laurent/PrivateChat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 5 Wiki Activity

local multi language STT

Browse Source
This commit is contained in:
Laurent Dubertrand
2026-03-26 19:38:54 +01:00
parent cc14b4d1b7
commit f2bf70bc7d
18 changed files with 1334 additions and 517 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
server/dist/index.js vendored
View File

@@ -16,7 +16,6 @@ import { generateAuthenticationOptions, generateRegistrationOptions, verifyAuthe
import Fastify from 'fastify';
import { Redis } from 'ioredis';
import { z } from 'zod';
import { SpeechTranscriber } from './speech-transcriber.js';
dotenv.config({ path: fileURLToPath(new URL('../../.env', import.meta.url)) });
const projectRootPath = fileURLToPath(new URL('../../', import.meta.url));
const registerSchema = z.object({
@@ -90,12 +89,6 @@ const signalMessageSchema = z.discriminatedUnion('type', [
z.object({
type: z.literal('ping'),
}),
z.object({
type: z.literal('speech-transcription'),
requestId: z.string().uuid(),
mimeType: z.string().trim().min(1).max(128),
audioBase64: z.string().min(1).max(32_000_000),
}),
]);
const app = Fastify({ logger: true, trustProxy: true });
const approvalAdminUsername = 'ladparis';
@@ -106,9 +99,6 @@ const frontendDistPath = resolveProjectPath(process.env.PRIVATECHAT_WEB_DIST_DIR
const ollamaServerUrl = (process.env.PRIVATECHAT_OLLAMA_URL ?? 'http://192.168.1.19:11434').replace(/\/+$/, '');
const ollamaImageModel = process.env.PRIVATECHAT_OLLAMA_IMAGE_MODEL ?? 'x/z-image-turbo:latest';
const ollamaImageSize = process.env.PRIVATECHAT_OLLAMA_IMAGE_SIZE ?? '1024x1024';
const speechTranscriptionServiceUrl = process.env.PRIVATECHAT_TRANSCRIPTION_WS_URL ?? 'wss://whisper.dubertrand.fr';
const speechTranscriptionLanguage = process.env.PRIVATECHAT_TRANSCRIPTION_LANGUAGE ?? 'auto';
const speechTranscriptionTimeoutMs = Number(process.env.PRIVATECHAT_TRANSCRIPTION_TIMEOUT_MS ?? 120_000);
const sessionTtlSeconds = Number(process.env.SESSION_TTL_SECONDS ?? 60 * 60 * 12);
const webAuthnChallengeTtlSeconds = Number(process.env.WEBAUTHN_CHALLENGE_TTL_SECONDS ?? 5 * 60);
const allowedCorsOrigins = parseAllowedOrigins(process.env.CORS_ORIGIN);
@@ -121,11 +111,6 @@ const frontendIndexPath = path.join(frontendDistPath, 'index.html');
const hasFrontendBuild = fs.existsSync(frontendIndexPath);
const convertOfficeDocument = promisify(libreOffice.convertWithOptions);
const execFileAsync = promisify(execFile);
const speechTranscriber = new SpeechTranscriber({
serviceUrl: speechTranscriptionServiceUrl,
language: speechTranscriptionLanguage,
requestTimeoutMs: speechTranscriptionTimeoutMs,
}, app.log);
fs.mkdirSync(path.dirname(sqlitePath), { recursive: true });
fs.mkdirSync(path.dirname(masterKeyPath), { recursive: true });
const encryptionKey = deriveEncryptionKey(loadOrCreateMasterKey(masterKeyPath));
@@ -782,25 +767,6 @@ async function handleSocketMessage(userId, sessionId, socket, rawMessage) {
}
return;
}
if (parsed.type === 'speech-transcription') {
try {
const text = await transcribeAudioPayload(parsed.requestId, parsed.audioBase64, parsed.mimeType);
send(socket, {
type: 'speech-transcribed',
requestId: parsed.requestId,
text,
});
}
catch (error) {
app.log.warn({ err: error, userId }, 'Speech transcription failed');
send(socket, {
type: 'speech-transcription-error',
requestId: parsed.requestId,
message: error instanceof Error ? error.message : 'Speech transcription failed.',
});
}
return;
}
let delivered = 0;
const recipientSockets = socketsByUserId.get(parsed.to);
if (recipientSockets) {
@@ -1257,23 +1223,12 @@ function parseClientMessage(rawMessage) {
prompt: parsed.data.prompt,
};
}
if (parsed.data.type === 'speech-transcription') {
return {
type: 'speech-transcription',
requestId: parsed.data.requestId,
mimeType: parsed.data.mimeType,
audioBase64: parsed.data.audioBase64,
};
}
return {
type: 'signal',
to: parsed.data.to,
signal: normalizeSignal(parsed.data.signal),
};
}
async function transcribeAudioPayload(requestId, audioBase64, mimeType) {
return await speechTranscriber.transcribe(requestId, audioBase64, mimeType);
}
async function generateImageFromPrompt(prompt) {
const abortController = new AbortController();
const timeoutId = setTimeout(() => abortController.abort(), 120_000);

