jb/paseo
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

simplify speech bootstrap and clean daemon startup locking

Browse Source
This commit is contained in:
Mohamed Boudra
2026-02-17 12:16:02 +07:00
parent efde557d6f
commit 34e35621f0
24 changed files with 68 additions and 136 deletions
Download Patch File Download Diff File Expand all files Collapse all files
packages/server/src/server/bootstrap.ts
+1 -11
View File
@@ -63,7 +63,7 @@ import type {
AgentProvider,
} from "./agent/agent-sdk-types.js";
import type { AgentProviderRuntimeSettingsMap } from "./agent/provider-launch-config.js";
import { acquirePidLock, releasePidLock, isLocked, PidLockError } from "./pid-lock.js";
import { acquirePidLock, releasePidLock } from "./pid-lock.js";
import { isHostAllowed, type AllowedHostsConfig } from "./allowed-hosts.js";
import {
createVoiceMcpSocketBridgeManager,
@@ -137,16 +137,6 @@ export async function createPaseoDaemon(
): Promise<PaseoDaemon> {
const logger = rootLogger.child({ module: "bootstrap" });
// Fail fast before expensive bootstrap (speech/runtime/model checks) if another daemon is active.
const lockState = await isLocked(config.paseoHome);
if (lockState.locked) {
const existingLock = lockState.info;
throw new PidLockError(
`Another Paseo daemon is already running (PID ${existingLock?.pid ?? "unknown"}, started ${existingLock?.startedAt ?? "unknown"})`,
existingLock
);
}
// Acquire PID lock before expensive bootstrap work so duplicate starts fail immediately.
await acquirePidLock(config.paseoHome, config.listen);
packages/server/src/server/daemon-e2e/connection-offer.e2e.test.ts
+2 -1
View File
@@ -186,7 +186,8 @@ describe("ConnectionOfferV2 (daemon E2E)", () => {
PASEO_HOME: tempHome,
PASEO_LISTEN: `0.0.0.0:${port}`,
OPENAI_API_KEY: "",
PASEO_LOCAL_AUTO_DOWNLOAD: "0",
PASEO_DICTATION_ENABLED: "0",
PASEO_VOICE_MODE_ENABLED: "0",
PASEO_LOG_FORMAT: "json",
};
packages/server/src/server/index.ts
-6
View File
@@ -1,9 +1,3 @@
import dotenv from "dotenv";
import { fileURLToPath } from "node:url";
import path from "node:path";
const __dirname = path.dirname(fileURLToPath(import.meta.url));
dotenv.config({ path: path.resolve(__dirname, "../../.env") });
import { createPaseoDaemon } from "./bootstrap.js";
import { loadConfig } from "./config.js";
import { resolvePaseoHome } from "./paseo-home.js";
packages/server/src/server/persisted-config.ts
+30 -2
View File
@@ -22,7 +22,6 @@ const ProviderCredentialsSchema = z
const LocalSpeechProviderSchema = z
.object({
modelsDir: z.string().min(1).optional(),
autoDownload: z.boolean().optional(),
})
.strict();
@@ -175,6 +174,34 @@ function getLogger(logger: LoggerLike | undefined): LoggerLike | undefined {
return logger?.child({ module: "config" });
}
function stripDeprecatedLocalSpeechConfigFields(parsed: unknown): unknown {
if (!parsed || typeof parsed !== "object" || Array.isArray(parsed)) {
return parsed;
}
const root = { ...(parsed as Record<string, unknown>) };
const providers = root.providers;
if (!providers || typeof providers !== "object" || Array.isArray(providers)) {
