@ai-ecoverse/pi-audio 1.15.0 → 1.16.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +2 -0
- package/dist/index.js +2 -0
- package/dist/index.js.map +2 -2
- package/dist/soundscape.js +104 -0
- package/dist/soundscape.js.map +7 -0
- package/dist/turns.js +28 -4
- package/dist/turns.js.map +2 -2
- package/dist/types/index.d.ts +2 -0
- package/dist/types/soundscape.d.ts +47 -0
- package/dist/types/turns.d.ts +3 -1
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -27,6 +27,8 @@ pi-audio does not apply that patch. Hosts such as slicc-agent record monkeypatch
|
|
|
27
27
|
|
|
28
28
|
`@ai-ecoverse/pi-audio/elevenlabs` registers ElevenLabs speech adapters when `registerElevenLabs()` is called. `scribe_v2` posts multipart audio to `/speech-to-text`. `scribe_v2_realtime` streams base64 audio on `wss://…/speech-to-text/realtime` and commits turns with `turn_prediction` (`turnDetection: 'semantic'`). `eleven_flash_v2_5` and `eleven_multilingual_v2` post JSON to `/text-to-speech/{voice}/with-timestamps` for a string and stream text on `stream-input` for an async iterable, with character alignment. `eleven_v3` and `eleven_v4` stay on the HTTP path (`textStreaming: false`). Scribe prices are $0.22 per hour for batch and $0.39 per hour for realtime.
|
|
29
29
|
|
|
30
|
+
`followBed(bed, soundscape)` cues a `BedChannel` from `VoicePhase`. A loop waits `graceMs` before it starts, so a short gap stays silent. `keep` leaves the current loop running across chained tools. Speech ducks the bed to `duckUnderSpeech`; listening ducks it to `duckUnderUser`. `voiceTurns` does this when `options.soundscape` is set and `out` is an `AudioOut` (`speech`, `bed`, `reference`, `playback`). A plain `AudioSink` still receives only the spoken audio.
|
|
31
|
+
|
|
30
32
|
`voiceTurns` captions include `agent_spoken`. `upTo` follows `AudioSink` playback: per character when the chunk has alignment, and evenly across the sentence when it does not. `captionsFromHistory(entries)` rebuilds user captions from entries whose `data.source` is `voice`, and agent captions from the assistant text. A `voice.truncate` entry adds `agent_spoken` and `agent_cut` at `heardUpTo`, so the rest of the original text stays available as unheard.
|
|
31
33
|
`@ai-ecoverse/pi-audio/xai` registers xAI speech adapters when `registerXAI()` is called. `grok-voice-transcribe-2.0` posts multipart audio to `/stt` (fields before `file`) and streams raw binary frames on `wss://…/stt` after `transcript.created`, with Smart Turn (`turnDetection: 'semantic'`). The catalog id is shared, so `audioCost` uses the streaming rate of $0.20 per hour. `grok-voice-tts-1.0` posts JSON to `/tts` for a string and streams `text.delta` / `audio.delta` on `wss://…/tts` for an async iterable (`textStreaming: true`). Voices are eve, ara, rex, sal, and leo. Codecs are mp3, pcm, wav, mulaw, and alaw.
|
|
32
34
|
|
package/dist/index.js
CHANGED
|
@@ -23,6 +23,7 @@ import {
|
|
|
23
23
|
speak,
|
|
24
24
|
transcribe
|
|
25
25
|
} from "./registry.js";
|
|
26
|
+
import { followBed } from "./soundscape.js";
|
|
26
27
|
import { SpeechEventStream } from "./speech.js";
|
|
27
28
|
import { askAgent, bindThinker, talkerTools } from "./talker.js";
|
|
28
29
|
import { mintRealtimeToken } from "./tokens.js";
|
|
@@ -46,6 +47,7 @@ export {
|
|
|
46
47
|
decodeAudioBase64,
|
|
47
48
|
encodeAudioBase64,
|
|
48
49
|
energyVad,
|
|
50
|
+
followBed,
|
|
49
51
|
getVoiceModel,
|
|
50
52
|
listVoiceModels,
|
|
51
53
|
listen,
|
package/dist/index.js.map
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"version": 3,
|
|
3
3
|
"sources": ["../src/index.ts"],
|
|
4
|
-
"sourcesContent": ["export type {\n AudioChunk,\n AudioEncoding,\n AudioFormat,\n AudioSink,\n AudioSource,\n AudioTag,\n CharAlignment,\n TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport type { CaptionHistoryEntry } from './captions.ts';\nexport { captionsFromHistory } from './captions.ts';\nexport {\n convertAudio,\n convertAudioSource,\n decodeAudioBase64,\n encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n Delivery,\n DeliverySupport,\n KnownRealtimeApi,\n KnownSpeechApi,\n KnownTranscriptionApi,\n RealtimeApi,\n RealtimeModel,\n SpeechApi,\n SpeechModel,\n TranscriptionApi,\n TranscriptionModel,\n VoiceInfo,\n VoiceModel,\n VoiceModelType,\n} from './models.ts';\nexport {\n getVoiceModel,\n listVoiceModels,\n registerVoiceModel,\n} from './models.ts';\nexport type {\n DeliveryPlan,\n RealtimeApiProvider,\n RealtimeContext,\n RealtimeEvent,\n RealtimeOptions,\n RealtimeSession,\n RealtimeSocket,\n RealtimeToken,\n TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n connectRealtime,\n listen,\n registerRealtimeApiProvider,\n registerSpeechApiProvider,\n registerTranscriptionApiProvider,\n speak,\n transcribe,\n} from './registry.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';\nexport type {\n BindThinkerOptions,\n BoundThinker,\n Thinker,\n ThinkerCaption,\n ThinkerPhase,\n ThinkerUpdate,\n} from './talker.ts';\nexport { askAgent, bindThinker, talkerTools } from './talker.ts';\nexport type { RealtimeConstraints } from './tokens.ts';\nexport { mintRealtimeToken } from './tokens.ts';\nexport type {\n AudioInput,\n ListenOptions,\n Transcript,\n TranscriptionApiProvider,\n TranscriptionEvent,\n TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n BargeIn,\n CaptionEvent,\n EchoRecommendation,\n VoiceConversation,\n VoiceCut,\n VoiceModels,\n VoicePhase,\n VoiceSession,\n VoiceTurnMeta,\n VoiceTurnOptions,\n} from './turns.ts';\nexport { voiceTurns } from './turns.ts';\nexport type {\n AudioCost,\n AudioCounters,\n AudioUsage,\n RealtimeAudioCost,\n SpeechAudioCost,\n TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
|
|
5
|
-
"mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAE3B,SAAS,2BAA2B;AACpC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,yBAAyB;AAUlC,SAAS,UAAU,aAAa,mBAAmB;AAEnD,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AAazC,SAAS,kBAAkB;AAS3B,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
|
|
4
|
+
"sourcesContent": ["export type {\n AudioChunk,\n AudioEncoding,\n AudioFormat,\n AudioSink,\n AudioSource,\n AudioTag,\n CharAlignment,\n TextSpan,\n} from './audio.ts';\nexport { resolveVoiceAuth } from './auth.ts';\nexport { blankUsage } from './blank-usage.ts';\nexport type { CaptionHistoryEntry } from './captions.ts';\nexport { captionsFromHistory } from './captions.ts';\nexport {\n convertAudio,\n convertAudioSource,\n decodeAudioBase64,\n encodeAudioBase64,\n} from './convert.ts';\nexport { VoiceRequestError } from './errors.ts';\nexport type {\n Delivery,\n DeliverySupport,\n KnownRealtimeApi,\n KnownSpeechApi,\n KnownTranscriptionApi,\n RealtimeApi,\n RealtimeModel,\n SpeechApi,\n SpeechModel,\n TranscriptionApi,\n TranscriptionModel,\n VoiceInfo,\n VoiceModel,\n VoiceModelType,\n} from './models.ts';\nexport {\n getVoiceModel,\n listVoiceModels,\n registerVoiceModel,\n} from './models.ts';\nexport type {\n DeliveryPlan,\n RealtimeApiProvider,\n RealtimeContext,\n RealtimeEvent,\n RealtimeOptions,\n RealtimeSession,\n RealtimeSocket,\n RealtimeToken,\n TruncationPlan,\n} from './realtime.ts';\nexport { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';\nexport {\n connectRealtime,\n listen,\n registerRealtimeApiProvider,\n registerSpeechApiProvider,\n registerTranscriptionApiProvider,\n speak,\n transcribe,\n} from './registry.ts';\nexport type { AudioOut, BedChannel, Cue, CueSource, Soundscape } from './soundscape.ts';\nexport { followBed } from './soundscape.ts';\nexport type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';\nexport { SpeechEventStream } from './speech.ts';\nexport type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';\nexport type {\n BindThinkerOptions,\n BoundThinker,\n Thinker,\n ThinkerCaption,\n ThinkerPhase,\n ThinkerUpdate,\n} from './talker.ts';\nexport { askAgent, bindThinker, talkerTools } from './talker.ts';\nexport type { RealtimeConstraints } from './tokens.ts';\nexport { mintRealtimeToken } from './tokens.ts';\nexport type {\n AudioInput,\n ListenOptions,\n Transcript,\n TranscriptionApiProvider,\n TranscriptionEvent,\n TranscriptSegment,\n} from './transcription.ts';\nexport { TranscriptionEventStream } from './transcription.ts';\nexport type {\n BargeIn,\n CaptionEvent,\n EchoRecommendation,\n VoiceConversation,\n VoiceCut,\n VoiceModels,\n VoicePhase,\n VoiceSession,\n VoiceTurnMeta,\n VoiceTurnOptions,\n} from './turns.ts';\nexport { voiceTurns } from './turns.ts';\nexport type {\n AudioCost,\n AudioCounters,\n AudioUsage,\n RealtimeAudioCost,\n SpeechAudioCost,\n TranscriptionAudioCost,\n} from './usage.ts';\nexport { addAudioCounters, priceAudio } from './usage.ts';\nexport type { VadHook, VadUpdate } from './vad.ts';\nexport { energyVad } from './vad.ts';\n"],
