken";
+    params: unknown;
+} | {
+    type: "bufferIsNonEmpty";
+    params: unknown;
+}, "FILLER_DELAY", "GetToken" | "Fail" | "Ponyfill" | {
+    Ready?: "Idle" | {
+        BufferedSpeaker?: Required<{
+            Buffer?: "BufferIdle" | "Buffering" | "BufferingDone";
+            Speaker?: "SpeakingIdle" | "PrepareSpeech" | {
+                Speak?: "Go" | "Paused";
+            };
+        }>;
+        Speaking?: "Go" | "Paused";
+    };
+}, string, TTSInit, {}, import("xstate").ResolveTypegenMeta<import("xstate").TypegenDisabled, TTSEvent, {
+    src: "getToken";
+    logic: import("xstate").PromiseActorLogic<string, any>;
+    id: string;
+} | {
+    src: "createEventsFromStream";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, Agenda>;
+    id: string;
+} | {
+    src: "ponyfill";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, {}>;
+    id: string;
+} | {
+    src: "start";
+    logic: import("xstate").CallbackActorLogic<import("xstate").EventObject, {}>;
+    id: string;
+}, {
+    type: "ttsStop";
+    params: unknown;
+} | {
+    type: "addFiller";
+    params: {};
+}, {
+    type: "bufferContainsUtterancePartReadyToBeSpoken";
+    params: unknown;
+} | {
+    type: "bufferIsNonEmpty";
+    params: unknown;
+}, "FILLER_DELAY", string>>;
 export {};
 //# sourceMappingURL=tts.d.ts.map

package/dist/tts.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,gBAAgB,EAAE,MAAM,EAAE,MAAM,SAAS,CAAC;AAEnD,UAAU,0BAA2B,SAAQ,wBAAwB;IACnE,KAAK,CAAC,EAAE,MAAM,OAAE;CACjB;AAED,UAAU,OAAO;IACf,YAAY,EAAE,YAAY,CAAC;IAC3B,gBAAgB,EAAE,MAAM,GAAG,gBAAgB,CAAC;IAC5C,eAAe,EAAE,MAAM,CAAC;CACzB;AAED,UAAU,UAAW,SAAQ,OAAO;IAClC,uBAAuB,CAAC,EAAE,MAAM,CAAC;IACjC,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,eAAe,CAAC;IACzB,QAAQ,CAAC,EAAE,oBAAoB,CAAC;IAChC,MAAM,CAAC,EAAE,0BAA0B,CAAC;IACpC,MAAM,CAAC,EAAE,MAAM,CAAC;CACjB;AAED,KAAK,QAAQ,GACT;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAChB;IACE,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE;QACL,MAAM,EAAE,eAAe,CAAC;QACxB,MAAM,EAAE,0BAA0B,CAAC;KACpC,CAAC;CACH,GACD;IAAE,IAAI,EAAE,OAAO,CAAA;CAAE,GACjB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GAChC;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,gBAAgB,CAAA;CAAE,CAAC;AAE/B,eAAO,MAAM,UAAU,sdA+KtB,CAAC"}
+{"version":3,"file":"tts.d.ts","sourceRoot":"","sources":["../src/tts.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,sBAAsB,EAAE,MAAM,EAAE,MAAM,SAAS,CAAC;AAEzD,UAAU,0BAA2B,SAAQ,wBAAwB;IACnE,KAAK,CAAC,EAAE,MAAM,OAAE;CACjB;AAED,UAAU,OAAO;IACf,YAAY,EAAE,YAAY,CAAC;IAC3B,gBAAgB,EAAE,MAAM,GAAG,sBAAsB,CAAC;IAClD,eAAe,EAAE,MAAM,CAAC;CACzB;AAED,UAAU,UAAW,SAAQ,OAAO;IAClC,uBAAuB,CAAC,EAAE,MAAM,CAAC;IACjC,UAAU,CAAC,EAAE,MAAM,CAAC;IACpB,MAAM,CAAC,EAAE,eAAe,CAAC;IACzB,QAAQ,CAAC,EAAE,oBAAoB,CAAC;IAChC,MAAM,CAAC,EAAE,0BAA0B,CAAC;IACpC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,mBAAmB,CAAC,EAAE,MAAM,CAAC;CAC9B;AAED,KAAK,QAAQ,GACT;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,SAAS,CAAA;CAAE,GACnB;IAAE,IAAI,EAAE,MAAM,CAAA;CAAE,GAChB;IACE,IAAI,EAAE,OAAO,CAAC;IACd,KAAK,EAAE;QACL,MAAM,EAAE,eAAe,CAAC;QACxB,MAAM,EAAE,0BAA0B,CAAC;KACpC,CAAC;CACH,GACD;IAAE,IAAI,EAAE,OAAO,CAAA;CAAE,GACjB;IAAE,IAAI,EAAE,OAAO,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GAChC;IAAE,IAAI,EAAE,aAAa,CAAA;CAAE,GACvB;IAAE,IAAI,EAAE,iBAAiB,CAAC;IAAC,KAAK,EAAE,MAAM,CAAA;CAAE,GAC1C;IAAE,IAAI,EAAE,gBAAgB,CAAA;CAAE,GAC1B;IAAE,IAAI,EAAE,gBAAgB,CAAA;CAAE,CAAC;AAI/B,eAAO,MAAM,UAAU;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;2BA6arB,CAAC"}

package/dist/tts.js

@@ -5,13 +5,107 @@ const xstate_1 = require("xstate");
 const getToken_1 = require("./getToken");
 const TextToSpeech_1 = require("web-speech-cognitive-services/lib/SpeechServices/TextToSpeech");
 const REGION = "northeurope";
-exports.ttsMachine = (0, xstate_1.createMachine)({
-    id: "tts",
+const UTTERANCE_CHUNK_REGEX = /(^.*([!?]+|([.,]+\s)))/;
+exports.ttsMachine = (0, xstate_1.setup)({
     types: {},
+    actions: {
+        ttsStop: ({ context }) => {
+            context.wsaTTS.cancel();
+        },
+        addFiller: (0, xstate_1.assign)(({ context }) => {
+            const spaceIndex = context.buffer.lastIndexOf(" ");