return root;
}
const providersRecord = { ...(providers as Record<string, unknown>) };
const local = providersRecord.local;
if (!local || typeof local !== "object" || Array.isArray(local)) {
root.providers = providersRecord;
return root;
}
const localRecord = { ...(local as Record<string, unknown>) };
if ("autoDownload" in localRecord) {
delete localRecord.autoDownload;
}
providersRecord.local = localRecord;
root.providers = providersRecord;
return root;
}
export function loadPersistedConfig(
paseoHome: string,
logger?: LoggerLike
@@ -209,7 +236,8 @@ export function loadPersistedConfig(
throw new Error(`[Config] Invalid JSON in ${configPath}: ${message}`);
}
const result = PersistedConfigSchema.safeParse(parsed);
const migrated = stripDeprecatedLocalSpeechConfigFields(parsed);
const result = PersistedConfigSchema.safeParse(migrated);
if (!result.success) {
const issues = result.error.issues
.map((i) => ` - ${i.path.join(".")}: ${i.message}`)
packages/server/src/server/session.ts
-1
View File
@@ -2879,7 +2879,6 @@ export class Session {
await ensureLocalSpeechModels({
modelsDir,
modelIds,
autoDownload: true,
logger: this.sessionLogger,
});
this.emit({
packages/server/src/server/speech/providers/local/config.ts
+1 -18
View File
@@ -24,7 +24,6 @@ export type LocalSpeechModelConfig = {
export type LocalSpeechProviderConfig = {
modelsDir: string;
autoDownload?: boolean;
models: LocalSpeechModelConfig;
};
@@ -36,17 +35,6 @@ export type { LocalSpeechModelId, LocalSttModelId, LocalTtsModelId };
const DEFAULT_LOCAL_MODELS_SUBDIR = path.join("models", "local-speech");
const BooleanStringSchema = z
.string()
.trim()
.toLowerCase()
.pipe(z.enum(["1", "0", "true", "false", "yes", "no"]))
.transform((value) => value === "1" || value === "true" || value === "yes");
const OptionalBooleanFlagSchema = z
.union([z.boolean(), BooleanStringSchema])
.optional();
const NumberLikeSchema = z.union([
z.number(),
z.string().trim().min(1),
@@ -63,7 +51,6 @@ const OptionalIntegerSchema = NumberLikeSchema
const LocalSpeechResolutionSchema = z.object({
includeProviderConfig: z.boolean(),
modelsDir: z.string().trim().min(1),
autoDownload: OptionalBooleanFlagSchema.default(true),
dictationLocalSttModel: LocalSttModelIdSchema.default(DEFAULT_LOCAL_STT_MODEL),
voiceLocalSttModel: LocalSttModelIdSchema.default(DEFAULT_LOCAL_STT_MODEL),
voiceLocalTtsModel: LocalTtsModelIdSchema.default(DEFAULT_LOCAL_TTS_MODEL),
@@ -98,7 +85,7 @@ function shouldIncludeLocalProviderConfig(params: {
return (
localRequestedByFeature ||
params.env.PASEO_LOCAL_MODELS_DIR !== undefined ||
params.persisted.providers?.local !== undefined
params.persisted.providers?.local?.modelsDir !== undefined
);
}
@@ -116,9 +103,6 @@ export function resolveLocalSpeechConfig(params: {
params.env.PASEO_LOCAL_MODELS_DIR ??
params.persisted.providers?.local?.modelsDir ??
path.join(params.paseoHome, DEFAULT_LOCAL_MODELS_SUBDIR),
autoDownload:
params.env.PASEO_LOCAL_AUTO_DOWNLOAD ??
params.persisted.providers?.local?.autoDownload,
dictationLocalSttModel:
params.env.PASEO_DICTATION_LOCAL_STT_MODEL ??
