| 1 | import { createStore } from "/js/AlpineStore.js"; |
| 2 | import { toastFrontendError } from "/components/notifications/notification-store.js"; |
| 3 | import { callJsonApi } from "/js/api.js"; |
| 4 | import { sttService } from "/js/stt-service.js"; |
| 5 | import { ttsService } from "/js/tts-service.js"; |
| 6 | import { sendMessage, updateChatInput } from "/index.js"; |
| 7 | |
| 8 | const PLUGIN_NAME = "_whisper_stt"; |
| 9 | |
| 10 | const Status = { |
| 11 | INACTIVE: "inactive", |
| 12 | ACTIVATING: "activating", |
| 13 | LISTENING: "listening", |
| 14 | RECORDING: "recording", |
| 15 | WAITING: "waiting", |
| 16 | PROCESSING: "processing", |
| 17 | }; |
| 18 | |
| 19 | const MicButtonClasses = [ |
| 20 | "mic-disabled", |
| 21 | "mic-inactive", |
| 22 | "mic-activating", |
| 23 | "mic-listening", |
| 24 | "mic-recording", |
| 25 | "mic-waiting", |
| 26 | "mic-processing", |
| 27 | ]; |
| 28 | |
| 29 | const MicStatusLabels = { |
| 30 | disabled: "Whisper STT disabled", |
| 31 | inactive: "Microphone standby", |
| 32 | activating: "Microphone activating", |
| 33 | listening: "Listening for speech", |
| 34 | recording: "Recording voice", |
| 35 | waiting: "Waiting for final silence", |
| 36 | processing: "Transcribing voice", |
| 37 | }; |
| 38 | |
| 39 | function clearMicrophoneTooltip(element) { |
| 40 | const tooltip = globalThis.bootstrap?.Tooltip?.getInstance?.(element); |
| 41 | tooltip?.dispose?.(); |
| 42 | element.removeAttribute("title"); |
| 43 | element.removeAttribute("data-bs-original-title"); |
| 44 | element.removeAttribute("data-bs-toggle"); |
| 45 | element.removeAttribute("data-bs-trigger"); |
| 46 | element.removeAttribute("data-bs-tooltip-initialized"); |
| 47 | } |
| 48 | |
| 49 | const model = { |
| 50 | runtimeInitialized: false, |
| 51 | statusLoaded: false, |
| 52 | loading: false, |
| 53 | error: "", |
| 54 | enabled: false, |
| 55 | config: { |
| 56 | model_size: "base", |
| 57 | language: "en", |
| 58 | message_mode: "send", |
| 59 | silence_threshold: 0.3, |
| 60 | silence_duration: 1000, |
| 61 | waiting_timeout: 2000, |
| 62 | }, |
| 63 | modelReady: false, |
| 64 | modelLoading: false, |
| 65 | loadedModel: "", |
| 66 | packageVersion: "", |
| 67 | providerCleanup: null, |
| 68 | microphoneInput: null, |
| 69 | finalTranscriptHandler: null, |
| 70 | isProcessingClick: false, |
| 71 | devices: [], |
| 72 | selectedDevice: "", |
| 73 | requestingPermission: false, |
| 74 | _ttsListener: null, |
| 75 | _deviceChangeListenerBound: false, |
| 76 | |
| 77 | async initRuntime() { |
| 78 | if (this.runtimeInitialized) return; |
| 79 | |
| 80 | this.runtimeInitialized = true; |
| 81 | await this.loadDevices(); |
| 82 | await this.refreshStatus({ suppressError: true }); |
| 83 | |
| 84 | if (!this._deviceChangeListenerBound) { |
| 85 | navigator.mediaDevices?.addEventListener?.("devicechange", () => { |
| 86 | void this.loadDevices(); |
| 87 | }); |
| 88 | this._deviceChangeListenerBound = true; |
| 89 | } |
| 90 | |
| 91 | if (!this._ttsListener) { |
| 92 | this._ttsListener = (event) => { |