+            return {
+                buffer: context.buffer.substring(0, spaceIndex) +
+                    " um," +
+                    context.buffer.substring(spaceIndex),
+            };
+        }),
+    },
+    actors: {
+        getToken: getToken_1.getToken,
+        createEventsFromStream: (0, xstate_1.fromCallback)(({ sendBack, input }) => {
+            const eventSource = new EventSource(input.stream);
+            eventSource.addEventListener("STREAMING_DONE", (_event) => {
+                console.log("received streaming done - closing event stream");
+                sendBack({ type: "STREAMING_DONE" });
+                eventSource.close();
+            });
+            eventSource.addEventListener("STREAMING_RESET", (_event) => {
+                console.log("received streaming reset");
+            });
+            eventSource.addEventListener("STREAMING_CHUNK", (event) => {
+                console.log("received streaming chunk:", event);
+                sendBack({ type: "STREAMING_CHUNK", value: event.data });
+            });
+        }),
+        ponyfill: (0, xstate_1.fromCallback)(({ sendBack, input }) => {
+            const ponyfill = (0, TextToSpeech_1.default)({
+                audioContext: input.audioContext,
+                credentials: {
+                    region: REGION, // TODO
+                    authorizationToken: input.azureAuthorizationToken,
+                },
+            });
+            const { speechSynthesis, SpeechSynthesisUtterance } = ponyfill;
+            const tts = speechSynthesis;
+            const ttsUtterance = SpeechSynthesisUtterance;
+            tts.addEventListener("voiceschanged", () => {
+                const voices = tts.getVoices();
+                if (voices.length > 0) {
+                    console.debug("[TTS] READY");
+                    sendBack({
+                        type: "READY",
+                        value: { wsaTTS: tts, wsaUtt: ttsUtterance },
+                    });
+                }
+                else {
+                    console.error("[TTS] No voices available");
+                    sendBack({ type: "ERROR" });
+                }
+            });
+        }),
+        start: (0, xstate_1.fromCallback)(({ sendBack, input }) => {
+            if (["", " "].includes(input.utterance)) {
+                console.debug("[TTS] SPEAK: ", input.utterance);
+                input.wsaTTS.speak("");
+            }
+            else {
+                console.debug("[TTS] SPEAK: ", input.utterance);
+                const content = wrapSSML(input.utterance, input.voice, input.ttsLexicon, 1); // todo speech rate;
+                const utterance = new input.wsaUtt(content);
+                utterance.addEventListener("start", () => {
+                    sendBack({ type: "TTS_STARTED" });
+                    console.debug("[TTS] TTS_STARTED");
+                });
+                utterance.addEventListener("end", () => {
+                    sendBack({ type: "SPEAK_COMPLETE" });
+                    console.debug("[TTS] SPEAK_COMPLETE");
+                });
+                input.wsaTTS.speak(utterance);
+            }
+        }),
+    },
+    guards: {
+        bufferContainsUtterancePartReadyToBeSpoken: ({ context }) => {
+            const m = context.buffer.match(UTTERANCE_CHUNK_REGEX);
+            return !!m;
+        },
+        bufferIsNonEmpty: ({ context }) => {
+            return !!context.buffer;
+        },
+    },
+    delays: {
+        FILLER_DELAY: ({ context }) => {
+            return context.agenda.fillerDelay;
+        },
+    },
+}).createMachine({
+    id: "tts",
     context: ({ input }) => ({
         ttsDefaultVoice: input.ttsDefaultVoice || "en-US-DavisNeural",
         audioContext: input.audioContext,
         azureCredentials: input.azureCredentials,
+        buffer: "",
     }),
     initial: "GetToken",
     on: {
@@ -33,9 +127,197 @@ exports.ttsMachine = (0, xstate_1.createMachine)({
             states: {
                 Idle: {
                     on: {
-                        SPEAK: {
-                            target: "Speaking",