persistedLocalFeatureModel(
@@ -160,7 +144,6 @@ export function resolveLocalSpeechConfig(params: {
parsed.includeProviderConfig
? {
modelsDir: parsed.modelsDir,
autoDownload: parsed.autoDownload,
models: {
dictationStt: parsed.dictationLocalSttModel,
voiceStt: parsed.voiceLocalSttModel,
packages/server/src/server/speech/providers/local/models.ts
-2
View File
@@ -36,13 +36,11 @@ export function getLocalSpeechModelDir(modelsDir: string, modelId: LocalSpeechMo
export async function ensureLocalSpeechModels(options: {
modelsDir: string;
modelIds: LocalSpeechModelId[];
autoDownload?: boolean;
logger: import("pino").Logger;
}): Promise<Record<LocalSpeechModelId, string>> {
return ensureSherpaOnnxModels({
modelsDir: options.modelsDir,
modelIds: options.modelIds,
autoDownload: options.autoDownload ?? true,
logger: options.logger,
});
}
packages/server/src/server/speech/providers/local/runtime.ts
+6 -13
View File
@@ -36,7 +36,6 @@ type ResolvedLocalModels = {
type LocalSpeechAvailability = {
configured: boolean;
modelsDir: string | null;
autoDownload: boolean | null;
};
export type InitializedLocalSpeech = {
@@ -79,7 +78,6 @@ export function getLocalSpeechAvailability(
return {
configured: Boolean(localConfig),
modelsDir: localConfig?.modelsDir ?? null,
autoDownload: localConfig?.autoDownload ?? null,
};
}
@@ -186,12 +184,10 @@ export async function initializeLocalSpeechServices(params: {
providers: RequestedSpeechProviders;
speechConfig: PaseoSpeechConfig | null;
logger: Logger;
modelEnsureAutoDownload?: boolean;
}): Promise<InitializedLocalSpeech> {
const { providers, logger, speechConfig } = params;
const localConfig = speechConfig?.local ?? null;
const localModels = resolveConfiguredLocalModels(speechConfig);
const modelEnsureAutoDownload = params.modelEnsureAutoDownload ?? (localConfig?.autoDownload ?? true);
let sttService: SpeechToTextProvider | null = null;
let ttsService: TextToSpeechProvider | null = null;
@@ -209,28 +205,25 @@ export async function initializeLocalSpeechServices(params: {
{
modelsDir: localConfig.modelsDir,
modelIds: requiredLocalModelIds,
autoDownload: modelEnsureAutoDownload,
},
"Ensuring local speech models"
);
await ensureLocalSpeechModels({
modelsDir: localConfig.modelsDir,
modelIds: requiredLocalModelIds,
autoDownload: modelEnsureAutoDownload,
logger,
});
} catch (err) {
logger.error(
logger.warn(
{
err,
modelsDir: localConfig.modelsDir,
modelIds: requiredLocalModelIds,
autoDownload: modelEnsureAutoDownload,
hint:
"Use `paseo speech models` to inspect status and " +
"`paseo speech download --model <MODEL_ID>` to fetch missing models.",
},
"Failed to ensure local speech models"
"Local speech model bootstrap failed"
);
}
}
@@ -256,14 +249,14 @@ export async function initializeLocalSpeechServices(params: {
localSttEngines.set(modelId, created);
return created;
} catch (err) {
logger.error(
logger.warn(
{
err,
modelsDir: localConfig.modelsDir,
modelId,
hint: buildModelDownloadHint(modelId),
},
"Failed to initialize local STT engine (models missing or invalid)"
"Local STT engine unavailable"
);
return null;
}
@@ -340,14 +333,14 @@ export async function initializeLocalSpeechServices(params: {
}
ttsService = localVoiceTtsProvider;
} catch (err) {
logger.error(
logger.warn(
{
err,
modelsDir: localConfig.modelsDir,
modelId: localModels.voiceLocalTtsModel,
hint: buildModelDownloadHint(localModels.voiceLocalTtsModel),
},
"Failed to initialize local TTS engine (models missing or invalid)"
"Local TTS engine unavailable"
);
}
}
packages/server/src/server/speech/providers/local/sherpa/model-downloader.test.ts
-14
View File
@@ -35,25 +35,12 @@ describe("sherpa model downloader", () => {
const out = await ensureSherpaOnnxModel({
modelsDir,
modelId: "kitten-nano-en-v0_1-fp16",
autoDownload: false,
logger,
});
expect(out).toBe(modelDir);
});
test("ensureSherpaOnnxModel throws when missing and autoDownload is false", async () => {
const modelsDir = makeTmpDir();
await expect(
ensureSherpaOnnxModel({
modelsDir,
modelId: "zipformer-bilingual-zh-en-2023-02-20",
autoDownload: false,
logger,
})
).rejects.toThrow(/auto-download/i);
});
test("ensureSherpaOnnxModel logs lifecycle events without progress spam", async () => {
const modelsDir = makeTmpDir();
const infoMessages: string[] = [];
@@ -82,7 +69,6 @@ describe("sherpa model downloader", () => {
await ensureSherpaOnnxModel({
modelsDir,
modelId: "pocket-tts-onnx-int8",
autoDownload: true,
logger: loggerWithSpy,
});
} finally {
packages/server/src/server/speech/providers/local/sherpa/model-downloader.ts
-10
View File
@@ -11,7 +11,6 @@ import { getSherpaOnnxModelSpec, type SherpaOnnxModelId } from "./model-catalog.