|
|
5
|
+
"mappings": "AAUA,SAAS,wBAAwB;AACjC,SAAS,kBAAkB;AAE3B,SAAS,2BAA2B;AACpC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AACP,SAAS,yBAAyB;AAiBlC;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAYP,SAAS,cAAc,gBAAgB,gBAAgB,sBAAsB;AAC7E;AAAA,EACE;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,OACK;AAEP,SAAS,iBAAiB;AAE1B,SAAS,yBAAyB;AAUlC,SAAS,UAAU,aAAa,mBAAmB;AAEnD,SAAS,yBAAyB;AASlC,SAAS,gCAAgC;AAazC,SAAS,kBAAkB;AAS3B,SAAS,kBAAkB,kBAAkB;AAE7C,SAAS,iBAAiB;",
|
|
6
6
|
"names": []
|
|
7
7
|
}
|
|
@@ -0,0 +1,104 @@
|
|
|
1
|
+
function isAudioOut(out) {
|
|
2
|
+
return "speech" in out && "bed" in out;
|
|
3
|
+
}
|
|
4
|
+
function followBed(bed, soundscape, wait) {
|
|
5
|
+
const schedule = wait === void 0 ? (run, delayMs) => setTimeout(run, delayMs) : wait;
|
|
6
|
+
let previous = { phase: "idle" };
|
|
7
|
+
let timer;
|
|
8
|
+
let generation = 0;
|
|
9
|
+
let playing;
|
|
10
|
+
const clearTimer = () => {
|
|
11
|
+
if (timer !== void 0) clearTimeout(timer);
|
|
12
|
+
timer = void 0;
|
|
13
|
+
};
|
|
14
|
+
const stop = () => {
|
|
15
|
+
generation += 1;
|
|
16
|
+
clearTimer();
|
|
17
|
+
};
|
|
18
|
+
const observe = (phase) => {
|
|
19
|
+
const cues = listCues(soundscape.cue(phase, previous));
|
|
20
|
+
previous = phase;
|
|
21
|
+
const fade = fadeOf(soundscape);
|
|
22
|
+
bed.duck(duckFor(phase, soundscape), fade);
|
|
23
|
+
for (const cue of cues) {
|
|
24
|
+
if (cue.kind === "oneShot") bed.play(cue, fade);
|
|
25
|
+
}
|
|
26
|
+
if (cues.some((cue) => cue.kind === "silence")) {
|
|
27
|
+
generation += 1;
|
|
28
|
+
clearTimer();
|
|
29
|
+
playing = void 0;
|
|
30
|
+
bed.stop(fade);
|
|
31
|
+
return;
|
|
32
|
+
}
|
|
33
|
+
const loop = firstLoop(cues);
|
|
34
|
+
if (!loop) {
|
|
35
|
+
generation += 1;
|
|
36
|
+
clearTimer();
|
|
37
|
+
return;
|
|
38
|
+
}
|
|
39
|
+
const key = sourceKey(loop.source);
|
|
40
|
+
if (playing === key) {
|
|
41
|
+
generation += 1;
|
|
42
|
+
clearTimer();
|
|
43
|
+
return;
|
|
44
|
+
}
|
|
45
|
+
generation += 1;
|
|
46
|
+
const token = generation;
|
|
47
|
+
clearTimer();
|
|
48
|
+
const start = () => {
|
|
49
|
+
if (token !== generation) return;
|
|
50
|
+
playing = key;
|
|
51
|
+
bed.play(loop, fade);
|
|
52
|
+
};
|
|
53
|
+
const grace = graceOf(soundscape);
|
|
54
|
+
if (grace <= 0) {
|
|
55
|
+
start();
|
|
56
|
+
return;
|
|
57
|
+
}
|
|
58
|
+
timer = schedule(start, grace);
|
|
59
|
+
};
|
|
60
|
+
return { observe, stop };
|
|
61
|
+
}
|
|
62
|
+
function listCues(cue) {
|
|
63
|
+
if ("kind" in cue) return [cue];
|
|
64
|
+
return cue;
|
|
65
|
+
}
|
|
66
|
+
function firstLoop(cues) {
|
|
67
|
+
for (const cue of cues) {
|
|
68
|
+
if (cue.kind === "loop") return cue;
|
|
69
|
+
}
|
|
70
|
+
return void 0;
|
|
71
|
+
}
|
|
72
|
+
function sourceKey(source) {
|
|
73
|
+
if ("url" in source) return `url:${source.url}`;
|
|
74
|
+
if ("buffer" in source) return `buffer:${source.buffer.length}`;
|
|
75
|
+
return "synth";
|
|
76
|
+
}
|
|
77
|
+
function fadeOf(soundscape) {
|
|
78
|
+
const fade = soundscape.fadeMs;
|
|
79
|
+
if (fade === void 0) return 0;
|
|
80
|
+
return fade;
|
|
81
|
+
}
|
|
82
|
+
function graceOf(soundscape) {
|
|
83
|
+
const grace = soundscape.graceMs;
|
|
84
|
+
if (grace === void 0) return 0;
|
|
85
|
+
return grace;
|
|
86
|
+
}
|
|
87
|
+
function duckFor(phase, soundscape) {
|
|
88
|
+
if (phase.phase === "speaking") {
|
|
89
|
+
const gain = soundscape.duckUnderSpeech;
|
|
90
|
+
if (gain === void 0) return 1;
|
|
91
|
+
return gain;
|
|
92
|
+
}
|
|
93
|
+
if (phase.phase === "listening") {
|
|
94
|
+
const gain = soundscape.duckUnderUser;
|
|
95
|
+
if (gain === void 0) return 1;
|
|
96
|
+
return gain;
|
|
97
|
+
}
|
|
98
|
+
return 1;
|
|
99
|
+
}
|
|
100
|
+
export {
|
|
101
|
+
followBed,
|
|
102
|
+
isAudioOut
|
|
103
|
+
};
|
|
104
|
+
//# sourceMappingURL=soundscape.js.map
|
|
@@ -0,0 +1,7 @@
|
|
|
1
|
+
{
|
|
2
|
+
"version": 3,
|
|
3
|
+
"sources": ["../src/soundscape.ts"],
|
|
4
|
+
"sourcesContent": ["import type { AudioSink, AudioSource } from './audio.ts';\nimport type { VoicePhase } from './turns.ts';\n\nexport type CueSource =\n | { url: string }\n | { buffer: AudioBuffer }\n | { synth: (ctx: BaseAudioContext) => AudioNode };\n\nexport type Cue =\n | { kind: 'loop'; source: CueSource; gain: number }\n | { kind: 'oneShot'; source: CueSource; gain: number }\n | { kind: 'keep' }\n | { kind: 'silence' };\n\nexport interface Soundscape {\n cue(next: VoicePhase, previous: VoicePhase): Cue | readonly Cue[];\n graceMs?: number;\n fadeMs?: number;\n duckUnderSpeech?: number;\n duckUnderUser?: number;\n}\n\nexport interface BedChannel {\n play(cue: Cue, fadeMs: number): void;\n duck(gain: number, rampMs: number): void;\n stop(fadeMs: number): void;\n}\n\nexport interface AudioOut {\n readonly context: BaseAudioContext;\n speech: AudioSink;\n bed: BedChannel;\n readonly reference: AudioSource;\n readonly playback: 'webaudio' | 'rtc-loopback';\n}\n\nexport interface BedFollower {\n observe(phase: VoicePhase): void;\n stop(): void;\n}\n\nexport function isAudioOut(out: AudioSink | AudioOut): out is AudioOut {\n return 'speech' in out && 'bed' in out;\n}\n\nexport function followBed(\n bed: BedChannel,\n soundscape: Soundscape,\n wait?: (run: () => void, delayMs: number) => ReturnType<typeof setTimeout>\n): BedFollower {\n const schedule =\n wait === undefined ? (run: () => void, delayMs: number) => setTimeout(run, delayMs) : wait;\n let previous: VoicePhase = { phase: 'idle' };\n let timer: ReturnType<typeof setTimeout> | undefined;\n let generation = 0;\n let playing: string | undefined;\n\n const clearTimer = () => {\n if (timer !== undefined) clearTimeout(timer);\n timer = undefined;\n };\n\n const stop = () => {\n generation += 1;\n clearTimer();\n };\n\n const observe = (phase: VoicePhase) => {\n const cues = listCues(soundscape.cue(phase, previous));\n previous = phase;\n const fade = fadeOf(soundscape);\n bed.duck(duckFor(phase, soundscape), fade);\n for (const cue of cues) {\n if (cue.kind === 'oneShot') bed.play(cue, fade);\n }\n if (cues.some((cue) => cue.kind === 'silence')) {\n generation += 1;\n clearTimer();\n playing = undefined;\n bed.stop(fade);\n return;\n }\n const loop = firstLoop(cues);\n if (!loop) {\n generation += 1;\n clearTimer();\n return;\n }\n const key = sourceKey(loop.source);\n if (playing === key) {\n generation += 1;\n clearTimer();\n return;\n }\n generation += 1;\n const token = generation;\n clearTimer();\n const start = () => {\n if (token !== generation) return;\n playing = key;\n bed.play(loop, fade);\n };\n const grace = graceOf(soundscape);\n if (grace <= 0) {\n start();\n return;\n }\n timer = schedule(start, grace);\n };\n\n return { observe, stop };\n}\n\nfunction listCues(cue: Cue | readonly Cue[]): readonly Cue[] {\n if ('kind' in cue) return [cue];\n return cue;\n}\n\nfunction firstLoop(cues: readonly Cue[]): Extract<Cue, { kind: 'loop' }> | undefined {\n for (const cue of cues) {\n if (cue.kind === 'loop') return cue;\n }\n return undefined;\n}\n\nfunction sourceKey(source: CueSource): string {\n if ('url' in source) return `url:${source.url}`;\n if ('buffer' in source) return `buffer:${source.buffer.length}`;\n return 'synth';\n}\n\nfunction fadeOf(soundscape: Soundscape): number {\n const fade = soundscape.fadeMs;\n if (fade === undefined) return 0;\n return fade;\n}\n\nfunction graceOf(soundscape: Soundscape): number {\n const grace = soundscape.graceMs;\n if (grace === undefined) return 0;\n return grace;\n}\n\nfunction duckFor(phase: VoicePhase, soundscape: Soundscape): number {\n if (phase.phase === 'speaking') {\n const gain = soundscape.duckUnderSpeech;\n if (gain === undefined) return 1;\n return gain;\n }\n if (phase.phase === 'listening') {\n const gain = soundscape.duckUnderUser;\n if (gain === undefined) return 1;\n return gain;\n }\n return 1;\n}\n"],
|
|
5
|
+
