speech: replace preprocess parsing with strict zod schemas

packages/server/src/server/speech/providers/local/config.ts

@@ -32,44 +32,29 @@ export type { LocalSpeechModelId, LocalSttModelId, LocalTtsModelId };
 
 const DEFAULT_LOCAL_MODELS_SUBDIR = path.join("models", "local-speech");
 
-const OptionalBooleanFlagSchema = z.preprocess((value) => {
+const BooleanStringSchema = z
-  if (typeof value === "boolean") {
+  .string()
-    return value;
+  .trim()
-  }
+  .toLowerCase()
-  if (typeof value !== "string") {
+  .pipe(z.enum(["1", "0", "true", "false", "yes", "no"]))
-    return value;
+  .transform((value) => value === "1" || value === "true" || value === "yes");
-  }
-  const normalized = value.trim().toLowerCase();
-  if (normalized === "1" || normalized === "true" || normalized === "yes") {
-    return true;
-  }
-  if (normalized === "0" || normalized === "false" || normalized === "no") {
-    return false;
-  }
-  return undefined;
-}, z.boolean().optional());
 
-const OptionalFiniteNumberSchema = z.preprocess((value) => {
+const OptionalBooleanFlagSchema = z
-  if (typeof value === "number") {
+  .union([z.boolean(), BooleanStringSchema])
-    return Number.isFinite(value) ? value : undefined;
+  .optional();
-  }
-  if (typeof value !== "string") {
-    return value;
-  }
-  const parsed = Number.parseFloat(value);
-  return Number.isFinite(parsed) ? parsed : undefined;
-}, z.number().optional());
 
-const OptionalIntegerSchema = z.preprocess((value) => {
+const NumberLikeSchema = z.union([
-  if (typeof value === "number") {
+  z.number(),
-    return Number.isInteger(value) ? value : undefined;
+  z.string().trim().min(1),
-  }
+]);
-  if (typeof value !== "string") {
+
-    return value;
+const OptionalFiniteNumberSchema = NumberLikeSchema
-  }
+  .pipe(z.coerce.number().finite())
-  const parsed = Number.parseInt(value, 10);
+  .optional();
-  return Number.isFinite(parsed) ? parsed : undefined;
+
-}, z.number().int().optional());
+const OptionalIntegerSchema = NumberLikeSchema
+  .pipe(z.coerce.number().int())
+  .optional();
 
 const LocalSpeechResolutionSchema = z.object({
   includeProviderConfig: z.boolean(),

packages/server/src/server/speech/providers/local/sherpa/model-catalog.ts

@@ -166,18 +166,20 @@ function buildAliasMap<T extends string>(modelIds: readonly T[]): Record<string,
 function createAliasedModelIdSchema<T extends string>(params: {
   modelIds: readonly T[];
   aliases: Record<string, T>;
-}): z.ZodType<T> {
+}): z.ZodType<T, z.ZodTypeDef, string> {
   const validIds = new Set(params.modelIds);
-  return z.preprocess((value) => {
+  return z
-    if (typeof value !== "string") {
+    .string()
-      return value;
+    .trim()
-    }
+    .toLowerCase()
-    const normalized = value.trim().toLowerCase();
+    .refine(
-    if (!normalized) {
+      (value): value is T =>
-      return value;
+        validIds.has(value as T) || Object.prototype.hasOwnProperty.call(params.aliases, value),
-    }
+      {
-    return params.aliases[normalized] ?? normalized;
+        message: "Invalid model id",
-  }, z.string().refine((value): value is T => validIds.has(value as T))) as z.ZodType<T>;
+      }
+    )
+    .transform((value) => params.aliases[value] ?? (value as T));
 }
 
 const STT_MODEL_ALIASES = buildAliasMap(LOCAL_STT_MODEL_IDS);

packages/server/src/server/speech/providers/openai/config.ts

@@ -26,43 +26,33 @@ const OpenAiTtsVoiceSchema = z.enum([
 
 const OpenAiTtsModelSchema = z.enum(["tts-1", "tts-1-hd"]);
 
-const OptionalFiniteNumberSchema = z.preprocess((value) => {
+const NumberLikeSchema = z.union([
-  if (typeof value === "number") {
+  z.number(),
-    return Number.isFinite(value) ? value : undefined;
+  z.string().trim().min(1),
-  }
+]);
-  if (typeof value !== "string") {
-    return value;
-  }
-  const parsed = Number.parseFloat(value);
-  return Number.isFinite(parsed) ? parsed : undefined;
-}, z.number().optional());
 
-const OptionalTrimmedStringSchema = z.preprocess((value) => {
+const OptionalFiniteNumberSchema = NumberLikeSchema
-  if (typeof value !== "string") {
+  .pipe(z.coerce.number().finite())
-    return value;
+  .optional();
-  }
+
-  const trimmed = value.trim();
+const OptionalTrimmedStringSchema = z.string().trim().min(1).optional();
-  return trimmed.length > 0 ? trimmed : undefined;
-}, z.string().optional());
 
 const OpenAiSpeechResolutionSchema = z.object({
   apiKey: OptionalTrimmedStringSchema,
   sttConfidenceThreshold: OptionalFiniteNumberSchema,
   sttModel: OptionalTrimmedStringSchema,
-  ttsVoice: z.preprocess((value) => {
+  ttsVoice: z
-    if (typeof value !== "string") {
+    .string()
-      return value;
+    .trim()
-    }
+    .toLowerCase()
-    const normalized = value.trim().toLowerCase();
+    .pipe(OpenAiTtsVoiceSchema)
-    return normalized.length > 0 ? normalized : undefined;
+    .default("alloy"),
-  }, OpenAiTtsVoiceSchema.default("alloy")),
+  ttsModel: z
-  ttsModel: z.preprocess((value) => {
+    .string()
-    if (typeof value !== "string") {
+    .trim()
-      return value;
+    .toLowerCase()
-    }
+    .pipe(OpenAiTtsModelSchema)
-    const normalized = value.trim().toLowerCase();
+    .default(DEFAULT_OPENAI_TTS_MODEL),
-    return normalized.length > 0 ? normalized : undefined;
-  }, OpenAiTtsModelSchema.default(DEFAULT_OPENAI_TTS_MODEL)),
   realtimeTranscriptionModel: OptionalTrimmedStringSchema.default(
     DEFAULT_OPENAI_REALTIME_TRANSCRIPTION_MODEL
   ),

packages/server/src/server/speech/speech-config-resolver.ts

@@ -9,13 +9,12 @@ import {
   type RequestedSpeechProviders,
 } from "./speech-types.js";
 
-const OptionalSpeechProviderSchema = z.preprocess((value) => {
+const OptionalSpeechProviderSchema = z
-  if (typeof value !== "string") {
+  .string()
-    return value;
+  .trim()
-  }
+  .toLowerCase()
-  const normalized = value.trim().toLowerCase();
+  .pipe(SpeechProviderIdSchema)
-  return normalized.length > 0 ? normalized : undefined;
+  .optional();
-}, SpeechProviderIdSchema.optional());
 
 const RequestedSpeechProvidersSchema = z.object({
   dictationSttProvider: OptionalSpeechProviderSchema.default("local"),