export type EnsureSherpaOnnxModelOptions = {
modelsDir: string;
modelId: SherpaOnnxModelId;
autoDownload: boolean;
logger: pino.Logger;
};
@@ -104,13 +103,6 @@ export async function ensureSherpaOnnxModel(options: EnsureSherpaOnnxModelOption
return modelDir;
}
if (!options.autoDownload) {
throw new Error(
`Missing local speech model files for ${options.modelId} in ${modelDir}. ` +
`Set PASEO_LOCAL_AUTO_DOWNLOAD=1 to auto-download.`
);
}
logger.info({ modelsDir: options.modelsDir }, "Starting model download");
try {
@@ -207,7 +199,6 @@ export async function ensureSherpaOnnxModel(options: EnsureSherpaOnnxModelOption
export async function ensureSherpaOnnxModels(options: {
modelsDir: string;
modelIds: SherpaOnnxModelId[];
autoDownload: boolean;
logger: pino.Logger;
}): Promise<Record<SherpaOnnxModelId, string>> {
const uniq = Array.from(new Set(options.modelIds));
@@ -216,7 +207,6 @@ export async function ensureSherpaOnnxModels(options: {
out[id] = await ensureSherpaOnnxModel({
modelsDir: options.modelsDir,
modelId: id,
autoDownload: options.autoDownload,
logger: options.logger,
});
}
packages/server/src/server/speech/providers/local/sherpa/speech-download.e2e.test.ts
-2
View File
@@ -106,7 +106,6 @@ describe("speech models (download E2E)", () => {
await ensureSherpaOnnxModels({
modelsDir,
modelIds,
autoDownload: true,
logger,
});
@@ -121,7 +120,6 @@ describe("speech models (download E2E)", () => {
},
local: {
modelsDir,
autoDownload: false,
models: {
dictationStt:
set === "parakeet-pocket"
packages/server/src/server/speech/speech-config-resolver.test.ts
-3
View File
@@ -35,7 +35,6 @@ describe("resolveSpeechConfig", () => {
});
expect(result.speech.local).toEqual({
modelsDir: path.join(paseoHome, "models", "local-speech"),
autoDownload: true,
models: {
dictationStt: "parakeet-tdt-0.6b-v2-int8",
voiceStt: "parakeet-tdt-0.6b-v2-int8",
@@ -67,7 +66,6 @@ describe("resolveSpeechConfig", () => {
PASEO_VOICE_LOCAL_TTS_SPEAKER_ID: "5",
PASEO_VOICE_LOCAL_TTS_SPEED: "1.35",
PASEO_LOCAL_MODELS_DIR: "/tmp/models",
PASEO_LOCAL_AUTO_DOWNLOAD: "0",
OPENAI_API_KEY: "env-key",
PASEO_VOICE_STT_PROVIDER: "openai",
PASEO_DICTATION_STT_PROVIDER: "local",
@@ -82,7 +80,6 @@ describe("resolveSpeechConfig", () => {
expect(result.speech.local).toEqual({
modelsDir: "/tmp/models",
autoDownload: false,
models: {
dictationStt: "zipformer-bilingual-zh-en-2023-02-20",
voiceStt: "parakeet-tdt-0.6b-v3-int8",
packages/server/src/server/speech/speech-runtime.test.ts
-2
View File
@@ -79,7 +79,6 @@ describe("initializeSpeechRuntime readiness", () => {
availability: {
configured: false,
modelsDir: null,
autoDownload: null,
},
cleanup: () => {},
});
@@ -115,7 +114,6 @@ describe("initializeSpeechRuntime readiness", () => {
availability: {
configured: false,
modelsDir: null,
autoDownload: null,
},
cleanup: () => {},
});
packages/server/src/server/speech/speech-runtime.ts
+9 -15
View File
@@ -205,7 +205,6 @@ function buildVoiceFeatureReadiness(params: {
missingLocalModelIds: LocalSpeechModelId[];
backgroundDownloadInProgress: boolean;
backgroundDownloadError: string | null;
localAutoDownloadEnabled: boolean;
}): SpeechReadinessState {
const enabled = params.realtimeVoice.enabled || params.dictation.enabled;
if (!enabled) {
@@ -246,7 +245,7 @@ function buildVoiceFeatureReadiness(params: {