"mappings": "AAyCO,SAAS,WAAW,KAA4C;AACrE,SAAO,YAAY,OAAO,SAAS;AACrC;AAEO,SAAS,UACd,KACA,YACA,MACa;AACb,QAAM,WACJ,SAAS,SAAY,CAAC,KAAiB,YAAoB,WAAW,KAAK,OAAO,IAAI;AACxF,MAAI,WAAuB,EAAE,OAAO,OAAO;AAC3C,MAAI;AACJ,MAAI,aAAa;AACjB,MAAI;AAEJ,QAAM,aAAa,MAAM;AACvB,QAAI,UAAU,OAAW,cAAa,KAAK;AAC3C,YAAQ;AAAA,EACV;AAEA,QAAM,OAAO,MAAM;AACjB,kBAAc;AACd,eAAW;AAAA,EACb;AAEA,QAAM,UAAU,CAAC,UAAsB;AACrC,UAAM,OAAO,SAAS,WAAW,IAAI,OAAO,QAAQ,CAAC;AACrD,eAAW;AACX,UAAM,OAAO,OAAO,UAAU;AAC9B,QAAI,KAAK,QAAQ,OAAO,UAAU,GAAG,IAAI;AACzC,eAAW,OAAO,MAAM;AACtB,UAAI,IAAI,SAAS,UAAW,KAAI,KAAK,KAAK,IAAI;AAAA,IAChD;AACA,QAAI,KAAK,KAAK,CAAC,QAAQ,IAAI,SAAS,SAAS,GAAG;AAC9C,oBAAc;AACd,iBAAW;AACX,gBAAU;AACV,UAAI,KAAK,IAAI;AACb;AAAA,IACF;AACA,UAAM,OAAO,UAAU,IAAI;AAC3B,QAAI,CAAC,MAAM;AACT,oBAAc;AACd,iBAAW;AACX;AAAA,IACF;AACA,UAAM,MAAM,UAAU,KAAK,MAAM;AACjC,QAAI,YAAY,KAAK;AACnB,oBAAc;AACd,iBAAW;AACX;AAAA,IACF;AACA,kBAAc;AACd,UAAM,QAAQ;AACd,eAAW;AACX,UAAM,QAAQ,MAAM;AAClB,UAAI,UAAU,WAAY;AAC1B,gBAAU;AACV,UAAI,KAAK,MAAM,IAAI;AAAA,IACrB;AACA,UAAM,QAAQ,QAAQ,UAAU;AAChC,QAAI,SAAS,GAAG;AACd,YAAM;AACN;AAAA,IACF;AACA,YAAQ,SAAS,OAAO,KAAK;AAAA,EAC/B;AAEA,SAAO,EAAE,SAAS,KAAK;AACzB;AAEA,SAAS,SAAS,KAA2C;AAC3D,MAAI,UAAU,IAAK,QAAO,CAAC,GAAG;AAC9B,SAAO;AACT;AAEA,SAAS,UAAU,MAAkE;AACnF,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,OAAQ,QAAO;AAAA,EAClC;AACA,SAAO;AACT;AAEA,SAAS,UAAU,QAA2B;AAC5C,MAAI,SAAS,OAAQ,QAAO,OAAO,OAAO,GAAG;AAC7C,MAAI,YAAY,OAAQ,QAAO,UAAU,OAAO,OAAO,MAAM;AAC7D,SAAO;AACT;AAEA,SAAS,OAAO,YAAgC;AAC9C,QAAM,OAAO,WAAW;AACxB,MAAI,SAAS,OAAW,QAAO;AAC/B,SAAO;AACT;AAEA,SAAS,QAAQ,YAAgC;AAC/C,QAAM,QAAQ,WAAW;AACzB,MAAI,UAAU,OAAW,QAAO;AAChC,SAAO;AACT;AAEA,SAAS,QAAQ,OAAmB,YAAgC;AAClE,MAAI,MAAM,UAAU,YAAY;AAC9B,UAAM,OAAO,WAAW;AACxB,QAAI,SAAS,OAAW,QAAO;AAC/B,WAAO;AAAA,EACT;AACA,MAAI,MAAM,UAAU,aAAa;AAC/B,UAAM,OAAO,WAAW;AACxB,QAAI,SAAS,OAAW,QAAO;AAC/B,WAAO;AAAA,EACT;AACA,SAAO;AACT;",
|
|
6
|
+
"names": []
|
|
7
|
+
}
|
package/dist/turns.js
CHANGED
|
@@ -4,6 +4,10 @@ import {
|
|
|
4
4
|
import { spokenOffset, utteranceIndex } from "./captions.js";
|
|
5
5
|
import { VoiceRequestError } from "./errors.js";
|
|
6
6
|
import { pushQueue } from "./queue.js";
|
|
7
|
+
import {
|
|
8
|
+
followBed,
|
|
9
|
+
isAudioOut
|
|
10
|
+
} from "./soundscape.js";
|
|
7
11
|
import {
|
|
8
12
|
createSpokenState,
|
|
9
13
|
spokenPieces
|
|
@@ -15,12 +19,13 @@ function voiceTurns(models, conversation, mic, out, options, signal) {
|
|
|
15
19
|
models,
|
|
16
20
|
conversation,
|
|
17
21
|
mic,
|
|
18
|
-
out,
|
|
22
|
+
out: speechOf(out),
|
|
19
23
|
options,
|
|
20
24
|
signal,
|
|
21
25
|
phases,
|
|
22
26
|
captions,
|
|
23
|
-
maps: []
|
|
27
|
+
maps: [],
|
|
28
|
+
bed: attachBed(out, options.soundscape)
|
|
24
29
|
};
|
|
25
30
|
const done = runSession(session).finally(() => {
|
|
26
31
|
phases.end();
|
|
@@ -45,7 +50,7 @@ async function runSession(session) {
|
|
|
45
50
|
if (!active) return;
|
|
46
51
|
notePlayback(session, active, tag, offsetMs);
|
|
47
52
|
});
|
|
48
|
-
const setPhase = phaseSetter(session.phases);
|
|
53
|
+
const setPhase = phaseSetter(session.phases, session.bed);
|
|
49
54
|
try {
|
|
50
55
|
if (session.signal.aborted) return;
|
|
51
56
|
setPhase({ phase: "listening" });
|
|
@@ -74,6 +79,7 @@ async function runSession(session) {
|
|
|
74
79
|
else if (active) await active.done;
|
|
75
80
|
stopPlayback();
|
|
76
81
|
setPhase({ phase: "idle" });
|
|
82
|
+
session.bed.stop();
|
|
77
83
|
}
|
|
78
84
|
}
|
|
79
85
|
function turnMeta(options, confidence) {
|
|
@@ -317,13 +323,31 @@ function shouldBarge(mode, event) {
|
|
|
317
323
|
if (mode === "interrupt") return event.type === "speech_start";
|
|
318
324
|
return event.type === "partial" && event.text.trim().length > 0;
|
|
319
325
|
}
|
|
320
|
-
function
|
|
326
|
+
function speechOf(out) {
|
|
327
|
+
if (isAudioOut(out)) return out.speech;
|
|
328
|
+
return out;
|
|
329
|
+
}
|
|
330
|
+
function attachBed(out, soundscape) {
|
|
331
|
+
if (!soundscape) return idleBed();
|
|
332
|
+
if (!isAudioOut(out)) return idleBed();
|
|
333
|
+
return followBed(out.bed, soundscape);
|
|
334
|
+
}
|
|
335
|
+
function idleBed() {
|
|
336
|
+
return {
|
|
337
|
+
observe() {
|
|
338
|
+
},
|
|
339
|
+
stop() {
|
|
340
|
+
}
|
|
341
|
+
};
|
|
342
|
+
}
|
|
343
|
+
function phaseSetter(phases, bed) {
|
|
321
344
|
let previous = "";
|
|
322
345
|
return (phase) => {
|
|
323
346
|
const key = JSON.stringify(phase);
|
|
324
347
|
if (key === previous) return;
|
|
325
348
|
previous = key;
|
|
326
349
|
phases.push(phase);
|
|
350
|
+
bed.observe(phase);
|
|
327
351
|
};
|
|
328
352
|
}
|
|
329
353
|
function linkSignal(parent) {
|
package/dist/turns.js.map
CHANGED
|
@@ -1,7 +1,7 @@
|
|
|
1
1
|
{
|
|
2
2
|
"version": 3,
|
|
3
3
|
"sources": ["../src/turns.ts"],
|
|
4
|
-
"sourcesContent": ["import type {\n AssistantMessageEvent,\n JsonObject,\n ProviderRequestOptions,\n ToolCall,\n UserMessage,\n} from '@earendil-works/pi-ai';\nimport {\n type AudioSink,\n type AudioSource,\n type AudioTag,\n chunkSeconds,\n type TextSpan,\n} from './audio.ts';\nimport { type SpokenClock, spokenOffset, utteranceIndex } from './captions.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { SpeechModel, TranscriptionModel } from './models.ts';\nimport { pushQueue } from './queue.ts';\nimport type { SpeechEvent, SpeechInput } from './speech.ts';\nimport {\n createSpokenState,\n type SpokenPiece,\n spokenPieces,\n type VoiceAgentEvent,\n} from './spoken.ts';\nimport type { ListenOptions, TranscriptionEvent } from './transcription.ts';\n\nexport interface VoiceTurnMeta {\n source: 'voice';\n transcriptionModel: string;\n confidence?: number;\n}\n\nexport interface VoiceCut {\n entryId: string;\n contentIndex: number;\n heardUpTo: number;\n heardMs: number;\n}\n\nexport interface VoiceConversation {\n prompt(\n message: UserMessage,\n meta: VoiceTurnMeta,\n signal: AbortSignal\n ): AsyncIterable<VoiceAgentEvent>;\n truncate(cut: VoiceCut): Promise<void>;\n}\n\nexport type BargeIn = 'interrupt' | 'onWords' | 'ignore' | 'auto';\n\nexport type EchoRecommendation = 'interrupt' | 'onWords' | 'headphones';\n\nexport interface VoiceTurnOptions {\n ears: TranscriptionModel<string>;\n mouth: SpeechModel<string>;\n voice: string;\n speakable?: (event: AssistantMessageEvent) => string | undefined;\n announce?: (call: ToolCall) => string | undefined;\n bargeIn?: BargeIn;\n echoRecommendation?: EchoRecommendation;\n}\n\nexport interface VoiceModels {\n listen(\n model: TranscriptionModel<string>,\n audio: AudioSource,\n options?: ListenOptions\n ): AsyncIterable<TranscriptionEvent>;\n speak(\n model: SpeechModel<string>,\n input: SpeechInput,\n options?: ProviderRequestOptions\n ): AsyncIterable<SpeechEvent>;\n}\n\nexport type VoicePhase =\n | { phase: 'listening' }\n | { phase: 'heard'; text: string }\n | { phase: 'thinking' }\n | { phase: 'tool'; toolName: string; toolCallId: string; args: JsonObject }\n | { phase: 'retrying'; attempt: number; at: number }\n | { phase: 'compacting' }\n | { phase: 'speaking' }\n | { phase: 'idle' };\n\nexport type CaptionEvent =\n | { type: 'user_partial'; turnId: string; segmentId: string; text: string }\n | { type: 'user_final'; turnId: string; segmentId: string; text: string; confidence?: number }\n | { type: 'agent_generated'; turnId: string; utteranceId: string; text: string }\n | { type: 'agent_spoken'; turnId: string; utteranceId: string; upTo: number }\n | { type: 'agent_cut'; turnId: string; utteranceId: string; heardUpTo: number }\n | { type: 'announce'; turnId: string; text: string }\n | { type: 'thinker_progress'; turnId: string; callId: string; text: string }\n | { type: 'thinker_result'; turnId: string; callId: string; summary: string; isError: boolean };\n\nexport interface VoiceSession {\n phases: AsyncIterable<VoicePhase>;\n captions: AsyncIterable<CaptionEvent>;\n done: Promise<void>;\n}\n\ninterface SpokenMap {\n spokenStart: number;\n spokenEnd: number;\n contentIndex: number;\n start: number;\n end: number;\n entryId?: string;\n kind: SpokenPiece['kind'];\n}\n\ninterface ActiveTurn {\n controller: AbortController;\n unlink: () => void;\n done: Promise<void>;\n turnId: string;\n utteranceId: string;\n spokenUpTo: number;\n cut?: VoiceCut;\n}\n\nexport function voiceTurns(\n models: VoiceModels,\n conversation: VoiceConversation,\n mic: AudioSource,\n out: AudioSink,\n options: VoiceTurnOptions,\n signal: AbortSignal\n): VoiceSession {\n const phases = pushQueue<VoicePhase>();\n const captions = pushQueue<CaptionEvent>();\n const session: Session = {\n models,\n conversation,\n mic,\n out,\n options,\n signal,\n phases,\n captions,\n maps: [],\n };\n const done = runSession(session).finally(() => {\n phases.end();\n captions.end();\n });\n return { phases: phases.iterable, captions: captions.iterable, done };\n}\n\ninterface Session {\n models: VoiceModels;\n conversation: VoiceConversation;\n mic: AudioSource;\n out: AudioSink;\n options: VoiceTurnOptions;\n signal: AbortSignal;\n phases: ReturnType<typeof