feat(transcription): add shared voice input support (#4232)

* feat(webui): add voice transcription input * feat(webui): render ANSI output in code blocks * refactor(webui): isolate voice recorder logic * refactor(transcription): keep websocket ingress thin * refactor(transcription): resolve channel audio settings on demand * style(webui): neutralize voice waveform color * feat(webui): add voice input tooltip * feat(webui): add voice input keyboard shortcut * fix(webui): distinguish voice shortcut platforms * fix(webui): place voice button after model selector * refactor(webui): share voice hold recording helpers * fix(desktop): allow microphone voice input * fix(webui): stabilize token usage month labels * feat(webui): show voice input on settings overview * fix(webui): label voice capability as recognition * fix(webui): align capability overview status * refactor(webui): isolate transcription socket handling * fix(webui): soften silent voice waveform * refactor(audio): clarify transcription service location * docs(transcription): clarify audio and provider boundaries * fix(exec): reduce session output polling flake
2026-06-09 01:08:49 +08:00
parent 06d454a225
commit 9c81280300
49 changed files with 3071 additions and 257 deletions
@@ -412,6 +412,61 @@ describe("NanobotClient", () => {
    );
  });

+  it("sends transcription requests and resolves transcription results outside chat dispatch", async () => {
+    const client = new NanobotClient({
+      url: "ws://test",
+      reconnect: false,
+      socketFactory: (url) => new FakeSocket(url) as unknown as WebSocket,
+    });
+    const handler = vi.fn();
+    client.onChat("chat-a", handler);
+    client.connect();
+    lastSocket().fakeOpen();
+
+    const promise = client.transcribeAudio("data:audio/webm;base64,AAAA", {
+      durationMs: 1234,
+      timeoutMs: 1_000,
+    });
+    const frame = JSON.parse(lastSocket().sent.at(-1) as string);
+    expect(frame).toMatchObject({
+      type: "transcribe_audio",
+      data_url: "data:audio/webm;base64,AAAA",
+      duration_ms: 1234,
+    });
+    expect(typeof frame.request_id).toBe("string");
+
+    lastSocket().fakeMessage({
+      event: "transcription_result",
+      request_id: frame.request_id,
+      text: "hello from voice",
+    });
+    await expect(promise).resolves.toBe("hello from voice");
+    expect(handler).not.toHaveBeenCalled();
+  });
+
+  it("rejects pending transcription requests on server errors and socket close", async () => {
+    const client = new NanobotClient({
+      url: "ws://test",
+      reconnect: false,
+      socketFactory: (url) => new FakeSocket(url) as unknown as WebSocket,
+    });
+    client.connect();
+    lastSocket().fakeOpen();
+
+    const errored = client.transcribeAudio("data:audio/webm;base64,AAAA", { timeoutMs: 1_000 });
+    const errorFrame = JSON.parse(lastSocket().sent.at(-1) as string);
+    lastSocket().fakeMessage({
+      event: "transcription_error",
+      request_id: errorFrame.request_id,
+      detail: "not_configured",
+    });
+    await expect(errored).rejects.toThrow("not_configured");
+
+    const dropped = client.transcribeAudio("data:audio/webm;base64,BBBB", { timeoutMs: 1_000 });
+    lastSocket().close();
+    await expect(dropped).rejects.toThrow("socket closed");
+  });
+
  it("queues sends while connecting and flushes on open", () => {
    const client = new NanobotClient({
      url: "ws://test",