| 93 | if (event?.detail?.isSpeaking && this.micStatus !== Status.INACTIVE) { |
| 94 | this.stop(); |
| 95 | } |
| 96 | }; |
| 97 | ttsService.addEventListener("statechange", this._ttsListener); |
| 98 | } |
| 99 | }, |
| 100 | |
| 101 | async ensureStatusLoaded({ force = false, suppressError = true } = {}) { |
| 102 | if ((!this.statusLoaded || force) && !this.loading) { |
| 103 | await this.refreshStatus({ suppressError }); |
| 104 | } |
| 105 | }, |
| 106 | |
| 107 | async refreshStatus({ suppressError = false } = {}) { |
| 108 | this.loading = true; |
| 109 | this.error = ""; |
| 110 | |
| 111 | try { |
| 112 | const status = await callJsonApi(`/plugins/${PLUGIN_NAME}/status`, {}); |
| 113 | this.statusLoaded = true; |
| 114 | this.enabled = !!status?.enabled; |
| 115 | this.config = { |
| 116 | model_size: status?.config?.model_size || "base", |
| 117 | language: status?.config?.language || "en", |
| 118 | message_mode: |
| 119 | status?.config?.message_mode === "draft" ? "draft" : "send", |
| 120 | silence_threshold: Number(status?.config?.silence_threshold ?? 0.3), |
| 121 | silence_duration: Number(status?.config?.silence_duration ?? 1000), |
| 122 | waiting_timeout: Number(status?.config?.waiting_timeout ?? 2000), |
| 123 | }; |
| 124 | this.modelReady = !!status?.model?.ready; |
| 125 | this.modelLoading = !!status?.model?.loading; |
| 126 | this.loadedModel = status?.model?.loaded_model || ""; |
| 127 | this.packageVersion = status?.package?.version || ""; |
| 128 | |
| 129 | if (this.enabled) { |
| 130 | this.registerProvider(); |
| 131 | } else { |
| 132 | this.unregisterProvider(); |
| 133 | } |
| 134 | } catch (error) { |
| 135 | this.error = error instanceof Error ? error.message : String(error); |
| 136 | this.unregisterProvider(); |
| 137 | if (!suppressError) { |
| 138 | void toastFrontendError(this.error, "Whisper STT"); |
| 139 | } |
| 140 | } finally { |
| 141 | this.loading = false; |
| 142 | this.updateMicrophoneButtonUI(); |
| 143 | } |
| 144 | }, |
| 145 | |
| 146 | registerProvider() { |
| 147 | if (this.providerCleanup || !this.enabled) return; |
| 148 | |
| 149 | this.providerCleanup = sttService.registerProvider(PLUGIN_NAME, { |
| 150 | handleMicrophoneClick: async () => await this.handleMicrophoneClick(), |
| 151 | requestMicrophonePermission: async () => |
| 152 | await this.requestMicrophonePermission(), |
| 153 | updateMicrophoneButtonUI: () => this.updateMicrophoneButtonUI(), |
| 154 | stop: () => this.stop(), |
| 155 | getStatus: () => this.micStatus, |
| 156 | }); |
| 157 | |
| 158 | sttService.emitStatusChange(this.micStatus); |
| 159 | this.updateMicrophoneButtonUI(); |
| 160 | }, |
| 161 | |
| 162 | unregisterProvider() { |
| 163 | if (!this.providerCleanup) return; |
| 164 | |
| 165 | this.stop(); |
| 166 | this.providerCleanup(); |
| 167 | this.providerCleanup = null; |
| 168 | }, |
| 169 | |
| 170 | async openConfig() { |
| 171 | const { store } = await import("/components/plugins/plugin-settings-store.js"); |
| 172 | await store.openConfig(PLUGIN_NAME); |
| 173 | }, |
| 174 | |
| 175 | openPanel() { |