available: false,
reasonCode: "models_missing",
message: `Voice features are unavailable: missing local models (${joinModelIds(missingModelIds)}).`,
retryable: params.localAutoDownloadEnabled,
retryable: true,
missingModelIds,
};
}
@@ -328,7 +327,6 @@ export async function initializeSpeechRuntime(params: {
const openaiConfig = params.openaiConfig;
const providers = resolveRequestedSpeechProviders(speechConfig);
const requestedProviders = describeRequestedProviders(providers);
const localAutoDownloadEnabled = speechConfig?.local?.autoDownload ?? true;
validateOpenAiCredentialRequirements({
providers,
@@ -382,7 +380,6 @@ export async function initializeSpeechRuntime(params: {
missingLocalModelIds,
backgroundDownloadInProgress,
backgroundDownloadError,
localAutoDownloadEnabled,
});
return {
generatedAt: new Date().toISOString(),
@@ -459,12 +456,11 @@ export async function initializeSpeechRuntime(params: {
});
};
const reconcileServices = async (modelEnsureAutoDownload: boolean): Promise<void> => {
const reconcileServices = async (): Promise<void> => {
const nextLocalSpeech = await initializeLocalSpeechServices({
providers,
speechConfig,
logger,
modelEnsureAutoDownload,
});
const nextOpenAiSpeech = initializeOpenAiSpeechServices({
providers,
@@ -521,13 +517,13 @@ export async function initializeSpeechRuntime(params: {
}
};
const runReconcile = async (params: { modelEnsureAutoDownload: boolean }): Promise<void> => {
const runReconcile = async (): Promise<void> => {
if (reconcileInFlight) {
await reconcileInFlight;
publishReadinessIfChanged();
return;
}
reconcileInFlight = reconcileServices(params.modelEnsureAutoDownload).finally(() => {
reconcileInFlight = reconcileServices().finally(() => {
reconcileInFlight = null;
});
await reconcileInFlight;
@@ -550,7 +546,7 @@ export async function initializeSpeechRuntime(params: {
}
const modelsDir = localModelConfig?.modelsDir ?? null;
const modelIds = [...missingLocalModelIds];
if (!modelsDir || modelIds.length === 0 || !localAutoDownloadEnabled) {
if (!modelsDir || modelIds.length === 0) {
return;
}
@@ -571,10 +567,9 @@ export async function initializeSpeechRuntime(params: {
await ensureLocalSpeechModels({
modelsDir,
modelIds,
autoDownload: true,
logger,
});
await runReconcile({ modelEnsureAutoDownload: false });
await runReconcile();
backgroundDownloadError = null;
} catch (error) {
backgroundDownloadError = error instanceof Error ? error.message : String(error);
@@ -610,12 +605,11 @@ export async function initializeSpeechRuntime(params: {
missingLocalModelIds.length === 0 &&
!backgroundDownloadInProgress
) {
await runReconcile({ modelEnsureAutoDownload: false });
await runReconcile();
}
if (
missingLocalModelIds.length > 0 &&
localAutoDownloadEnabled &&
!backgroundDownloadInProgress &&
!backgroundDownloadError
) {
@@ -629,10 +623,10 @@ export async function initializeSpeechRuntime(params: {
}
};
await runReconcile({ modelEnsureAutoDownload: false });
await runReconcile();
const snapshot = computeReadinessSnapshot();
if (snapshot.voiceFeature.enabled && !snapshot.voiceFeature.available) {
if (missingLocalModelIds.length > 0 && localAutoDownloadEnabled) {
if (missingLocalModelIds.length > 0) {
startBackgroundDownload();
}
scheduleMonitor();