-                            actions: (0, xstate_1.assign)({ agenda: ({ event }) => event.value }),
+                        SPEAK: [
+                            {
+                                target: "BufferedSpeaker",
+                                guard: ({ event }) => !!event.value.stream,
+                                actions: (0, xstate_1.assign)({
+                                    agenda: ({ event }) => event.value.fillerDelay
+                                        ? event.value
+                                        : Object.assign(Object.assign({}, event.value), { fillerDelay: 500 }),
+                                }),
+                            },
+                            {
+                                target: "Speaking",
+                                actions: (0, xstate_1.assign)({ agenda: ({ event }) => event.value }),
+                            },
+                        ],
+                    },
+                },
+                BufferedSpeaker: {
+                    type: "parallel",
+                    invoke: {
+                        id: "createEventsFromStream",
+                        src: "createEventsFromStream",
+                        input: ({ context }) => context.agenda,
+                    },
+                    on: {
+                        STOP: {
+                            target: "Idle",
+                        },
+                        SPEAK_COMPLETE: [
+                            {
+                                guard: (0, xstate_1.stateIn)("#BufferingDone"),
+                                target: "Idle",
+                                actions: [(0, xstate_1.sendParent)({ type: "SPEAK_COMPLETE" })],
+                            },
+                        ],
+                    },
+                    states: {
+                        Buffer: {
+                            initial: "BufferIdle",
+                            states: {
+                                BufferIdle: {
+                                    id: "BufferIdle",
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry BufferIdle", event),
+                                    ],
+                                    on: {
+                                        STREAMING_CHUNK: {
+                                            actions: [
+                                                ({ event }) => console.debug("=================STREAMING_CHUNK: BufferIdle => Buffering", event),
+                                            ],
+                                            target: "Buffering",
+                                        },
+                                    },
+                                },
+                                Buffering: {
+                                    id: "Buffering",
+                                    on: {
+                                        STREAMING_CHUNK: [
+                                            {
+                                                actions: [
+                                                    ({ event }) => console.debug("=================STREAMING_CHUNK: Buffering => Buffering", event),
+                                                ],
+                                                target: "Buffering",
+                                                reenter: true,
+                                            },
+                                        ],
+                                        STREAMING_DONE: [
+                                            {
+                                                target: "BufferingDone",
+                                                actions: [
+                                                    ({ event }) => console.debug("=================STREAMING_DONE: Buffering => BufferingDone", event),
+                                                ],
+                                            },
+                                        ],
+                                    },
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry Buffering", event),