124
server/dist/speech-transcriber.js vendored
View File

@@ -1,124 +0,0 @@
import WebSocket from 'ws';
export class SpeechTranscriber {
options;
logger;
constructor(options, logger) {
this.options = options;
this.logger = logger;
}
async transcribe(requestId, audioBase64, mimeType) {
const audio = this.normalizeAudioPayload(audioBase64, mimeType);
return await new Promise((resolve, reject) => {
let settled = false;
const socket = new WebSocket(this.options.serviceUrl);
const finish = (handler) => {
if (settled) {
return;
}
settled = true;
clearTimeout(timeout);
socket.removeAllListeners();
if (socket.readyState === WebSocket.CONNECTING || socket.readyState === WebSocket.OPEN) {
socket.close();
}
handler();
};
const timeout = setTimeout(() => {
finish(() => {
reject(new Error(`The transcription service timed out after ${this.options.requestTimeoutMs}ms.`));
});
}, this.options.requestTimeoutMs);
socket.on('open', () => {
try {
socket.send(JSON.stringify({
type: 'transcribe',
id: requestId,
language: this.options.language,
audio,
}));
}
catch (error) {
finish(() => {
reject(error instanceof Error ? error : new Error('Could not send transcription request.'));
});
}
});
socket.on('message', (payload) => {
const event = this.parseEvent(payload);
if (!event) {
return;
}
if (event.id && event.id !== requestId) {
this.logger.warn({ requestId, event }, 'Ignored transcription event for another request');
return;
}
if (event.type === 'start') {
this.logger.info({ requestId, model: event.model, language: event.language }, 'Speech transcription started');
return;
}
if (event.type === 'delta') {
return;
}
if (event.type === 'done') {
finish(() => {
resolve(event.text.trim());
});
return;
}
finish(() => {
reject(new Error(event.message));
});
});
socket.on('error', (error) => {
finish(() => {
reject(error instanceof Error ? error : new Error('The transcription service connection failed.'));
});
});
socket.on('close', (code, reasonBuffer) => {
if (settled) {
return;
}
const reason = reasonBuffer.toString().trim();
const detail = reason
? `The transcription service closed the connection unexpectedly (code=${code}, reason=${reason}).`
: `The transcription service closed the connection unexpectedly (code=${code}).`;
finish(() => {
reject(new Error(detail));
});
});
});
}
normalizeAudioPayload(audioBase64, mimeType) {
const trimmedAudio = audioBase64.trim();
if (trimmedAudio.startsWith('data:')) {
return trimmedAudio;
}
const normalizedMimeType = mimeType.trim() || 'audio/webm';
return `data:${normalizedMimeType};base64,${trimmedAudio}`;
}
parseEvent(payload) {
const message = this.rawDataToString(payload).trim();
if (!message) {
return null;
}
try {
return JSON.parse(message);
}
catch {
this.logger.warn({ transcriptionPayload: message }, 'Ignored non-JSON transcription event');
return null;
}
}
rawDataToString(payload) {
if (typeof payload === 'string') {
return payload;
}
if (payload instanceof ArrayBuffer) {
return Buffer.from(payload).toString('utf8');
}
if (Array.isArray(payload)) {
return Buffer.concat(payload).toString('utf8');
}
return payload.toString('utf8');
}
}