pushQueue<VoicePhase>>;\n captions: ReturnType<typeof pushQueue<CaptionEvent>>;\n maps: SpokenMap[];\n playing?: SpokenClock;\n}\n\nasync function runSession(session: Session): Promise<void> {\n const mode = bargeMode(session.options);\n let active: ActiveTurn | undefined;\n let confidence: number | undefined;\n let reserved = 0;\n let openId = '';\n const openTurn = () => {\n if (!openId) {\n reserved += 1;\n openId = String(reserved);\n }\n return openId;\n };\n const stopPlayback = session.out.onPlayback((tag, offsetMs) => {\n if (!active) return;\n notePlayback(session, active, tag, offsetMs);\n });\n const setPhase = phaseSetter(session.phases);\n try {\n if (session.signal.aborted) return;\n setPhase({ phase: 'listening' });\n for await (const event of session.models.listen(session.options.ears, session.mic, {\n signal: session.signal,\n })) {\n if (event.type === 'error') {\n if (event.reason === 'aborted' || session.signal.aborted) return;\n throw new VoiceRequestError(event.reason, event.error);\n }\n const captionId =\n event.type === 'partial' || event.type === 'final' || event.type === 'turn_end'\n ? openTurn()\n : openId;\n confidence = noteUser(session, event, captionId, confidence);\n if (shouldBarge(mode, event)) active = await interrupt(session, active);\n if (event.type !== 'turn_end') continue;\n const turnId = openId || openTurn();\n openId = '';\n if (mode === 'ignore') active = await finishTurn(active);\n else active = await interrupt(session, active);\n setPhase({ phase: 'heard', text: event.text });\n const meta = turnMeta(session.options, confidence);\n confidence = undefined;\n active = startTurn(session, setPhase, turnId, event, meta);\n }\n } finally {\n if (session.signal.aborted && active) await interrupt(session, active);\n else if (active) await active.done;\n stopPlayback();\n setPhase({ phase: 'idle' });\n }\n}\n\nfunction turnMeta(options: VoiceTurnOptions, confidence: number | undefined): VoiceTurnMeta {\n return {\n source: 'voice',\n transcriptionModel: options.ears.id,\n ...(confidence === undefined ? {} : { confidence }),\n };\n}\n\nfunction noteUser(\n session: Session,\n event: TranscriptionEvent,\n turnId: string,\n confidence: number | undefined\n): number | undefined {\n if (event.type === 'partial') {\n session.captions.push({\n type: 'user_partial',\n turnId,\n segmentId: event.segmentId,\n text: event.text,\n });\n }\n if (event.type === 'final') {\n session.captions.push({\n type: 'user_final',\n turnId,\n segmentId: event.segmentId,\n text: event.text,\n ...(event.confidence === undefined ? {} : { confidence: event.confidence }),\n });\n return event.confidence ?? confidence;\n }\n if (event.type === 'turn_end') {\n session.captions.push({ type: 'user_final', turnId, segmentId: 'turn', text: event.text });\n }\n return confidence;\n}\n\nasync function finishTurn(active: ActiveTurn | undefined): Promise<undefined> {\n if (!active) return undefined;\n await active.done;\n return undefined;\n}\n\nfunction startTurn(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n turnId: string,\n event: Extract<TranscriptionEvent, { type: 'turn_end' }>,\n meta: VoiceTurnMeta\n): ActiveTurn {\n const linked = linkSignal(session.signal);\n const active: ActiveTurn = {\n controller: linked.controller,\n unlink: linked.unlink,\n turnId,\n utteranceId: `${turnId}:agent`,\n spokenUpTo: 0,\n done: Promise.resolve(),\n };\n active.done = performTurn(session, setPhase, active, event.text, event.at, meta).finally(() => {\n linked.unlink();\n });\n return active;\n}\n\nasync function performTurn(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n text: string,\n at: number,\n meta: VoiceTurnMeta\n): Promise<void> {\n const signal = active.controller.signal;\n try {\n setPhase({ phase: 'thinking' });\n const message: UserMessage = { role: 'user', content: text, timestamp: at };\n const feed = textFeed(signal);\n const maps: SpokenMap[] = [];\n session.maps = maps;\n const producing = readAgent(session, setPhase, active, message, meta, feed, maps);\n const audio = session.models.speak(\n session.options.mouth,\n { text: feed.iterable, voice: session.options.voice },\n { signal }\n );\n await Promise.all([producing, pipeSpeech(session, setPhase, active, audio, maps)]);\n if (!signal.aborted) setPhase({ phase: 'listening' });\n } catch (error) {\n if (signal.aborted || session.signal.aborted) return;\n throw error;\n }\n}\n\nasync function readAgent(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n message: UserMessage,\n meta: VoiceTurnMeta,\n feed: TextFeed,\n maps: SpokenMap[]\n): Promise<void> {\n const state = createSpokenState();\n let spokenChars = 0;\n let generated = '';\n try {\n for await (const event of session.conversation.prompt(\n message,\n meta,\n active.controller.signal\n )) {\n if (active.controller.signal.aborted) break;\n notePhase(setPhase, event);\n for (const piece of spokenPieces(event, state, session.options)) {\n spokenChars += piece.text.length;\n generated = rememberPiece(\n session,\n active,\n piece,\n maps,\n spokenChars,\n generated,\n state.entryId\n );\n feed.push(piece.text);\n }\n if (state.entryId) bindEntry(maps, state.entryId);\n }\n } finally {\n feed.end();\n }\n}\n\nfunction rememberPiece(\n session: Session,\n active: ActiveTurn,\n piece: SpokenPiece,\n maps: SpokenMap[],\n spokenEnd: number,\n generated: string,\n entryId: string | undefined\n): string {\n const bound = piece.entryId ?? entryId;\n maps.push({\n spokenStart: spokenEnd - piece.text.length,\n spokenEnd,\n contentIndex: piece.contentIndex,\n start: piece.start,\n end: piece.end,\n kind: piece.kind,\n ...(bound ? { entryId: bound } : {}),\n });\n if (piece.kind === 'announce') {\n session.captions.push({ type: 'announce', turnId: active.turnId, text: piece.text });\n return generated;\n }\n const text = generated + piece.text;\n session.captions.push({\n type: 'agent_generated',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n text,\n });\n return text;\n}\n\nasync function pipeSpeech(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n audio: AsyncIterable<SpeechEvent>,\n maps: SpokenMap[]\n): Promise<void> {\n for await (const event of audio) {\n if (active.controller.signal.aborted) return;\n if (event.type === 'error') {\n if (event.reason === 'aborted') return;\n throw new VoiceRequestError(event.reason, event.error);\n }\n if (event.type !== 'audio') continue;\n setPhase({ phase: 'speaking' });\n const span = translateSpan(event.span, maps);\n const playing: SpokenClock = {\n start: event.span.start,\n end: event.span.end,\n durationMs: Math.round(chunkSeconds(event.chunk) * 1000),\n };\n if (event.alignment) playing.alignment = event.alignment;\n session.playing = playing;\n const tag: AudioTag = { span };\n if (event.alignment) tag.alignment = event.alignment;\n session.out.write(event.chunk, tag);\n }\n}\n\nfunction notePlayback(session: Session, active: ActiveTurn, tag: AudioTag, offsetMs: number): void {\n const playing = session.playing;\n if (playing) {\n const upTo = utteranceIndex(spokenOffset(playing, offsetMs), session.maps);\n if (upTo > active.spokenUpTo) {\n active.spokenUpTo = upTo;\n session.captions.push({\n type: 'agent_spoken',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n upTo,\n });\n }\n }\n if (!('entryId' in tag.span)) return;\n active.cut = {\n entryId: tag.span.entryId,\n contentIndex: tag.span.contentIndex,\n heardUpTo: tag.span.end,\n heardMs: session.out.playedMs,\n };\n}\n\nfunction translateSpan(span: TextSpan, maps: SpokenMap[]): TextSpan {\n if ('entryId' in span) return span;\n const map = answerMap(span, maps);\n if (!map?.entryId) return span;\n const spokenLen = map.spokenEnd - map.spokenStart;\n const sourceLen = map.end - map.start;\n const start = map.start + scale(Math.max(0, span.start - map.spokenStart), spokenLen, sourceLen);\n const end =\n map.start + scale(Math.min(spokenLen, span.end - map.spokenStart), spokenLen, sourceLen);\n return { entryId: map.entryId, contentIndex: map.contentIndex, start, end };\n}\n\nfunction answerMap(span: { start: number; end: number }, maps: SpokenMap[]): SpokenMap | undefined {\n for (let index = maps.length - 1; index >= 0; index -= 1) {\n const item = maps[index];\n if (item?.kind !