| 176 | window.openModal?.(`/plugins/${PLUGIN_NAME}/webui/main.html`); |
| 177 | }, |
| 178 | |
| 179 | updateMicrophoneButtonUI() { |
| 180 | const status = this.enabled ? this.micStatus : "disabled"; |
| 181 | const label = MicStatusLabels[status] || "Microphone"; |
| 182 | const microphoneButtons = document.querySelectorAll( |
| 183 | "[data-whisper-microphone], #microphone-button", |
| 184 | ); |
| 185 | for (const microphoneButton of microphoneButtons) { |
| 186 | clearMicrophoneTooltip(microphoneButton); |
| 187 | microphoneButton.classList.remove(...MicButtonClasses); |
| 188 | microphoneButton.classList.add(`mic-${status}`); |
| 189 | microphoneButton.setAttribute("data-status", status); |
| 190 | microphoneButton.setAttribute("aria-label", label); |
| 191 | microphoneButton.setAttribute( |
| 192 | "aria-pressed", |
| 193 | String( |
| 194 | status !== "disabled" && |
| 195 | status !== Status.INACTIVE && |
| 196 | status !== Status.ACTIVATING, |
| 197 | ), |
| 198 | ); |
| 199 | } |
| 200 | }, |
| 201 | |
| 202 | async loadDevices() { |
| 203 | try { |
| 204 | const devices = await navigator.mediaDevices.enumerateDevices(); |
| 205 | this.devices = devices.filter( |
| 206 | (device) => device.kind === "audioinput" && device.deviceId, |
| 207 | ); |
| 208 | |
| 209 | const saved = localStorage.getItem("whisperSttSelectedDevice") || ""; |
| 210 | const savedStillExists = this.devices.some( |
| 211 | (device) => device.deviceId === saved, |
| 212 | ); |
| 213 | |
| 214 | if (savedStillExists) { |
| 215 | this.selectedDevice = saved; |
| 216 | return; |
| 217 | } |
| 218 | |
| 219 | const defaultDevice = |
| 220 | this.devices.find((device) => device.deviceId === "default") || |
| 221 | this.devices[0]; |
| 222 | this.selectedDevice = defaultDevice?.deviceId || ""; |
| 223 | } catch (error) { |
| 224 | console.error("[Whisper STT] Failed to enumerate audio devices", error); |
| 225 | this.devices = []; |
| 226 | this.selectedDevice = ""; |
| 227 | } |
| 228 | }, |
| 229 | |
| 230 | async selectDevice(deviceId) { |
| 231 | this.selectedDevice = deviceId || ""; |
| 232 | localStorage.setItem("whisperSttSelectedDevice", this.selectedDevice); |
| 233 | |
| 234 | if (this.microphoneInput?.selectedDeviceId !== this.selectedDevice) { |
| 235 | this.stop(); |
| 236 | this.microphoneInput = null; |
| 237 | } |
| 238 | }, |
| 239 | |
| 240 | getSelectedDevice() { |
| 241 | let device = this.devices.find( |
| 242 | (candidate) => candidate.deviceId === this.selectedDevice, |
| 243 | ); |
| 244 | |
| 245 | if (!device && this.devices.length > 0) { |
| 246 | device = |
| 247 | this.devices.find((candidate) => candidate.deviceId === "default") || |
| 248 | this.devices[0]; |
| 249 | } |
| 250 | |
| 251 | return device || null; |
| 252 | }, |
| 253 | |
| 254 | async requestMicrophonePermission() { |
| 255 | this.requestingPermission = true; |
| 256 | |
| 257 | try { |
| 258 | const stream = await navigator.mediaDevices.getUserMedia({ audio: true }); |
| 259 | stream.getTracks().forEach((track) => track.stop()); |
| 260 | await this.loadDevices(); |