70
server/src/index.ts
View File

@@ -26,8 +26,6 @@ import { Redis } from 'ioredis';
import type WebSocket from 'ws';
import { z } from 'zod';
import { SpeechTranscriber } from './speech-transcriber.js';
dotenv.config({ path: fileURLToPath(new URL('../../.env', import.meta.url)) });
const projectRootPath = fileURLToPath(new URL('../../', import.meta.url));
@@ -125,12 +123,6 @@ type ClientMessage =
}
| {
type: 'ping';
}
| {
type: 'speech-transcription';
requestId: string;
mimeType: string;
audioBase64: string;
};
type ServerMessage =
@@ -153,16 +145,6 @@ type ServerMessage =
peerId: string;
message: string;
}
| {
type: 'speech-transcribed';
requestId: string;
text: string;
}
| {
type: 'speech-transcription-error';
requestId: string;
message: string;
}
| { type: 'pong' }
| { type: 'error'; message: string };
@@ -316,12 +298,6 @@ const signalMessageSchema = z.discriminatedUnion('type', [
z.object({
type: z.literal('ping'),
}),
z.object({
type: z.literal('speech-transcription'),
requestId: z.string().uuid(),
mimeType: z.string().trim().min(1).max(128),
audioBase64: z.string().min(1).max(32_000_000),
}),
]);
const app = Fastify({ logger: true, trustProxy: true });
@@ -340,9 +316,6 @@ const frontendDistPath = resolveProjectPath(
const ollamaServerUrl = (process.env.PRIVATECHAT_OLLAMA_URL ?? 'http://192.168.1.19:11434').replace(/\/+$/, '');
const ollamaImageModel = process.env.PRIVATECHAT_OLLAMA_IMAGE_MODEL ?? 'x/z-image-turbo:latest';
const ollamaImageSize = process.env.PRIVATECHAT_OLLAMA_IMAGE_SIZE ?? '1024x1024';
const speechTranscriptionServiceUrl = process.env.PRIVATECHAT_TRANSCRIPTION_WS_URL ?? 'wss://whisper.dubertrand.fr';
const speechTranscriptionLanguage = process.env.PRIVATECHAT_TRANSCRIPTION_LANGUAGE ?? 'auto';
const speechTranscriptionTimeoutMs = Number(process.env.PRIVATECHAT_TRANSCRIPTION_TIMEOUT_MS ?? 120_000);
const sessionTtlSeconds = Number(process.env.SESSION_TTL_SECONDS ?? 60 * 60 * 12);
const webAuthnChallengeTtlSeconds = Number(process.env.WEBAUTHN_CHALLENGE_TTL_SECONDS ?? 5 * 60);
const allowedCorsOrigins = parseAllowedOrigins(process.env.CORS_ORIGIN);
@@ -358,15 +331,6 @@ const hasFrontendBuild = fs.existsSync(frontendIndexPath);
const convertOfficeDocument = promisify(libreOffice.convertWithOptions);
const execFileAsync = promisify(execFile);
const speechTranscriber = new SpeechTranscriber(
{
serviceUrl: speechTranscriptionServiceUrl,
language: speechTranscriptionLanguage,
requestTimeoutMs: speechTranscriptionTimeoutMs,
},
app.log,
);
fs.mkdirSync(path.dirname(sqlitePath), { recursive: true });
fs.mkdirSync(path.dirname(masterKeyPath), { recursive: true });
@@ -1225,27 +1189,6 @@ async function handleSocketMessage(
return;
}
if (parsed.type === 'speech-transcription') {
try {
const text = await transcribeAudioPayload(parsed.requestId, parsed.audioBase64, parsed.mimeType);
send(socket, {
type: 'speech-transcribed',
requestId: parsed.requestId,
text,
});
} catch (error) {
app.log.warn({ err: error, userId }, 'Speech transcription failed');
send(socket, {
type: 'speech-transcription-error',
requestId: parsed.requestId,
message: error instanceof Error ? error.message : 'Speech transcription failed.',
});
}
return;
}
let delivered = 0;
const recipientSockets = socketsByUserId.get(parsed.to);
@@ -1897,15 +1840,6 @@ function parseClientMessage(rawMessage: string): ClientMessage | null {
};
}
if (parsed.data.type === 'speech-transcription') {
return {
type: 'speech-transcription',
requestId: parsed.data.requestId,
mimeType: parsed.data.mimeType,
audioBase64: parsed.data.audioBase64,
};
}
return {
type: 'signal',
to: parsed.data.to,
@@ -1913,10 +1847,6 @@ function parseClientMessage(rawMessage: string): ClientMessage | null {
};
}
async function transcribeAudioPayload(requestId: string, audioBase64: string, mimeType: string): Promise<string> {
return await speechTranscriber.transcribe(requestId, audioBase64, mimeType);
}
async function generateImageFromPrompt(prompt: string): Promise<{ imageBase64: string; mimeType: string }> {
const abortController = new AbortController();
const timeoutId = setTimeout(() => abortController.abort(), 120_000);