+                                        (0, xstate_1.assign)({
+                                            buffer: ({ context, event }) => context.buffer + event.value,
+                                        }),
+                                    ],
+                                },
+                                BufferingDone: {
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry BufferingDone", event),
+                                    ],
+                                    id: "BufferingDone",
+                                },
+                            },
+                        },
+                        Speaker: {
+                            initial: "SpeakingIdle",
+                            states: {
+                                SpeakingIdle: {
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry SpeakingIdle", event),
+                                    ],
+                                    always: [
+                                        {
+                                            target: "Speak",
+                                            guard: (0, xstate_1.stateIn)("#BufferingDone"),
+                                            actions: [
+                                                ({ event }) => console.debug("========== in BufferingDone: SpeakingIdle => Speak", event),
+                                                (0, xstate_1.assign)({
+                                                    utteranceFromStream: ({ context }) => context.buffer,
+                                                }),
+                                                (0, xstate_1.assign)({
+                                                    buffer: "",
+                                                }),
+                                            ],
+                                        },
+                                        {
+                                            target: "PrepareSpeech",
+                                            guard: "bufferContainsUtterancePartReadyToBeSpoken",
+                                        },
+                                    ],
+                                    after: {
+                                        FILLER_DELAY: {
+                                            target: "SpeakingIdle",
+                                            reenter: true,
+                                            actions: "addFiller",
+                                            guard: ({ context }) => context.buffer.includes(" "),
+                                        },
+                                    },
+                                },
+                                PrepareSpeech: {
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry PrepareSpeech", event),
+                                        (0, xstate_1.assign)(({ context }) => {
+                                            let utterancePart;
+                                            let restOfBuffer;
+                                            const match = context.buffer.match(UTTERANCE_CHUNK_REGEX);
+                                            utterancePart = match[0];
+                                            restOfBuffer = context.buffer.substring(utterancePart.length);
+                                            return {
+                                                buffer: restOfBuffer,
+                                                utteranceFromStream: utterancePart,
+                                            };
+                                        }),
+                                    ],
+                                    always: [
+                                        {
+                                            target: "Speak",
+                                        },
+                                    ],
+                                },
+                                Speak: {
+                                    entry: [
+                                        ({ event }) => console.debug("=== Entry Speak", event),
+                                    ],
+                                    initial: "Go",