| 261 | return true; |
| 262 | } catch (error) { |
| 263 | console.error("[Whisper STT] Microphone permission denied", error); |
| 264 | globalThis.toast?.( |
| 265 | "Microphone access denied. Please enable microphone access in your browser settings.", |
| 266 | "error", |
| 267 | ); |
| 268 | return false; |
| 269 | } finally { |
| 270 | this.requestingPermission = false; |
| 271 | } |
| 272 | }, |
| 273 | |
| 274 | async handleMicrophoneClick(finalTranscriptHandler = null) { |
| 275 | if (this.isProcessingClick) return; |
| 276 | |
| 277 | this.finalTranscriptHandler = |
| 278 | typeof finalTranscriptHandler === "function" ? finalTranscriptHandler : null; |
| 279 | this.isProcessingClick = true; |
| 280 | try { |
| 281 | await this.ensureStatusLoaded({ force: true, suppressError: false }); |
| 282 | if (!this.enabled) { |
| 283 | this.finalTranscriptHandler = null; |
| 284 | globalThis.justToast?.("Whisper STT is disabled.", "info"); |
| 285 | return; |
| 286 | } |
| 287 | |
| 288 | ttsService.stop(); |
| 289 | |
| 290 | const selectedDevice = this.getSelectedDevice(); |
| 291 | if ( |
| 292 | this.microphoneInput && |
| 293 | this.microphoneInput.selectedDeviceId !== (selectedDevice?.deviceId || "") |
| 294 | ) { |
| 295 | this.stop(); |
| 296 | this.microphoneInput = null; |
| 297 | } |
| 298 | |
| 299 | if (!this.microphoneInput) { |
| 300 | await this.initMicrophone(); |
| 301 | } |
| 302 | |
| 303 | if (this.microphoneInput) { |
| 304 | await this.microphoneInput.toggle(); |
| 305 | } |
| 306 | } finally { |
| 307 | setTimeout(() => { |
| 308 | this.isProcessingClick = false; |
| 309 | }, 300); |
| 310 | } |
| 311 | }, |
| 312 | |
| 313 | async initMicrophone() { |
| 314 | if (this.microphoneInput) return this.microphoneInput; |
| 315 | |
| 316 | const input = new MicrophoneInput(this, async (text, isFinal) => { |
| 317 | if (isFinal) { |
| 318 | await this.deliverVoiceMessage(text); |
| 319 | } |
| 320 | }); |
| 321 | |
| 322 | const initialized = await input.initialize(); |
| 323 | this.microphoneInput = initialized ? input : null; |
| 324 | return this.microphoneInput; |
| 325 | }, |
| 326 | |
| 327 | async deliverVoiceMessage(text) { |
| 328 | if (this.finalTranscriptHandler) { |
| 329 | await this.finalTranscriptHandler(text, { |
| 330 | messageMode: this.config.message_mode, |
| 331 | sendImmediately: this.sendsImmediately, |
| 332 | }); |
| 333 | return; |
| 334 | } |
| 335 | await this.sendVoiceMessage(text); |
| 336 | }, |
| 337 | |
| 338 | async sendVoiceMessage(text) { |
| 339 | const message = String(text || "").trim(); |
| 340 | if (!message) return; |
| 341 | |
| 342 | updateChatInput(message); |
| 343 | |
| 344 | if (!this.sendsImmediately) { |
| 345 | this.stop(); |
| 346 | return; |
| 347 | } |
| 348 | |
| 349 | if (!this.microphoneInput?.messageSent) { |
| 350 | this.microphoneInput.messageSent = true; |
| 351 | await sendMessage(); |
| 352 | } |
| 353 | }, |
| 354 | |
| 355 | notifyStatusChange() { |
| 356 | this.updateMicrophoneButtonUI(); |
| 357 | sttService.emitStatusChange(this.micStatus); |
| 358 | }, |
