From 1e3b76f26b079f076f996dac6d196ef7dc2f9479 Mon Sep 17 00:00:00 2001
From: Alix-007 <li.long15@xydigit.com>
Date: Tue, 21 Jul 2026 09:56:52 +0800
Subject: [PATCH] fix(microsoft): keep valid voices from malformed catalogs
 (#110784)

---
 extensions/microsoft/speech-provider.test.ts | 46 ++++++++++++++++
 extensions/microsoft/speech-provider.ts      | 57 +++++++++-----------
 2 files changed, 72 insertions(+), 31 deletions(-)

diff --git a/extensions/microsoft/speech-provider.test.ts b/extensions/microsoft/speech-provider.test.ts
index 729c4840eacf..1f340c5040bf 100644
--- a/extensions/microsoft/speech-provider.test.ts
+++ b/extensions/microsoft/speech-provider.test.ts
@@ -105,6 +105,52 @@ describe("listMicrosoftVoices", () => {
     );
   });
 
+  it("returns an empty catalog for a malformed top-level payload", async () => {
+    globalThis.fetch = vi
+      .fn()
+      .mockResolvedValue(
+        new Response("null", { status: 200 }),
+      ) as unknown as typeof globalThis.fetch;
+
+    await expect(listVoicesThroughProvider()).resolves.toEqual([]);
+  });
+
+  it("skips malformed rows without discarding valid voices", async () => {
+    globalThis.fetch = vi.fn().mockResolvedValue(
+      new Response(
+        JSON.stringify([
+          null,
+          "unexpected",
+          [],
+          { ShortName: 42 },
+          {
+            ShortName: "en-US-AvaNeural",
+            FriendlyName: "Microsoft Ava Online (Natural) - English (United States)",
+            Locale: "en-US",
+            Gender: "Female",
+            VoiceTag: {
+              ContentCategories: [null, "General"],
+              VoicePersonalities: [false, "Friendly", "Positive"],
+            },
+          },
+        ]),
+        { status: 200 },
+      ),
+    ) as unknown as typeof globalThis.fetch;
+
+    await expect(listVoicesThroughProvider()).resolves.toEqual([
+      {
+        id: "en-US-AvaNeural",
+        name: "Microsoft Ava Online (Natural) - English (United States)",
+        category: "General",
+        description: "Friendly, Positive",
+        locale: "en-US",
+        gender: "Female",
+        personalities: ["Friendly", "Positive"],
+      },
+    ]);
+  });
+
   it("throws on Microsoft voice list failures", async () => {
     globalThis.fetch = vi
       .fn()
diff --git a/extensions/microsoft/speech-provider.ts b/extensions/microsoft/speech-provider.ts
index 8f4f45f4a078..508b5d9bbe07 100644
--- a/extensions/microsoft/speech-provider.ts
+++ b/extensions/microsoft/speech-provider.ts
@@ -47,17 +47,6 @@ type MicrosoftProviderConfig = {
   timeoutMs?: number;
 };
 
-type MicrosoftVoiceListEntry = {
-  ShortName?: string;
-  FriendlyName?: string;
-  Locale?: string;
-  Gender?: string;
-  VoiceTag?: {
-    ContentCategories?: string[];
-    VoicePersonalities?: string[];
-  };
-};
-
 function normalizeMicrosoftProviderConfig(
   rawConfig: Record<string, unknown>,
 ): MicrosoftProviderConfig {
@@ -114,9 +103,10 @@ function buildMicrosoftVoiceHeaders(): Record<string, string> {
   };
 }
 
-function formatMicrosoftVoiceDescription(entry: MicrosoftVoiceListEntry): string | undefined {
-  const personalities = entry.VoiceTag?.VoicePersonalities?.filter(Boolean) ?? [];
-  return personalities.length > 0 ? personalities.join(", ") : undefined;
+function readMicrosoftVoiceTagStrings(value: unknown): string[] | undefined {
+  return Array.isArray(value)
+    ? value.filter((entry): entry is string => typeof entry === "string" && entry.trim().length > 0)
+    : undefined;
 }
 
 function isCjkDominant(text: string): boolean {
@@ -173,24 +163,29 @@ async function listMicrosoftVoices(
       });
     }
     await assertOkOrThrowProviderError(response, "Microsoft voices API error");
-    const voices = await readProviderJsonResponse<MicrosoftVoiceListEntry[]>(
-      response,
-      "microsoft.speech-voices",
-    );
+    const voices = await readProviderJsonResponse<unknown>(response, "microsoft.speech-voices");
     return Array.isArray(voices)
-      ? voices
-          .map((voice) => ({
-            id: voice.ShortName?.trim() ?? "",
-            name: trimToUndefined(voice.FriendlyName) ?? trimToUndefined(voice.ShortName),
-            category: voice.VoiceTag?.ContentCategories?.find((value) => value.trim().length > 0),
-            description: formatMicrosoftVoiceDescription(voice),
-            locale: trimToUndefined(voice.Locale),
-            gender: trimToUndefined(voice.Gender),
-            personalities: voice.VoiceTag?.VoicePersonalities?.filter(
-              (value): value is string => value.trim().length > 0,
-            ),
-          }))
-          .filter((voice) => voice.id.length > 0)
+      ? voices.flatMap((value) => {
+          const voice = asObject(value);
+          const id = trimToUndefined(voice?.ShortName);
+          if (!voice || !id) {
+            return [];
+          }
+          const voiceTag = asObject(voice.VoiceTag);
+          const categories = readMicrosoftVoiceTagStrings(voiceTag?.ContentCategories);
+          const personalities = readMicrosoftVoiceTagStrings(voiceTag?.VoicePersonalities);
+          return [
+            {
+              id,
+              name: trimToUndefined(voice.FriendlyName) ?? id,
+              category: categories?.[0],
+              description: personalities?.length ? personalities.join(", ") : undefined,
+              locale: trimToUndefined(voice.Locale),
+              gender: trimToUndefined(voice.Gender),
+              personalities,
+            },
+          ];
+        })
       : [];
   } finally {
     await release();