+                                    on: {
+                                        TTS_STARTED: {
+                                            actions: (0, xstate_1.sendParent)({ type: "TTS_STARTED" }),
+                                        },
+                                        SPEAK_COMPLETE: [
+                                            {
+                                                guard: (0, xstate_1.stateIn)("#Buffering"),
+                                                target: "SpeakingIdle",
+                                            },
+                                            {
+                                                guard: "bufferIsNonEmpty",
+                                                target: "SpeakingIdle",
+                                            },
+                                        ],
+                                    },
+                                    states: {
+                                        Go: {
+                                            invoke: {
+                                                src: "start",
+                                                input: ({ context }) => ({
+                                                    wsaTTS: context.wsaTTS,
+                                                    wsaUtt: context.wsaUtt,
+                                                    ttsLexicon: context.ttsLexicon,
+                                                    voice: context.agenda.voice || context.ttsDefaultVoice,
+                                                    utterance: context.utteranceFromStream,
+                                                }),
+                                            },
+                                            on: {
+                                                CONTROL: "Paused",
+                                            },
+                                            exit: "ttsStop",
+                                        },
+                                        Paused: {
+                                            on: {
+                                                CONTROL: "Go",
+                                            },
+                                        },
+                                    },
+                                },
+                            },
                         },
                     },
                 },
@@ -113,61 +395,6 @@ exports.ttsMachine = (0, xstate_1.createMachine)({
             },
         },
     },
-}, {
-    actions: {
-        ttsStop: ({ context }) => {
-            context.wsaTTS.cancel();
-        },
-    },
-    actors: {
-        getToken: getToken_1.getToken,
-        ponyfill: (0, xstate_1.fromCallback)(({ sendBack, input }) => {
-            const ponyfill = (0, TextToSpeech_1.default)({
-                audioContext: input.audioContext,
-                credentials: {
-                    region: REGION, // TODO
-                    authorizationToken: input.azureAuthorizationToken,
-                },
-            });
-            const { speechSynthesis, SpeechSynthesisUtterance } = ponyfill;
-            const tts = speechSynthesis;
-            const ttsUtterance = SpeechSynthesisUtterance;
-            tts.addEventListener("voiceschanged", () => {
-                const voices = tts.getVoices();
-                if (voices.length > 0) {
-                    console.debug("[TTS] READY");
-                    sendBack({
-                        type: "READY",
-                        value: { wsaTTS: tts, wsaUtt: ttsUtterance },
-                    });
-                }
-                else {
-                    console.error("[TTS] No voices available");
-                    sendBack({ type: "ERROR" });
-                }
-            });
-        }),
-        start: (0, xstate_1.fromCallback)(({ sendBack, input }) => {
-            if (["", " "].includes(input.utterance)) {
-                console.debug("[TTS] SPEAK: ", input.utterance);
-                input.wsaTTS.speak("");
-            }
-            else {