| 359 | |
| 360 | stop() { |
| 361 | if (this.microphoneInput) { |
| 362 | this.microphoneInput.status = Status.INACTIVE; |
| 363 | this.microphoneInput.dispose(); |
| 364 | this.microphoneInput = null; |
| 365 | } |
| 366 | |
| 367 | this.finalTranscriptHandler = null; |
| 368 | |
| 369 | this.notifyStatusChange(); |
| 370 | }, |
| 371 | |
| 372 | get micStatus() { |
| 373 | return this.microphoneInput?.status || Status.INACTIVE; |
| 374 | }, |
| 375 | |
| 376 | get sendsImmediately() { |
| 377 | return this.config.message_mode !== "draft"; |
| 378 | }, |
| 379 | |
| 380 | get messageModeLabel() { |
| 381 | return this.sendsImmediately ? "Send immediately" : "Draft in composer"; |
| 382 | }, |
| 383 | |
| 384 | get statusText() { |
| 385 | if (!this.enabled) return "Disabled"; |
| 386 | if (this.modelLoading) return "Loading"; |
| 387 | if (this.modelReady) return "Ready"; |
| 388 | return "Idle"; |
| 389 | }, |
| 390 | |
| 391 | get statusClass() { |
| 392 | if (!this.enabled) return "warn"; |
| 393 | if (this.modelLoading) return "warn"; |
| 394 | if (this.modelReady) return "ok"; |
| 395 | return "warn"; |
| 396 | }, |
| 397 | |
| 398 | get selectedDeviceLabel() { |
| 399 | const device = this.getSelectedDevice(); |
| 400 | if (!device) return "System default"; |
| 401 | return device.label || "System default"; |
| 402 | }, |
| 403 | }; |
| 404 | |
| 405 | class MicrophoneInput { |
| 406 | constructor(owner, updateCallback) { |
| 407 | this.owner = owner; |
| 408 | this.updateCallback = updateCallback; |
| 409 | this.mediaStream = null; |
| 410 | this.mediaRecorder = null; |
| 411 | this.audioContext = null; |
| 412 | this.mediaStreamSource = null; |
| 413 | this.analyserNode = null; |
| 414 | this.audioChunks = []; |
| 415 | this.lastChunk = null; |
| 416 | this.messageSent = false; |
| 417 | this.lastAudioTime = null; |
| 418 | this.waitingTimer = null; |
| 419 | this.silenceStartTime = null; |
| 420 | this.hasStartedRecording = false; |
| 421 | this.analysisFrame = null; |
| 422 | this.selectedDeviceId = ""; |
| 423 | this._status = Status.INACTIVE; |
| 424 | } |
| 425 | |
| 426 | get status() { |
| 427 | return this._status; |
| 428 | } |
| 429 | |
| 430 | set status(nextStatus) { |
| 431 | if (this._status === nextStatus) return; |
| 432 | |
| 433 | const previousStatus = this._status; |
| 434 | this._status = nextStatus; |
| 435 | this.handleStatusChange(previousStatus, nextStatus); |
| 436 | this.owner.notifyStatusChange(); |
| 437 | } |
| 438 | |
| 439 | async initialize() { |
| 440 | this.status = Status.ACTIVATING; |
| 441 | |
| 442 | try { |
| 443 | const selectedDevice = this.owner.getSelectedDevice(); |
| 444 | const stream = await navigator.mediaDevices.getUserMedia({ |
| 445 | audio: { |
| 446 | deviceId: |
| 447 | selectedDevice?.deviceId |
| 448 | ? { exact: selectedDevice.deviceId } |
| 449 | : undefined, |
| 450 | echoCancellation: true, |
| 451 | noiseSuppression: true, |
| 452 | channelCount: 1, |
| 453 | }, |
| 454 | }); |
| 455 | |
| 456 | this.selectedDeviceId = selectedDevice?.deviceId || ""; |