== 'answer' || !item.entryId) continue;\n if (item.spokenStart < span.end && item.spokenEnd > span.start) return item;\n }\n return undefined;\n}\n\nfunction scale(offset: number, spokenLen: number, sourceLen: number): number {\n if (spokenLen <= 0) return 0;\n return Math.round((offset * sourceLen) / spokenLen);\n}\n\nfunction bindEntry(maps: SpokenMap[], entryId: string): void {\n for (const map of maps) {\n if (map.kind === 'answer' && map.entryId === undefined) map.entryId = entryId;\n }\n}\n\nfunction notePhase(setPhase: (phase: VoicePhase) => void, event: VoiceAgentEvent): void {\n if (event.type === 'run_start' || event.type === 'turn_start') setPhase({ phase: 'thinking' });\n if (event.type === 'compaction_start') setPhase({ phase: 'compacting' });\n if (event.type === 'auto_retry_start' && event.attempt !== undefined && event.at !== undefined) {\n setPhase({ phase: 'retrying', attempt: event.attempt, at: event.at });\n }\n if (event.type === 'tool_execution_start' && event.toolName && event.toolCallId) {\n setPhase({\n phase: 'tool',\n toolName: event.toolName,\n toolCallId: event.toolCallId,\n args: event.args ?? {},\n });\n }\n}\n\nasync function interrupt(session: Session, active: ActiveTurn | undefined): Promise<undefined> {\n if (!active) return undefined;\n active.controller.abort();\n session.out.flush();\n await active.done;\n if (active.cut) {\n session.captions.push({\n type: 'agent_cut',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n heardUpTo: active.cut.heardUpTo,\n });\n await session.conversation.truncate(active.cut);\n }\n return undefined;\n}\n\nfunction bargeMode(options: VoiceTurnOptions): 'interrupt' | 'onWords' | 'ignore' {\n const mode = options.bargeIn ?? 'interrupt';\n if (mode !== 'auto') return mode;\n if (options.echoRecommendation === 'interrupt') return 'interrupt';\n if (options.echoRecommendation === 'headphones') return 'ignore';\n return 'onWords';\n}\n\nfunction shouldBarge(mode: 'interrupt' | 'onWords' | 'ignore', event: TranscriptionEvent): boolean {\n if (mode === 'ignore') return false;\n if (mode === 'interrupt') return event.type === 'speech_start';\n return event.type === 'partial' && event.text.trim().length > 0;\n}\n\nfunction phaseSetter(phases: Session['phases']): (phase: VoicePhase) => void {\n let previous = '';\n return (phase) => {\n const key = JSON.stringify(phase);\n if (key === previous) return;\n previous = key;\n phases.push(phase);\n };\n}\n\nfunction linkSignal(parent: AbortSignal): { controller: AbortController; unlink: () => void } {\n const controller = new AbortController();\n const abort = () => controller.abort();\n if (parent.aborted) controller.abort();\n else parent.addEventListener('abort', abort);\n return {\n controller,\n unlink: () => parent.removeEventListener('abort', abort),\n };\n}\n\ninterface TextFeed {\n push(text: string): void;\n end(): void;\n iterable: AsyncIterable<string>;\n}\n\nfunction textFeed(signal: AbortSignal): TextFeed {\n const queue: string[] = [];\n let ended = false;\n let wake: (() => void) | undefined;\n const notify = () => {\n const current = wake;\n wake = undefined;\n current?.();\n };\n const onAbort = () => notify();\n signal.addEventListener('abort', onAbort);\n return {\n push(text) {\n if (ended || signal.aborted || text.length === 0) return;\n queue.push(text);\n notify();\n },\n end() {\n ended = true;\n notify();\n },\n iterable: {\n async *[Symbol.asyncIterator]() {\n try {\n while (!signal.aborted) {\n const next = queue.shift();\n if (next !== undefined) {\n yield next;\n continue;\n }\n if (ended) return;\n await new Promise<void>((resolve) => {\n if (queue.length > 0 || ended || signal.aborted) resolve();\n else wake = resolve;\n });\n }\n } finally {\n signal.removeEventListener('abort', onAbort);\n }\n },\n },\n };\n}\n"],
|
|
5
|
-
"mappings": "AAOA;AAAA,EAIE;AAAA,OAEK;AACP,SAA2B,cAAc,sBAAsB;AAC/D,SAAS,yBAAyB;AAElC,SAAS,iBAAiB;
|
|
4
|
+
"sourcesContent": ["import type {\n AssistantMessageEvent,\n JsonObject,\n ProviderRequestOptions,\n ToolCall,\n UserMessage,\n} from '@earendil-works/pi-ai';\nimport {\n type AudioSink,\n type AudioSource,\n type AudioTag,\n chunkSeconds,\n type TextSpan,\n} from './audio.ts';\nimport { type SpokenClock, spokenOffset, utteranceIndex } from './captions.ts';\nimport { VoiceRequestError } from './errors.ts';\nimport type { SpeechModel, TranscriptionModel } from './models.ts';\nimport { pushQueue } from './queue.ts';\nimport {\n type AudioOut,\n type BedFollower,\n followBed,\n isAudioOut,\n type Soundscape,\n} from './soundscape.ts';\nimport type { SpeechEvent, SpeechInput } from './speech.ts';\nimport {\n createSpokenState,\n type SpokenPiece,\n spokenPieces,\n type VoiceAgentEvent,\n} from './spoken.ts';\nimport type { ListenOptions, TranscriptionEvent } from './transcription.ts';\n\nexport interface VoiceTurnMeta {\n source: 'voice';\n transcriptionModel: string;\n confidence?: number;\n}\n\nexport interface VoiceCut {\n entryId: string;\n contentIndex: number;\n heardUpTo: number;\n heardMs: number;\n}\n\nexport interface VoiceConversation {\n prompt(\n message: UserMessage,\n meta: VoiceTurnMeta,\n signal: AbortSignal\n ): AsyncIterable<VoiceAgentEvent>;\n truncate(cut: VoiceCut): Promise<void>;\n}\n\nexport type BargeIn = 'interrupt' | 'onWords' | 'ignore' | 'auto';\n\nexport type EchoRecommendation = 'interrupt' | 'onWords' | 'headphones';\n\nexport interface VoiceTurnOptions {\n ears: TranscriptionModel<string>;\n mouth: SpeechModel<string>;\n voice: string;\n speakable?: (event: AssistantMessageEvent) => string | undefined;\n announce?: (call: ToolCall) => string | undefined;\n bargeIn?: BargeIn;\n echoRecommendation?: EchoRecommendation;\n soundscape?: Soundscape;\n}\n\nexport interface VoiceModels {\n listen(\n model: TranscriptionModel<string>,\n audio: AudioSource,\n options?: ListenOptions\n ): AsyncIterable<TranscriptionEvent>;\n speak(\n model: SpeechModel<string>,\n input: SpeechInput,\n options?: ProviderRequestOptions\n ): AsyncIterable<SpeechEvent>;\n}\n\nexport type VoicePhase =\n | { phase: 'listening' }\n | { phase: 'heard'; text: string }\n | { phase: 'thinking' }\n | { phase: 'tool'; toolName: string; toolCallId: string; args: JsonObject }\n | { phase: 'retrying'; attempt: number; at: number }\n | { phase: 'compacting' }\n | { phase: 'speaking' }\n | { phase: 'idle' };\n\nexport type CaptionEvent =\n | { type: 'user_partial'; turnId: string; segmentId: string; text: string }\n | { type: 'user_final'; turnId: string; segmentId: string; text: string; confidence?: number }\n | { type: 'agent_generated'; turnId: string; utteranceId: string; text: string }\n | { type: 'agent_spoken'; turnId: string; utteranceId: string; upTo: number }\n | { type: 'agent_cut'; turnId: string; utteranceId: string; heardUpTo: number }\n | { type: 'announce'; turnId: string; text: string }\n | { type: 'thinker_progress'; turnId: string; callId: string; text: string }\n | { type: 'thinker_result'; turnId: string; callId: string; summary: string; isError: boolean };\n\nexport interface VoiceSession {\n phases: AsyncIterable<VoicePhase>;\n captions: AsyncIterable<CaptionEvent>;\n done: Promise<void>;\n}\n\ninterface SpokenMap {\n spokenStart: number;\n spokenEnd: number;\n contentIndex: number;\n start: number;\n end: number;\n entryId?: string;\n kind: SpokenPiece['kind'];\n}\n\ninterface ActiveTurn {\n controller: AbortController;\n unlink: () => void;\n done: Promise<void>;\n turnId: string;\n utteranceId: string;\n spokenUpTo: number;\n cut?: VoiceCut;\n}\n\nexport function voiceTurns(\n models: VoiceModels,\n conversation: VoiceConversation,\n mic: AudioSource,\n out: AudioSink | AudioOut,\n options: VoiceTurnOptions,\n signal: AbortSignal\n): VoiceSession {\n const phases = pushQueue<VoicePhase>();\n const captions = pushQueue<CaptionEvent>();\n const session: Session = {\n models,\n conversation,\n mic,\n out: speechOf(out),\n options,\n signal,\n phases,\n captions,\n maps: [],\n bed: attachBed(out, options.soundscape),\n };\n const done = runSession(session).finally(() => {\n phases.end();\n captions.end();\n });\n return { phases: phases.iterable, captions: captions.iterable, done };\n}\n\ninterface Session {\n models: VoiceModels;\n conversation: VoiceConversation;\n mic: AudioSource;\n out: AudioSink;\n options: VoiceTurnOptions;\n signal: AbortSignal;\n phases: ReturnType<typeof pushQueue<VoicePhase>>;\n captions: ReturnType<typeof pushQueue<CaptionEvent>>;\n maps: SpokenMap[];\n playing?: SpokenClock;\n bed: BedFollower;\n}\n\nasync function runSession(session: Session): Promise<void> {\n const mode = bargeMode(session.options);\n let active: ActiveTurn | undefined;\n let confidence: number | undefined;\n let reserved = 0;\n let openId = '';\n const openTurn = () => {\n if (!openId) {\n reserved += 1;\n openId = String(reserved);\n }\n return openId;\n };\n const stopPlayback = session.out.onPlayback((tag, offsetMs) => {\n if (!active) return;\n notePlayback(session, active, tag, offsetMs);\n });\n const setPhase = phaseSetter(session.phases, session.bed);\n try {\n if (session.signal.aborted) return;\n setPhase({ phase: 'listening' });\n for await (const event of session.models.listen(session.options.ears, session.mic, {\n signal: session.signal,\n })) {\n if (event.type === 'error') {\n if (event.reason === 'aborted' || session.signal.aborted) return;\n throw new VoiceRequestError(event.reason, event.error);\n }\n const captionId =\n event.type === 'partial' || event.type === 'final' || event.type === 'turn_end'\n ? openTurn()\n : openId;\n confidence = noteUser(session, event, captionId, confidence);\n if (shouldBarge(mode, event)) active = await interrupt(session, active);\n if (event.type !