-                console.debug("[TTS] SPEAK: ", input.utterance);
-                const content = wrapSSML(input.utterance, input.voice, input.ttsLexicon, 1); // todo speech rate;
-                const utterance = new input.wsaUtt(content);
-                utterance.addEventListener("start", () => {
-                    sendBack({ type: "TTS_STARTED" });
-                    console.debug("[TTS] TTS_STARTED");
-                });
-                utterance.addEventListener("end", () => {
-                    sendBack({ type: "SPEAK_COMPLETE" });
-                    console.debug("[TTS] SPEAK_COMPLETE");
-                });
-                input.wsaTTS.speak(utterance);
-            }
-        }),
-    },
 });
 const wrapSSML = (text, voice, lexicon, speechRate) => {
     let content = `<speak version="1.0" xmlns="http://www.w3.org/2001/10/synthesis" xmlns:mstts="http://www.w3.org/2001/mstts" xml:lang="en-US"><voice name="${voice}">`;

package/dist/types.d.ts

@@ -25,7 +25,8 @@ export interface Settings {
 export interface Agenda {
     utterance: string;
     voice?: string;
-    stream?: ReadableStream<string>;
+    stream?: string;
+    fillerDelay?: number;
 }
 export interface Hypothesis {
     utterance: string;

package/dist/types.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,MAAM,WAAW,sBAAsB;IACrC,QAAQ,EAAE,MAAM,CAAC;IACjB,GAAG,EAAE,MAAM,CAAC;CACb;AAED;;GAEG;AACH,MAAM,WAAW,gBAAiB,SAAQ,sBAAsB;CAAG;AAEnE,MAAM,WAAW,wBAAwB;IACvC,QAAQ,EAAE,MAAM,CAAC;IACjB,GAAG,EAAE,MAAM,CAAC;IACZ,WAAW,EAAE,MAAM,CAAC;IACpB,cAAc,EAAE,MAAM,CAAC;CACxB;AAED,MAAM,WAAW,QAAQ;IACvB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,gBAAgB,EAAE,MAAM,GAAG,gBAAgB,GAAG,sBAAsB,CAAC;IACrE,wBAAwB,CAAC,EAAE,wBAAwB,CAAC;IACpD,yBAAyB,CAAC,EAAE,MAAM,CAAC;IACnC,wBAAwB,CAAC,EAAE,MAAM,CAAC;IAClC,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,2BAA2B,CAAC,EAAE,MAAM,CAAC;CACtC;AAED,MAAM,WAAW,MAAM;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,cAAc,CAAC,MAAM,CAAC,CAAC;CACjC;AAED,MAAM,WAAW,UAAU;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,mBAAmB;IAClC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC;IACjB,GAAG,CAAC,EAAE,OAAO,GAAG,wBAAwB,CAAC;CAC1C"}
+{"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA,MAAM,WAAW,sBAAsB;IACrC,QAAQ,EAAE,MAAM,CAAC;IACjB,GAAG,EAAE,MAAM,CAAC;CACb;AAED;;GAEG;AACH,MAAM,WAAW,gBAAiB,SAAQ,sBAAsB;CAAG;AAEnE,MAAM,WAAW,wBAAwB;IACvC,QAAQ,EAAE,MAAM,CAAC;IACjB,GAAG,EAAE,MAAM,CAAC;IACZ,WAAW,EAAE,MAAM,CAAC;IACpB,cAAc,EAAE,MAAM,CAAC;CACxB;AAED,MAAM,WAAW,QAAQ;IACvB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,gBAAgB,EAAE,MAAM,GAAG,gBAAgB,GAAG,sBAAsB,CAAC;IACrE,wBAAwB,CAAC,EAAE,wBAAwB,CAAC;IACpD,yBAAyB,CAAC,EAAE,MAAM,CAAC;IACnC,wBAAwB,CAAC,EAAE,MAAM,CAAC;IAClC,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,2BAA2B,CAAC,EAAE,MAAM,CAAC;CACtC;AAED,MAAM,WAAW,MAAM;IACrB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;AAED,MAAM,WAAW,UAAU;IACzB,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;CACpB;AAED,MAAM,WAAW,mBAAmB;IAClC,cAAc,CAAC,EAAE,MAAM,CAAC;IACxB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,KAAK,CAAC,EAAE,MAAM,EAAE,CAAC;IACjB,GAAG,CAAC,EAAE,OAAO,GAAG,wBAAwB,CAAC;CAC1C"}

package/package.json

@@ -1,11 +1,13 @@
 {
   "name": "speechstate",
-  "version": "2.0.0-beta.9",
+  "version": "2.0.0",
   "license": "GPL-3.0",
   "homepage": "http://localhost/speechstate",
   "main": "./dist/index.js",
   "types": "./dist/index.d.ts",
-  "files": ["/dist"],
+  "files": [
+    "/dist"
+  ],
   "dependencies": {
     "@statelyai/inspect": "^0.2.3",
     "microsoft-cognitiveservices-speech-sdk": "^1.31.0",
@@ -18,10 +20,17 @@
     "test": "jest --silent=false"
   },
   "eslintConfig": {
-    "extends": ["react-app", "react-app/jest"]
+    "extends": [
+      "react-app",
+      "react-app/jest"
+    ]
   },
   "browserslist": {
-    "production": [">0.2%", "not dead", "not op_mini all"],
+    "production": [
+      ">0.2%",
+      "not dead",
+      "not op_mini all"
+    ],
     "development": [
       "last 1 chrome version",
       "last 1 firefox version",

package/README.plantuml

@@ -1,3 +0,0 @@
-@startuml
-Bob -> Alice : Hello World!
-@enduml