| 457 | this.mediaStream = stream; |
| 458 | this.mediaRecorder = new MediaRecorder(stream); |
| 459 | this.mediaRecorder.ondataavailable = (event) => { |
| 460 | if ( |
| 461 | event.data.size > 0 && |
| 462 | (this.status === Status.RECORDING || this.status === Status.WAITING) |
| 463 | ) { |
| 464 | if (this.lastChunk) { |
| 465 | this.audioChunks.push(this.lastChunk); |
| 466 | this.lastChunk = null; |
| 467 | } |
| 468 | this.audioChunks.push(event.data); |
| 469 | } else if (this.status === Status.LISTENING) { |
| 470 | this.lastChunk = event.data; |
| 471 | } |
| 472 | }; |
| 473 | |
| 474 | this.setupAudioAnalysis(stream); |
| 475 | return true; |
| 476 | } catch (error) { |
| 477 | console.error("[Whisper STT] Microphone initialization failed", error); |
| 478 | globalThis.toast?.( |
| 479 | "Failed to access the microphone. Please check browser permissions.", |
| 480 | "error", |
| 481 | ); |
| 482 | this.status = Status.INACTIVE; |
| 483 | this.dispose(); |
| 484 | return false; |
| 485 | } |
| 486 | } |
| 487 | |
| 488 | handleStatusChange(previousStatus, nextStatus) { |
| 489 | if (nextStatus !== Status.RECORDING) { |
| 490 | this.lastChunk = null; |
| 491 | } |
| 492 | |
| 493 | switch (nextStatus) { |
| 494 | case Status.INACTIVE: |
| 495 | this.handleInactiveState(); |
| 496 | break; |
| 497 | case Status.LISTENING: |
| 498 | this.handleListeningState(); |
| 499 | break; |
| 500 | case Status.RECORDING: |
| 501 | this.handleRecordingState(); |
| 502 | break; |
| 503 | case Status.WAITING: |
| 504 | this.handleWaitingState(); |
| 505 | break; |
| 506 | case Status.PROCESSING: |
| 507 | this.handleProcessingState(); |
| 508 | break; |
| 509 | } |
| 510 | } |
| 511 | |
| 512 | handleInactiveState() { |
| 513 | this.stopRecording(); |
| 514 | this.stopAudioAnalysis(); |
| 515 | clearTimeout(this.waitingTimer); |
| 516 | this.waitingTimer = null; |
| 517 | } |
| 518 | |
| 519 | handleListeningState() { |
| 520 | this.stopRecording(); |
| 521 | this.audioChunks = []; |
| 522 | this.hasStartedRecording = false; |
| 523 | this.silenceStartTime = null; |
| 524 | this.lastAudioTime = null; |
| 525 | this.messageSent = false; |
| 526 | this.startAudioAnalysis(); |
| 527 | } |
| 528 | |
| 529 | handleRecordingState() { |
| 530 | if (!this.mediaRecorder) return; |
| 531 | |
| 532 | if (!this.hasStartedRecording && this.mediaRecorder.state !== "recording") { |
| 533 | this.hasStartedRecording = true; |
| 534 | this.mediaRecorder.start(1000); |
| 535 | } |
| 536 | |
| 537 | clearTimeout(this.waitingTimer); |
| 538 | this.waitingTimer = null; |
| 539 | } |
| 540 | |
| 541 | handleWaitingState() { |
| 542 | clearTimeout(this.waitingTimer); |
| 543 | this.waitingTimer = setTimeout(() => { |
| 544 | if (this.status === Status.WAITING) { |
| 545 | this.status = Status.PROCESSING; |
| 546 | } |
| 547 | }, this.owner.config.waiting_timeout); |
| 548 | } |
| 549 | |
| 550 | handleProcessingState() { |
| 551 | this.stopRecording(); |
| 552 | void this.process(); |
| 553 | } |
| 554 | |
| 555 | setupAudioAnalysis(stream) { |