== 'turn_end') continue;\n const turnId = openId || openTurn();\n openId = '';\n if (mode === 'ignore') active = await finishTurn(active);\n else active = await interrupt(session, active);\n setPhase({ phase: 'heard', text: event.text });\n const meta = turnMeta(session.options, confidence);\n confidence = undefined;\n active = startTurn(session, setPhase, turnId, event, meta);\n }\n } finally {\n if (session.signal.aborted && active) await interrupt(session, active);\n else if (active) await active.done;\n stopPlayback();\n setPhase({ phase: 'idle' });\n session.bed.stop();\n }\n}\n\nfunction turnMeta(options: VoiceTurnOptions, confidence: number | undefined): VoiceTurnMeta {\n return {\n source: 'voice',\n transcriptionModel: options.ears.id,\n ...(confidence === undefined ? {} : { confidence }),\n };\n}\n\nfunction noteUser(\n session: Session,\n event: TranscriptionEvent,\n turnId: string,\n confidence: number | undefined\n): number | undefined {\n if (event.type === 'partial') {\n session.captions.push({\n type: 'user_partial',\n turnId,\n segmentId: event.segmentId,\n text: event.text,\n });\n }\n if (event.type === 'final') {\n session.captions.push({\n type: 'user_final',\n turnId,\n segmentId: event.segmentId,\n text: event.text,\n ...(event.confidence === undefined ? {} : { confidence: event.confidence }),\n });\n return event.confidence ?? confidence;\n }\n if (event.type === 'turn_end') {\n session.captions.push({ type: 'user_final', turnId, segmentId: 'turn', text: event.text });\n }\n return confidence;\n}\n\nasync function finishTurn(active: ActiveTurn | undefined): Promise<undefined> {\n if (!active) return undefined;\n await active.done;\n return undefined;\n}\n\nfunction startTurn(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n turnId: string,\n event: Extract<TranscriptionEvent, { type: 'turn_end' }>,\n meta: VoiceTurnMeta\n): ActiveTurn {\n const linked = linkSignal(session.signal);\n const active: ActiveTurn = {\n controller: linked.controller,\n unlink: linked.unlink,\n turnId,\n utteranceId: `${turnId}:agent`,\n spokenUpTo: 0,\n done: Promise.resolve(),\n };\n active.done = performTurn(session, setPhase, active, event.text, event.at, meta).finally(() => {\n linked.unlink();\n });\n return active;\n}\n\nasync function performTurn(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n text: string,\n at: number,\n meta: VoiceTurnMeta\n): Promise<void> {\n const signal = active.controller.signal;\n try {\n setPhase({ phase: 'thinking' });\n const message: UserMessage = { role: 'user', content: text, timestamp: at };\n const feed = textFeed(signal);\n const maps: SpokenMap[] = [];\n session.maps = maps;\n const producing = readAgent(session, setPhase, active, message, meta, feed, maps);\n const audio = session.models.speak(\n session.options.mouth,\n { text: feed.iterable, voice: session.options.voice },\n { signal }\n );\n await Promise.all([producing, pipeSpeech(session, setPhase, active, audio, maps)]);\n if (!signal.aborted) setPhase({ phase: 'listening' });\n } catch (error) {\n if (signal.aborted || session.signal.aborted) return;\n throw error;\n }\n}\n\nasync function readAgent(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n message: UserMessage,\n meta: VoiceTurnMeta,\n feed: TextFeed,\n maps: SpokenMap[]\n): Promise<void> {\n const state = createSpokenState();\n let spokenChars = 0;\n let generated = '';\n try {\n for await (const event of session.conversation.prompt(\n message,\n meta,\n active.controller.signal\n )) {\n if (active.controller.signal.aborted) break;\n notePhase(setPhase, event);\n for (const piece of spokenPieces(event, state, session.options)) {\n spokenChars += piece.text.length;\n generated = rememberPiece(\n session,\n active,\n piece,\n maps,\n spokenChars,\n generated,\n state.entryId\n );\n feed.push(piece.text);\n }\n if (state.entryId) bindEntry(maps, state.entryId);\n }\n } finally {\n feed.end();\n }\n}\n\nfunction rememberPiece(\n session: Session,\n active: ActiveTurn,\n piece: SpokenPiece,\n maps: SpokenMap[],\n spokenEnd: number,\n generated: string,\n entryId: string | undefined\n): string {\n const bound = piece.entryId ?? entryId;\n maps.push({\n spokenStart: spokenEnd - piece.text.length,\n spokenEnd,\n contentIndex: piece.contentIndex,\n start: piece.start,\n end: piece.end,\n kind: piece.kind,\n ...(bound ? { entryId: bound } : {}),\n });\n if (piece.kind === 'announce') {\n session.captions.push({ type: 'announce', turnId: active.turnId, text: piece.text });\n return generated;\n }\n const text = generated + piece.text;\n session.captions.push({\n type: 'agent_generated',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n text,\n });\n return text;\n}\n\nasync function pipeSpeech(\n session: Session,\n setPhase: (phase: VoicePhase) => void,\n active: ActiveTurn,\n audio: AsyncIterable<SpeechEvent>,\n maps: SpokenMap[]\n): Promise<void> {\n for await (const event of audio) {\n if (active.controller.signal.aborted) return;\n if (event.type === 'error') {\n if (event.reason === 'aborted') return;\n throw new VoiceRequestError(event.reason, event.error);\n }\n if (event.type !== 'audio') continue;\n setPhase({ phase: 'speaking' });\n const span = translateSpan(event.span, maps);\n const playing: SpokenClock = {\n start: event.span.start,\n end: event.span.end,\n durationMs: Math.round(chunkSeconds(event.chunk) * 1000),\n };\n if (event.alignment) playing.alignment = event.alignment;\n session.playing = playing;\n const tag: AudioTag = { span };\n if (event.alignment) tag.alignment = event.alignment;\n session.out.write(event.chunk, tag);\n }\n}\n\nfunction notePlayback(session: Session, active: ActiveTurn, tag: AudioTag, offsetMs: number): void {\n const playing = session.playing;\n if (playing) {\n const upTo = utteranceIndex(spokenOffset(playing, offsetMs), session.maps);\n if (upTo > active.spokenUpTo) {\n active.spokenUpTo = upTo;\n session.captions.push({\n type: 'agent_spoken',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n upTo,\n });\n }\n }\n if (!('entryId' in tag.span)) return;\n active.cut = {\n entryId: tag.span.entryId,\n contentIndex: tag.span.contentIndex,\n heardUpTo: tag.span.end,\n heardMs: session.out.playedMs,\n };\n}\n\nfunction translateSpan(span: TextSpan, maps: SpokenMap[]): TextSpan {\n if ('entryId' in span) return span;\n const map = answerMap(span, maps);\n if (!map?.entryId) return span;\n const spokenLen = map.spokenEnd - map.spokenStart;\n const sourceLen = map.end - map.start;\n const start = map.start + scale(Math.max(0, span.start - map.spokenStart), spokenLen, sourceLen);\n const end =\n map.start + scale(Math.min(spokenLen, span.end - map.spokenStart), spokenLen, sourceLen);\n return { entryId: map.entryId, contentIndex: map.contentIndex, start, end };\n}\n\nfunction answerMap(span: { start: number; end: number }, maps: SpokenMap[]): SpokenMap | undefined {\n for (let index = maps.length - 1; index >= 0; index -= 1) {\n const item = maps[index];\n if (item?.kind !== 'answer' || !item.entryId) continue;\n if (item.spokenStart < span.end && item.spokenEnd > span.start) return item;\n }\n return undefined;\n}\n\nfunction scale(offset: number, spokenLen: number, sourceLen: number): number {\n if (spokenLen <= 0) return 0;\n return Math.round((offset * sourceLen) / spokenLen);\n}\n\nfunction bindEntry(maps: SpokenMap[], entryId: string): void {\n for (const map of maps) {\n if (map.kind === 'answer' && map.entryId === undefined) map.entryId = entryId;\n }\n}\n\nfunction notePhase(setPhase: (phase: VoicePhase) => void, event: VoiceAgentEvent): void {\n if (event.type === 'run_start' || event.type === 'turn_start') setPhase({ phase: 'thinking' });\n if (event.type === 'compaction_start') setPhase({ phase: 'compacting' });\n if (event.type === 'auto_retry_start' && event.attempt !