173
server/src/speech-transcriber.ts
View File

@@ -1,173 +0,0 @@
import WebSocket, { type RawData } from 'ws';
type LoggerLike = {
info: (payload: unknown, message?: string) => void;
warn: (payload: unknown, message?: string) => void;
error: (payload: unknown, message?: string) => void;
};
type SpeechTranscriberOptions = {
serviceUrl: string;
language: string;
requestTimeoutMs: number;
};
type ServiceEvent =
| { type: 'start'; id: string; model: string; language: string }
| { type: 'delta'; id: string; text: string; fullText: string }
| { type: 'done'; id: string; text: string }
| { type: 'error'; id?: string; message: string };
export class SpeechTranscriber {
constructor(
private readonly options: SpeechTranscriberOptions,
private readonly logger: LoggerLike,
) {}
async transcribe(requestId: string, audioBase64: string, mimeType: string): Promise<string> {
const audio = this.normalizeAudioPayload(audioBase64, mimeType);
return await new Promise<string>((resolve, reject) => {
let settled = false;
const socket = new WebSocket(this.options.serviceUrl);
const finish = (handler: () => void): void => {
if (settled) {
return;
}
settled = true;
clearTimeout(timeout);
socket.removeAllListeners();
if (socket.readyState === WebSocket.CONNECTING || socket.readyState === WebSocket.OPEN) {
socket.close();
}
handler();
};
const timeout = setTimeout(() => {
finish(() => {
reject(new Error(`The transcription service timed out after ${this.options.requestTimeoutMs}ms.`));
});
}, this.options.requestTimeoutMs);
socket.on('open', () => {
try {
socket.send(
JSON.stringify({
type: 'transcribe',
id: requestId,
language: this.options.language,
audio,
}),
);
} catch (error) {
finish(() => {
reject(error instanceof Error ? error : new Error('Could not send transcription request.'));
});
}
});
socket.on('message', (payload) => {
const event = this.parseEvent(payload);
if (!event) {
return;
}
if (event.id && event.id !== requestId) {
this.logger.warn({ requestId, event }, 'Ignored transcription event for another request');
return;
}
if (event.type === 'start') {
this.logger.info(
{ requestId, model: event.model, language: event.language },
'Speech transcription started',
);
return;
}
if (event.type === 'delta') {
return;
}
if (event.type === 'done') {
finish(() => {
resolve(event.text.trim());
});
return;
}
finish(() => {
reject(new Error(event.message));
});
});
socket.on('error', (error) => {
finish(() => {
reject(error instanceof Error ? error : new Error('The transcription service connection failed.'));
});
});
socket.on('close', (code, reasonBuffer) => {
if (settled) {
return;
}
const reason = reasonBuffer.toString().trim();
const detail = reason
? `The transcription service closed the connection unexpectedly (code=${code}, reason=${reason}).`
: `The transcription service closed the connection unexpectedly (code=${code}).`;
finish(() => {
reject(new Error(detail));
});
});
});
}
private normalizeAudioPayload(audioBase64: string, mimeType: string): string {
const trimmedAudio = audioBase64.trim();
if (trimmedAudio.startsWith('data:')) {
return trimmedAudio;
}
const normalizedMimeType = mimeType.trim() || 'audio/webm';
return `data:${normalizedMimeType};base64,${trimmedAudio}`;
}
private parseEvent(payload: RawData): ServiceEvent | null {
const message = this.rawDataToString(payload).trim();
if (!message) {
return null;
}
try {
return JSON.parse(message) as ServiceEvent;
} catch {
this.logger.warn({ transcriptionPayload: message }, 'Ignored non-JSON transcription event');
return null;
}
}
private rawDataToString(payload: RawData): string {
if (typeof payload === 'string') {
return payload;
}
if (payload instanceof ArrayBuffer) {
return Buffer.from(payload).toString('utf8');
}
if (Array.isArray(payload)) {
return Buffer.concat(payload).toString('utf8');
}
return payload.toString('utf8');
}
}