| 556 | this.audioContext = new (window.AudioContext || window.webkitAudioContext)(); |
| 557 | this.mediaStreamSource = this.audioContext.createMediaStreamSource(stream); |
| 558 | this.analyserNode = this.audioContext.createAnalyser(); |
| 559 | this.analyserNode.fftSize = 2048; |
| 560 | this.analyserNode.minDecibels = -90; |
| 561 | this.analyserNode.maxDecibels = -10; |
| 562 | this.analyserNode.smoothingTimeConstant = 0.85; |
| 563 | this.mediaStreamSource.connect(this.analyserNode); |
| 564 | } |
| 565 | |
| 566 | startAudioAnalysis() { |
| 567 | const analyzeFrame = () => { |
| 568 | if (this.status === Status.INACTIVE || !this.analyserNode) return; |
| 569 | |
| 570 | const dataArray = new Uint8Array(this.analyserNode.fftSize); |
| 571 | this.analyserNode.getByteTimeDomainData(dataArray); |
| 572 | |
| 573 | let sum = 0; |
| 574 | for (let index = 0; index < dataArray.length; index += 1) { |
| 575 | const amplitude = (dataArray[index] - 128) / 128; |
| 576 | sum += amplitude * amplitude; |
| 577 | } |
| 578 | |
| 579 | const rms = Math.sqrt(sum / dataArray.length); |
| 580 | const now = Date.now(); |
| 581 | const silenceThreshold = this.densify(this.owner.config.silence_threshold); |
| 582 | |
| 583 | if (rms > silenceThreshold) { |
| 584 | this.lastAudioTime = now; |
| 585 | this.silenceStartTime = null; |
| 586 | |
| 587 | if ( |
| 588 | (this.status === Status.LISTENING || this.status === Status.WAITING) && |
| 589 | !ttsService.isSpeaking() |
| 590 | ) { |
| 591 | this.status = Status.RECORDING; |
| 592 | } |
| 593 | } else if (this.status === Status.RECORDING) { |
| 594 | if (!this.silenceStartTime) { |
| 595 | this.silenceStartTime = now; |
| 596 | } |
| 597 | |
| 598 | const silenceDuration = now - this.silenceStartTime; |
| 599 | if (silenceDuration >= this.owner.config.silence_duration) { |
| 600 | this.status = Status.WAITING; |
| 601 | } |
| 602 | } |
| 603 | |
| 604 | this.analysisFrame = requestAnimationFrame(analyzeFrame); |
| 605 | }; |
| 606 | |
| 607 | this.stopAudioAnalysis(); |
| 608 | this.analysisFrame = requestAnimationFrame(analyzeFrame); |
| 609 | } |
| 610 | |
| 611 | stopAudioAnalysis() { |
| 612 | if (this.analysisFrame) { |
| 613 | cancelAnimationFrame(this.analysisFrame); |
| 614 | this.analysisFrame = null; |
| 615 | } |
| 616 | } |
| 617 | |
| 618 | stopRecording() { |
| 619 | if (this.mediaRecorder?.state === "recording") { |
| 620 | this.mediaRecorder.stop(); |
| 621 | this.hasStartedRecording = false; |
| 622 | } |
| 623 | } |
| 624 | |
| 625 | densify(value) { |
| 626 | return Math.exp(-5 * (1 - value)); |
| 627 | } |
| 628 | |
| 629 | async process() { |
| 630 | if (this.audioChunks.length === 0) { |
| 631 | if (this.status === Status.PROCESSING) { |
| 632 | this.status = Status.LISTENING; |
| 633 | } |
| 634 | return; |
| 635 | } |
| 636 | |
| 637 | const audioBlob = new Blob(this.audioChunks, { type: "audio/wav" }); |
| 638 | const audio = await this.convertBlobToBase64(audioBlob); |
| 639 | |
| 640 | try { |
| 641 | const result = await callJsonApi(`/plugins/${PLUGIN_NAME}/transcribe`, { |
| 642 | audio, |
| 643 | }); |