== undefined && event.at !== undefined) {\n setPhase({ phase: 'retrying', attempt: event.attempt, at: event.at });\n }\n if (event.type === 'tool_execution_start' && event.toolName && event.toolCallId) {\n setPhase({\n phase: 'tool',\n toolName: event.toolName,\n toolCallId: event.toolCallId,\n args: event.args ?? {},\n });\n }\n}\n\nasync function interrupt(session: Session, active: ActiveTurn | undefined): Promise<undefined> {\n if (!active) return undefined;\n active.controller.abort();\n session.out.flush();\n await active.done;\n if (active.cut) {\n session.captions.push({\n type: 'agent_cut',\n turnId: active.turnId,\n utteranceId: active.utteranceId,\n heardUpTo: active.cut.heardUpTo,\n });\n await session.conversation.truncate(active.cut);\n }\n return undefined;\n}\n\nfunction bargeMode(options: VoiceTurnOptions): 'interrupt' | 'onWords' | 'ignore' {\n const mode = options.bargeIn ?? 'interrupt';\n if (mode !== 'auto') return mode;\n if (options.echoRecommendation === 'interrupt') return 'interrupt';\n if (options.echoRecommendation === 'headphones') return 'ignore';\n return 'onWords';\n}\n\nfunction shouldBarge(mode: 'interrupt' | 'onWords' | 'ignore', event: TranscriptionEvent): boolean {\n if (mode === 'ignore') return false;\n if (mode === 'interrupt') return event.type === 'speech_start';\n return event.type === 'partial' && event.text.trim().length > 0;\n}\n\nfunction speechOf(out: AudioSink | AudioOut): AudioSink {\n if (isAudioOut(out)) return out.speech;\n return out;\n}\n\nfunction attachBed(out: AudioSink | AudioOut, soundscape: Soundscape | undefined): BedFollower {\n if (!soundscape) return idleBed();\n if (!isAudioOut(out)) return idleBed();\n return followBed(out.bed, soundscape);\n}\n\nfunction idleBed(): BedFollower {\n return {\n observe() {},\n stop() {},\n };\n}\n\nfunction phaseSetter(phases: Session['phases'], bed: BedFollower): (phase: VoicePhase) => void {\n let previous = '';\n return (phase) => {\n const key = JSON.stringify(phase);\n if (key === previous) return;\n previous = key;\n phases.push(phase);\n bed.observe(phase);\n };\n}\n\nfunction linkSignal(parent: AbortSignal): { controller: AbortController; unlink: () => void } {\n const controller = new AbortController();\n const abort = () => controller.abort();\n if (parent.aborted) controller.abort();\n else parent.addEventListener('abort', abort);\n return {\n controller,\n unlink: () => parent.removeEventListener('abort', abort),\n };\n}\n\ninterface TextFeed {\n push(text: string): void;\n end(): void;\n iterable: AsyncIterable<string>;\n}\n\nfunction textFeed(signal: AbortSignal): TextFeed {\n const queue: string[] = [];\n let ended = false;\n let wake: (() => void) | undefined;\n const notify = () => {\n const current = wake;\n wake = undefined;\n current?.();\n };\n const onAbort = () => notify();\n signal.addEventListener('abort', onAbort);\n return {\n push(text) {\n if (ended || signal.aborted || text.length === 0) return;\n queue.push(text);\n notify();\n },\n end() {\n ended = true;\n notify();\n },\n iterable: {\n async *[Symbol.asyncIterator]() {\n try {\n while (!signal.aborted) {\n const next = queue.shift();\n if (next !== undefined) {\n yield next;\n continue;\n }\n if (ended) return;\n await new Promise<void>((resolve) => {\n if (queue.length > 0 || ended || signal.aborted) resolve();\n else wake = resolve;\n });\n }\n } finally {\n signal.removeEventListener('abort', onAbort);\n }\n },\n },\n };\n}\n"],
|
|
5
|
+
"mappings": "AAOA;AAAA,EAIE;AAAA,OAEK;AACP,SAA2B,cAAc,sBAAsB;AAC/D,SAAS,yBAAyB;AAElC,SAAS,iBAAiB;AAC1B;AAAA,EAGE;AAAA,EACA;AAAA,OAEK;AAEP;AAAA,EACE;AAAA,EAEA;AAAA,OAEK;AAmGA,SAAS,WACd,QACA,cACA,KACA,KACA,SACA,QACc;AACd,QAAM,SAAS,UAAsB;AACrC,QAAM,WAAW,UAAwB;AACzC,QAAM,UAAmB;AAAA,IACvB;AAAA,IACA;AAAA,IACA;AAAA,IACA,KAAK,SAAS,GAAG;AAAA,IACjB;AAAA,IACA;AAAA,IACA;AAAA,IACA;AAAA,IACA,MAAM,CAAC;AAAA,IACP,KAAK,UAAU,KAAK,QAAQ,UAAU;AAAA,EACxC;AACA,QAAM,OAAO,WAAW,OAAO,EAAE,QAAQ,MAAM;AAC7C,WAAO,IAAI;AACX,aAAS,IAAI;AAAA,EACf,CAAC;AACD,SAAO,EAAE,QAAQ,OAAO,UAAU,UAAU,SAAS,UAAU,KAAK;AACtE;AAgBA,eAAe,WAAW,SAAiC;AACzD,QAAM,OAAO,UAAU,QAAQ,OAAO;AACtC,MAAI;AACJ,MAAI;AACJ,MAAI,WAAW;AACf,MAAI,SAAS;AACb,QAAM,WAAW,MAAM;AACrB,QAAI,CAAC,QAAQ;AACX,kBAAY;AACZ,eAAS,OAAO,QAAQ;AAAA,IAC1B;AACA,WAAO;AAAA,EACT;AACA,QAAM,eAAe,QAAQ,IAAI,WAAW,CAAC,KAAK,aAAa;AAC7D,QAAI,CAAC,OAAQ;AACb,iBAAa,SAAS,QAAQ,KAAK,QAAQ;AAAA,EAC7C,CAAC;AACD,QAAM,WAAW,YAAY,QAAQ,QAAQ,QAAQ,GAAG;AACxD,MAAI;AACF,QAAI,QAAQ,OAAO,QAAS;AAC5B,aAAS,EAAE,OAAO,YAAY,CAAC;AAC/B,qBAAiB,SAAS,QAAQ,OAAO,OAAO,QAAQ,QAAQ,MAAM,QAAQ,KAAK;AAAA,MACjF,QAAQ,QAAQ;AAAA,IAClB,CAAC,GAAG;AACF,UAAI,MAAM,SAAS,SAAS;AAC1B,YAAI,MAAM,WAAW,aAAa,QAAQ,OAAO,QAAS;AAC1D,cAAM,IAAI,kBAAkB,MAAM,QAAQ,MAAM,KAAK;AAAA,MACvD;AACA,YAAM,YACJ,MAAM,SAAS,aAAa,MAAM,SAAS,WAAW,MAAM,SAAS,aACjE,SAAS,IACT;AACN,mBAAa,SAAS,SAAS,OAAO,WAAW,UAAU;AAC3D,UAAI,YAAY,MAAM,KAAK,EAAG,UAAS,MAAM,UAAU,SAAS,MAAM;AACtE,UAAI,MAAM,SAAS,WAAY;AAC/B,YAAM,SAAS,UAAU,SAAS;AAClC,eAAS;AACT,UAAI,SAAS,SAAU,UAAS,MAAM,WAAW,MAAM;AAAA,UAClD,UAAS,MAAM,UAAU,SAAS,MAAM;AAC7C,eAAS,EAAE,OAAO,SAAS,MAAM,MAAM,KAAK,CAAC;AAC7C,YAAM,OAAO,SAAS,QAAQ,SAAS,UAAU;AACjD,mBAAa;AACb,eAAS,UAAU,SAAS,UAAU,QAAQ,OAAO,IAAI;AAAA,IAC3D;AAAA,EACF,UAAE;AACA,QAAI,QAAQ,OAAO,WAAW,OAAQ,OAAM,UAAU,SAAS,MAAM;AAAA,aAC5D,OAAQ,OAAM,OAAO;AAC9B,iBAAa;AACb,aAAS,EAAE,OAAO,OAAO,CAAC;AAC1B,YAAQ,IAAI,KAAK;AAAA,EACnB;AACF;AAEA,SAAS,SAAS,SAA2B,YAA+C;AAC1F,SAAO;AAAA,IACL,QAAQ;AAAA,IACR,oBAAoB,QAAQ,KAAK;AAAA,IACjC,GAAI,eAAe,SAAY,CAAC,IAAI,EAAE,WAAW;AAAA,EACnD;AACF;AAEA,SAAS,SACP,SACA,OACA,QACA,YACoB;AACpB,MAAI,MAAM,SAAS,WAAW;AAC5B,YAAQ,SAAS,KAAK;AAAA,MACpB,MAAM;AAAA,MACN;AAAA,MACA,WAAW,MAAM;AAAA,MACjB,MAAM,MAAM;AAAA,IACd,CAAC;AAAA,EACH;AACA,MAAI,MAAM,SAAS,SAAS;AAC1B,YAAQ,SAAS,KAAK;AAAA,MACpB,MAAM;AAAA,MACN;AAAA,MACA,WAAW,MAAM;AAAA,MACjB,MAAM,MAAM;AAAA,MACZ,GAAI,MAAM,eAAe,SAAY,CAAC,IAAI,EAAE,YAAY,MAAM,WAAW;AAAA,IAC3E,CAAC;AACD,WAAO,MAAM,cAAc;AAAA,EAC7B;AACA,MAAI,MAAM,SAAS,YAAY;AAC7B,YAAQ,SAAS,KAAK,EAAE,MAAM,cAAc,QAAQ,WAAW,QAAQ,MAAM,MAAM,KAAK,CAAC;AAAA,EAC3F;AACA,SAAO;AACT;AAEA,eAAe,WAAW,QAAoD;AAC5E,MAAI,CAAC,OAAQ,QAAO;AACpB,QAAM,OAAO;AACb,SAAO;AACT;AAEA,SAAS,UACP,SACA,UACA,QACA,OACA,MACY;AACZ,QAAM,SAAS,WAAW,QAAQ,MAAM;AACxC,QAAM,SAAqB;AAAA,IACzB,YAAY,OAAO;AAAA,IACnB,QAAQ,OAAO;AAAA,IACf;AAAA,IACA,aAAa,GAAG,MAAM;AAAA,IACtB,YAAY;AAAA,IACZ,MAAM,QAAQ,QAAQ;AAAA,EACxB;AACA,SAAO,OAAO,YAAY,SAAS,UAAU,QAAQ,MAAM,MAAM,MAAM,IAAI,IAAI,EAAE,QAAQ,MAAM;AAC7F,WAAO,OAAO;AAAA,EAChB,CAAC;AACD,SAAO;AACT;AAEA,eAAe,YACb,SACA,UACA,QACA,MACA,IACA,MACe;AACf,QAAM,SAAS,OAAO,WAAW;AACjC,MAAI;AACF,aAAS,EAAE,OAAO,WAAW,CAAC;AAC9B,UAAM,UAAuB,EAAE,MAAM,QAAQ,SAAS,MAAM,WAAW,GAAG;AAC1E,UAAM,OAAO,SAAS,MAAM;AAC5B,UAAM,OAAoB,CAAC;AAC3B,YAAQ,OAAO;AACf,UAAM,YAAY,UAAU,SAAS,UAAU,QAAQ,SAAS,MAAM,MAAM,IAAI;AAChF,UAAM,QAAQ,QAAQ,OAAO;AAAA,MAC3B,QAAQ,QAAQ;AAAA,MAChB,EAAE,MAAM,KAAK,UAAU,OAAO,QAAQ,QAAQ,MAAM;AAAA,MACpD,EAAE,OAAO;AAAA,IACX;AACA,UAAM,QAAQ,IAAI,CAAC,WAAW,WAAW,SAAS,UAAU,QAAQ,OAAO,IAAI,CAAC,CAAC;AACjF,QAAI,CAAC,OAAO,QAAS,UAAS,EAAE,OAAO,YAAY,CAAC;AAAA,EACtD,SAAS,OAAO;AACd,QAAI,OAAO,WAAW,QAAQ,OAAO,QAAS;AAC9C,UAAM;AAAA,EACR;AACF;AAEA,eAAe,UACb,SACA,UACA,QACA,SACA,MACA,MACA,MACe;AACf,QAAM,QAAQ,kBAAkB;AAChC,MAAI,cAAc;AAClB,MAAI,YAAY;AAChB,MAAI;AACF,qBAAiB,SAAS,QAAQ,aAAa;AAAA,MAC7C;AAAA,MACA;AAAA,MACA,OAAO,WAAW;AAAA,IACpB,GAAG;AACD,UAAI,OAAO,WAAW,OAAO,QAAS;AACtC,gBAAU,UAAU,KAAK;AACzB,iBAAW,SAAS,aAAa,OAAO,OAAO,QAAQ,OAAO,GAAG;AAC/D,uBAAe,MAAM,KAAK;AAC1B,oBAAY;AAAA,UACV;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA,MAAM;AAAA,QACR;AACA,aAAK,KAAK,MAAM,IAAI;AAAA,MACtB;AACA,UAAI,MAAM,QAAS,WAAU,MAAM,MAAM,OAAO;AAAA,IAClD;AAAA,EACF,UAAE;AACA,SAAK,IAAI;AAAA,EACX;AACF;AAEA,SAAS,cACP,SACA,QACA,OACA,MACA,WACA,WACA,SACQ;AACR,QAAM,QAAQ,MAAM,WAAW;AAC/B,OAAK,KAAK;AAAA,IACR,aAAa,YAAY,MAAM,KAAK;AAAA,IACpC;AAAA,IACA,cAAc,MAAM;AAAA,IACpB,OAAO,MAAM;AAAA,IACb,KAAK,MAAM;AAAA,IACX,MAAM,MAAM;AAAA,IACZ,GAAI,QAAQ,EAAE,SAAS,MAAM,IAAI,CAAC;AAAA,EACpC,CAAC;AACD,MAAI,MAAM,SAAS,YAAY;AAC7B,YAAQ,SAAS,KAAK,EAAE,MAAM,YAAY,QAAQ,OAAO,QAAQ,MAAM,MAAM,KAAK,CAAC;AACnF,WAAO;AAAA,EACT;AACA,QAAM,OAAO,YAAY,MAAM;AAC/B,UAAQ,SAAS,KAAK;AAAA,IACpB,MAAM;AAAA,IACN,QAAQ,OAAO;AAAA,IACf,aAAa,OAAO;AAAA,IACpB;AAAA,EACF,CAAC;AACD,SAAO;AACT;AAEA,eAAe,WACb,SACA,UACA,QACA,OACA,MACe;AACf,mBAAiB,SAAS,OAAO;AAC/B,QAAI,OAAO,WAAW,OAAO,QAAS;AACtC,QAAI,MAAM,SAAS,SAAS;AAC1B,UAAI,MAAM,WAAW,UAAW;AAChC,YAAM,IAAI,kBAAkB,MAAM,QAAQ,MAAM,KAAK;AAAA,IACvD;AACA,QAAI,MAAM,SAAS,QAAS;AAC5B,aAAS,EAAE,OAAO,WAAW,CAAC;AAC9B,UAAM,OAAO,cAAc,MAAM,MAAM,IAAI;AAC3C,UAAM,UAAuB;AAAA,MAC3B,OAAO,MAAM,KAAK;AAAA,MAClB,KAAK,MAAM,KAAK;AAAA,MAChB,YAAY,KAAK,MAAM,aAAa,MAAM,KAAK,IAAI,GAAI;AAAA,IACzD;AACA,QAAI,MAAM,UAAW,SAAQ,YAAY,MAAM;AAC/C,YAAQ,UAAU;AAClB,UAAM,MAAgB,EAAE,KAAK;AAC7B,QAAI,MAAM,UAAW,KAAI,YAAY,MAAM;AAC3C,YAAQ,IAAI,MAAM,MAAM,OAAO,GAAG;AAAA,EACpC;AACF;AAEA,SAAS,aAAa,SAAkB,QAAoB,KAAe,UAAwB;AACjG,QAAM,UAAU,QAAQ;AACxB,MAAI,SAAS;AACX,UAAM,OAAO,eAAe,aAAa,SAAS,QAAQ,GAAG,QAAQ,IAAI;AACzE,QAAI,OAAO,OAAO,YAAY;AAC5B,aAAO,aAAa;AACpB,cAAQ,SAAS,KAAK;AAAA,QACpB,MAAM;AAAA,QACN,QAAQ,OAAO;AAAA,QACf,aAAa,OAAO;AAAA,QACpB;AAAA,MACF,CAAC;AAAA,IACH;AAAA,EACF;AACA,MAAI,EAAE,aAAa,IAAI,MAAO;AAC9B,SAAO,MAAM;AAAA,IACX,SAAS,IAAI,KAAK;AAAA,IAClB,cAAc,IAAI,KAAK;AAAA,IACvB,WAAW,IAAI,KAAK;AAAA,IACpB,SAAS,QAAQ,IAAI;AAAA,EACvB;AACF;AAEA,SAAS,cAAc,MAAgB,MAA6B;AAClE,MAAI,aAAa,KAAM,QAAO;AAC9B,QAAM,MAAM,UAAU,MAAM,IAAI;AAChC,MAAI,CAAC,KAAK,QAAS,QAAO;AAC1B,QAAM,YAAY,IAAI,YAAY,IAAI;AACtC,QAAM,YAAY,IAAI,MAAM,IAAI;AAChC,QAAM,QAAQ,IAAI,QAAQ,MAAM,KAAK,IAAI,GAAG,KAAK,QAAQ,IAAI,WAAW,GAAG,WAAW,SAAS;AAC/F,QAAM,MACJ,IAAI,QAAQ,MAAM,KAAK,IAAI,WAAW,KAAK,MAAM,IAAI,WAAW,GAAG,WAAW,SAAS;AACzF,SAAO,EAAE,SAAS,IAAI,SAAS,cAAc,IAAI,cAAc,OAAO,IAAI;AAC5E;AAEA,SAAS,UAAU,MAAsC,MAA0C;AACjG,WAAS,QAAQ,KAAK,SAAS,GAAG,SAAS,GAAG,SAAS,GAAG;AACxD,UAAM,OAAO,KAAK,KAAK;AACvB,QAAI,MAAM,SAAS,YAAY,CAAC,KAAK,QAAS;AAC9C,QAAI,KAAK,cAAc,KAAK,OAAO,KAAK,YAAY,KAAK,MAAO,QAAO;AAAA,EACzE;AACA,SAAO;AACT;AAEA,SAAS,MAAM,QAAgB,WAAmB,WAA2B;AAC3E,MAAI,aAAa,EAAG,QAAO;AAC3B,SAAO,KAAK,MAAO,SAAS,YAAa,SAAS;AACpD;AAEA,SAAS,UAAU,MAAmB,SAAuB;AAC3D,aAAW,OAAO,MAAM;AACtB,QAAI,IAAI,SAAS,YAAY,IAAI,YAAY,OAAW,KAAI,UAAU;AAAA,EACxE;AACF;AAEA,SAAS,UAAU,UAAuC,OAA8B;AACtF,MAAI,MAAM,SAAS,eAAe,MAAM,SAAS,aAAc,UAAS,EAAE,OAAO,WAAW,CAAC;AAC7F,MAAI,MAAM,SAAS,mBAAoB,UAAS,EAAE,OAAO,aAAa,CAAC;AACvE,MAAI,MAAM,SAAS,sBAAsB,MAAM,YAAY,UAAa,MAAM,OAAO,QAAW;AAC9F,aAAS,EAAE,OAAO,YAAY,SAAS,MAAM,SAAS,IAAI,MAAM,GAAG,CAAC;AAAA,EACtE;AACA,MAAI,MAAM,SAAS,0BAA0B,MAAM,YAAY,MAAM,YAAY;AAC/E,aAAS;AAAA,MACP,OAAO;AAAA,MACP,UAAU,MAAM;AAAA,MAChB,YAAY,MAAM;AAAA,MAClB,MAAM,MAAM,QAAQ,CAAC;AAAA,IACvB,CAAC;AAAA,EACH;AACF;AAEA,eAAe,UAAU,SAAkB,QAAoD;AAC7F,MAAI,CAAC,OAAQ,QAAO;AACpB,SAAO,WAAW,MAAM;AACxB,UAAQ,IAAI,MAAM;AAClB,QAAM,OAAO;AACb,MAAI,OAAO,KAAK;AACd,YAAQ,SAAS,KAAK;AAAA,MACpB,MAAM;AAAA,MACN,QAAQ,OAAO;AAAA,MACf,aAAa,OAAO;AAAA,MACpB,WAAW,OAAO,IAAI;AAAA,IACxB,CAAC;AACD,UAAM,QAAQ,aAAa,SAAS,OAAO,GAAG;AAAA,EAChD;AACA,SAAO;AACT;AAEA,SAAS,UAAU,SAA+D;AAChF,QAAM,OAAO,QAAQ,WAAW;AAChC,MAAI,SAAS,OAAQ,QAAO;AAC5B,MAAI,QAAQ,uBAAuB,YAAa,QAAO;AACvD,MAAI,QAAQ,uBAAuB,aAAc,QAAO;AACxD,SAAO;AACT;AAEA,SAAS,YAAY,MAA0C,OAAoC;AACjG,MAAI,SAAS,SAAU,QAAO;AAC9B,MAAI,SAAS,YAAa,QAAO,MAAM,SAAS;AAChD,SAAO,MAAM,SAAS,aAAa,MAAM,KAAK,KAAK,EAAE,SAAS;AAChE;AAEA,SAAS,SAAS,KAAsC;AACtD,MAAI,WAAW,GAAG,EAAG,QAAO,IAAI;AAChC,SAAO;AACT;AAEA,SAAS,UAAU,KAA2B,YAAiD;AAC7F,MAAI,CAAC,WAAY,QAAO,QAAQ;AAChC,MAAI,CAAC,WAAW,GAAG,EAAG,QAAO,QAAQ;AACrC,SAAO,UAAU,IAAI,KAAK,UAAU;AACtC;AAEA,SAAS,UAAuB;AAC9B,SAAO;AAAA,IACL,UAAU;AAAA,IAAC;AAAA,IACX,OAAO;AAAA,IAAC;AAAA,EACV;AACF;AAEA,SAAS,YAAY,QAA2B,KAA+C;AAC7F,MAAI,WAAW;AACf,SAAO,CAAC,UAAU;AAChB,UAAM,MAAM,KAAK,UAAU,KAAK;AAChC,QAAI,QAAQ,SAAU;AACtB,eAAW;AACX,WAAO,KAAK,KAAK;AACjB,QAAI,QAAQ,KAAK;AAAA,EACnB;AACF;AAEA,SAAS,WAAW,QAA0E;AAC5F,QAAM,aAAa,IAAI,gBAAgB;AACvC,QAAM,QAAQ,MAAM,WAAW,MAAM;AACrC,MAAI,OAAO,QAAS,YAAW,MAAM;AAAA,MAChC,QAAO,iBAAiB,SAAS,KAAK;AAC3C,SAAO;AAAA,IACL;AAAA,IACA,QAAQ,MAAM,OAAO,oBAAoB,SAAS,KAAK;AAAA,EACzD;AACF;AAQA,SAAS,SAAS,QAA+B;AAC/C,QAAM,QAAkB,CAAC;AACzB,MAAI,QAAQ;AACZ,MAAI;AACJ,QAAM,SAAS,MAAM;AACnB,UAAM,UAAU;AAChB,WAAO;AACP,cAAU;AAAA,EACZ;AACA,QAAM,UAAU,MAAM,OAAO;AAC7B,SAAO,iBAAiB,SAAS,OAAO;AACxC,SAAO;AAAA,IACL,KAAK,MAAM;AACT,UAAI,SAAS,OAAO,WAAW,KAAK,WAAW,EAAG;AAClD,YAAM,KAAK,IAAI;AACf,aAAO;AAAA,IACT;AAAA,IACA,MAAM;AACJ,cAAQ;AACR,aAAO;AAAA,IACT;AAAA,IACA,UAAU;AAAA,MACR,QAAQ,OAAO,aAAa,IAAI;AAC9B,YAAI;AACF,iBAAO,CAAC,OAAO,SAAS;AACtB,kBAAM,OAAO,MAAM,MAAM;AACzB,gBAAI,SAAS,QAAW;AACtB,oBAAM;AACN;AAAA,YACF;AACA,gBAAI,MAAO;AACX,kBAAM,IAAI,QAAc,CAAC,YAAY;AACnC,kBAAI,MAAM,SAAS,KAAK,SAAS,OAAO,QAAS,SAAQ;AAAA,kBACpD,QAAO;AAAA,YACd,CAAC;AAAA,UACH;AAAA,QACF,UAAE;AACA,iBAAO,oBAAoB,SAAS,OAAO;AAAA,QAC7C;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;",
|
|
6
6
|
"names": []
|
|
7
7
|
}
|
package/dist/types/index.d.ts
CHANGED
|
@@ -10,6 +10,8 @@ export { getVoiceModel, listVoiceModels, registerVoiceModel, } from './models.ts
|
|
|
10
10
|
export type { DeliveryPlan, RealtimeApiProvider, RealtimeContext, RealtimeEvent, RealtimeOptions, RealtimeSession, RealtimeSocket, RealtimeToken, TruncationPlan, } from './realtime.ts';
|
|
11
11
|
export { planDelivery, planTruncation, RealtimeLedger, realtimeSocket } from './realtime.ts';
|
|
12
12
|
export { connectRealtime, listen, registerRealtimeApiProvider, registerSpeechApiProvider, registerTranscriptionApiProvider, speak, transcribe, } from './registry.ts';
|
|
13
|
+
export type { AudioOut, BedChannel, Cue, CueSource, Soundscape } from './soundscape.ts';
|
|
14
|
+
export { followBed } from './soundscape.ts';
|
|
13
15
|
export type { SpeechApiProvider, SpeechEvent, SpeechInput } from './speech.ts';
|
|
14
16
|
export { SpeechEventStream } from './speech.ts';
|
|
15
17
|
export type { VoiceAgentEvent, VoiceMessageChange } from './spoken.ts';
|
|
@@ -0,0 +1,47 @@
|
|
|
1
|
+
import type { AudioSink, AudioSource } from './audio.ts';
|
|
2
|
+
import type { VoicePhase } from './turns.ts';
|
|
3
|
+
export type CueSource = {
|
|
4
|
+
url: string;
|
|
5
|
+
} | {
|
|
6
|
+
buffer: AudioBuffer;
|
|
7
|
+
} | {
|
|
8
|
+
synth: (ctx: BaseAudioContext) => AudioNode;
|
|
9
|
+
};
|
|
10
|
+
export type Cue = {
|
|
11
|
+
kind: 'loop';
|
|
12
|
+
source: CueSource;
|
|
13
|
+
gain: number;
|
|
14
|
+
} | {
|
|
15
|
+
kind: 'oneShot';
|
|
16
|
+
source: CueSource;
|
|
17
|
+
gain: number;
|
|
18
|
+
} | {
|
|
19
|
+
kind: 'keep';
|
|
20
|
+
} | {
|
|
21
|
+
kind: 'silence';
|
|
22
|
+
};
|
|
23
|
+
export interface Soundscape {
|
|
24
|
+
cue(next: VoicePhase, previous: VoicePhase): Cue | readonly Cue[];
|
|
25
|
+
graceMs?: number;
|
|
26
|
+
fadeMs?: number;
|
|
27
|
+
duckUnderSpeech?: number;
|
|
28
|
+
duckUnderUser?: number;
|
|
29
|
+
}
|
|
30
|
+
export interface BedChannel {
|
|
31
|
+
play(cue: Cue, fadeMs: number): void;
|
|
32
|
+
duck(gain: number, rampMs: number): void;
|
|
33
|
+
stop(fadeMs: number): void;
|
|
34
|
+
}
|
|
35
|
+
export interface AudioOut {
|
|
36
|
+
readonly context: BaseAudioContext;
|
|
37
|
+
speech: AudioSink;
|
|
38
|
+
bed: BedChannel;
|
|
39
|
+
readonly reference: AudioSource;
|
|
40
|
+
readonly playback: 'webaudio' | 'rtc-loopback';
|
|
41
|
+
}
|
|
42
|
+
export interface BedFollower {
|
|
43
|
+
observe(phase: VoicePhase): void;
|
|
44
|
+
stop(): void;
|
|
45
|
+
}
|
|
46
|
+
export declare function isAudioOut(out: AudioSink | AudioOut): out is AudioOut;
|
|
47
|
+
export declare function followBed(bed: BedChannel, soundscape: Soundscape, wait?: (run: () => void, delayMs: number) => ReturnType<typeof setTimeout>): BedFollower;
|
package/dist/types/turns.d.ts
CHANGED
|
@@ -1,6 +1,7 @@
|
|
|
1
1
|
import type { AssistantMessageEvent, JsonObject, ProviderRequestOptions, ToolCall, UserMessage } from '@earendil-works/pi-ai';
|
|
2
2
|
import { type AudioSink, type AudioSource } from './audio.ts';
|
|
3
3
|
import type { SpeechModel, TranscriptionModel } from './models.ts';
|
|
4
|
+
import { type AudioOut, type Soundscape } from './soundscape.ts';
|
|
4
5
|
import type { SpeechEvent, SpeechInput } from './speech.ts';
|
|
5
6
|
import { type VoiceAgentEvent } from './spoken.ts';
|
|
6
7
|
import type { ListenOptions, TranscriptionEvent } from './transcription.ts';
|
|
@@ -29,6 +30,7 @@ export interface VoiceTurnOptions {
|
|
|
29
30
|
announce?: (call: ToolCall) => string | undefined;
|
|
30
31
|
bargeIn?: BargeIn;
|
|
31
32
|
echoRecommendation?: EchoRecommendation;
|
|
33
|
+
soundscape?: Soundscape;
|
|
32
34
|
}
|
|
33
35
|
export interface VoiceModels {
|
|
34
36
|
listen(model: TranscriptionModel<string>, audio: AudioSource, options?: ListenOptions): AsyncIterable<TranscriptionEvent>;
|
|
@@ -104,4 +106,4 @@ export interface VoiceSession {
|
|
|
104
106
|
captions: AsyncIterable<CaptionEvent>;
|
|
105
107
|
done: Promise<void>;
|
|
106
108
|
}
|
|
107
|
-
export declare function voiceTurns(models: VoiceModels, conversation: VoiceConversation, mic: AudioSource, out: AudioSink, options: VoiceTurnOptions, signal: AbortSignal): VoiceSession;
|
|
109
|
+
export declare function voiceTurns(models: VoiceModels, conversation: VoiceConversation, mic: AudioSource, out: AudioSink | AudioOut, options: VoiceTurnOptions, signal: AbortSignal): VoiceSession;
|
package/package.json
CHANGED