| 644 | const text = this.filterResult(result?.text || ""); |
| 645 | if (text) { |
| 646 | await this.updateCallback(text, true); |
| 647 | } |
| 648 | } catch (error) { |
| 649 | console.error("[Whisper STT] Transcription failed", error); |
| 650 | window.toastFetchError?.("Transcription error", error); |
| 651 | } finally { |
| 652 | this.audioChunks = []; |
| 653 | if (this.status === Status.PROCESSING) { |
| 654 | this.status = Status.LISTENING; |
| 655 | } |
| 656 | } |
| 657 | } |
| 658 | |
| 659 | convertBlobToBase64(audioBlob) { |
| 660 | return new Promise((resolve, reject) => { |
| 661 | const reader = new FileReader(); |
| 662 | reader.onloadend = () => { |
| 663 | const result = String(reader.result || ""); |
| 664 | resolve(result.split(",")[1] || ""); |
| 665 | }; |
| 666 | reader.onerror = (error) => reject(error); |
| 667 | reader.readAsDataURL(audioBlob); |
| 668 | }); |
| 669 | } |
| 670 | |
| 671 | filterResult(text) { |
| 672 | const normalized = String(text || "").trim(); |
| 673 | if (!normalized) return ""; |
| 674 | |
| 675 | const wrapped = |
| 676 | (normalized.startsWith("{") && normalized.endsWith("}")) || |
| 677 | (normalized.startsWith("(") && normalized.endsWith(")")) || |
| 678 | (normalized.startsWith("[") && normalized.endsWith("]")); |
| 679 | |
| 680 | if (wrapped) { |
| 681 | console.log(`[Whisper STT] Discarding transcription: ${normalized}`); |
| 682 | return ""; |
| 683 | } |
| 684 | |
| 685 | return normalized; |
| 686 | } |
| 687 | |
| 688 | async toggle() { |
| 689 | const hasPermission = await this.requestPermission(); |
| 690 | if (!hasPermission) return; |
| 691 | |
| 692 | if ( |
| 693 | this.status === Status.INACTIVE || |
| 694 | this.status === Status.ACTIVATING |
| 695 | ) { |
| 696 | this.status = Status.LISTENING; |
| 697 | } else { |
| 698 | this.owner.stop(); |
| 699 | } |
| 700 | } |
| 701 | |
| 702 | async requestPermission() { |
| 703 | return await this.owner.requestMicrophonePermission(); |
| 704 | } |
| 705 | |
| 706 | dispose() { |
| 707 | clearTimeout(this.waitingTimer); |
| 708 | this.waitingTimer = null; |
| 709 | this.stopAudioAnalysis(); |
| 710 | |
| 711 | try { |
| 712 | this.mediaRecorder?.stream?.getTracks?.().forEach((track) => track.stop()); |
| 713 | } catch (_error) { |
| 714 | // Ignore media cleanup failures. |
| 715 | } |
| 716 | |
| 717 | try { |
| 718 | this.mediaStream?.getTracks?.().forEach((track) => track.stop()); |
| 719 | } catch (_error) { |
| 720 | // Ignore media cleanup failures. |
| 721 | } |
| 722 | |
| 723 | try { |
| 724 | this.audioContext?.close?.(); |
| 725 | } catch (_error) { |
| 726 | // Ignore audio context cleanup failures. |
| 727 | } |
| 728 | |
| 729 | this.mediaStream = null; |
| 730 | this.mediaRecorder = null; |
| 731 | this.mediaStreamSource = null; |
| 732 | this.analyserNode = null; |
| 733 | this.audioContext = null; |
| 734 | this.audioChunks = []; |
| 735 | this.lastChunk = null; |
| 736 | this.hasStartedRecording = false; |
| 737 | } |
| 738 | } |
| 739 | |
| 740 | export const store = createStore("whisperStt", model); |