Skip to content

armgit / cindy

Public Imported from GitHub
0

Consider it done. The open-source AI agent that works out of the box · 想到,就能做到。开源、开箱即用的 AI Agent。

Merge upstream/main and retain both speech route regressions

JiangJiahaoa29c6196a61b
diff --cc apps/desktop/src/renderer/features/bots/voice/__tests__/CompanionVoiceSettings.test.tsx
index 8dad6b08ca,d08f6dc289..4b5c38d9ef
--- a/apps/desktop/src/renderer/features/bots/voice/__tests__/CompanionVoiceSettings.test.tsx
+++ b/apps/desktop/src/renderer/features/bots/voice/__tests__/CompanionVoiceSettings.test.tsx
@@@ -435,51 -435,35 +435,80 @@@ it('reconciles the saved speed on mode
expect(saved).not.toHaveProperty('speed');
});
+// Unit regression at the actual settings entry: Gateway Qwen accepts 0.5–2,
+// although its transport is OpenAI-compatible. The old protocol-only range let
+// users save 4x and receive a Gateway error; retain the existing adjustment flow.
+it('clamps speed when switching to Gateway Qwen and previews the adjusted choice', async () => {
+ const openai = {
+ id: 'tts-1',
+ name: 'OpenAI speech',
+ providerId: 'xd',
+ protocol: 'openai' as const,
+ protocolLocked: true,
+ defaultVoice: 'alloy',
+ };
+ const qwen = {
+ ...openai,
+ id: 'qwen/qwen-audio-3.1-tts-plus',
+ name: 'Qwen speech',
+ defaultVoice: 'longanlingxin_v3.1',
+ };
+ models = [openai, qwen];
+ vi.mocked(window.electronAPI.voiceConversation.preference).mockResolvedValueOnce({
+ selection: { providerId: 'xd', modelId: openai.id, voice: 'alloy', speed: 4 },
+ customized: true,
+ });
+ renderSettings();
+ const model = await screen.findByRole('combobox', { name: 'bots.voiceMode.model' });
+ await waitFor(() => expect((model as HTMLSelectElement).disabled).toBe(false));
+ fireEvent.change(model, { target: { value: voiceModelKey(qwen) } });
+ await waitFor(() =>
+ expect(window.electronAPI.voiceConversation.setPreference).toHaveBeenLastCalledWith(
+ 'bot',
+ expect.objectContaining({ modelId: qwen.id, voice: qwen.defaultVoice, speed: 2 }),
+ ),
+ );
+ const slider = await screen.findByRole('slider', { name: 'bots.voiceMode.speed' });
+ expect(slider.getAttribute('aria-valuemin')).toBe('0.5');
+ expect(slider.getAttribute('aria-valuemax')).toBe('2');
+ expect(screen.getByText('bots.voiceMode.speedAdjusted')).toBeTruthy();
+ fireEvent.click(await screen.findByRole('button', { name: 'bots.voiceMode.preview' }));
+ await waitFor(() =>
+ expect(window.electronAPI.voiceConversation.preview).toHaveBeenCalledWith(
+ expect.objectContaining({ modelId: qwen.id, voice: qwen.defaultVoice, speed: 2 }),
+ ),
+ );
+});
+
+ // OpenRouter uses OpenAI transport, but does not promise OpenAI's speed range.
+ // Exercise real settings save/preview and call-start resolution with an old saved speed.
+ it('honors route speed limits when reopening, editing and previewing an OpenRouter voice', async () => {
+ const router: VoiceConversationModel = {
+ id: 'elevenlabs/eleven-v3', name: 'Eleven v3', providerId: 'router', agent: 'codex',
+ protocol: 'openai', protocolLocked: true, speedRange: null,
+ };
+ models = [router];
+ const saved = { providerId: 'router', agent: 'codex' as const, modelId: router.id, voice: 'sarah', speed: 1.5 };
+ vi.mocked(window.electronAPI.voiceConversation.preference).mockResolvedValue({ selection: saved, customized: true });
+ renderSettings();
+ await screen.findByText('bots.voiceMode.speedUnavailable');
+ expect(screen.queryByRole('slider')).toBeNull();
+ expect(resolveVoiceConfiguration(models, saved).selection).not.toHaveProperty('speed');
+ const voice = await screen.findByLabelText('bots.voiceMode.voiceId');
+ await waitFor(() => expect((voice as HTMLInputElement).disabled).toBe(false));
+ fireEvent.change(voice, { target: { value: 'custom-voice-id' } });
+ fireEvent.blur(voice);
+ await waitFor(() => expect(window.electronAPI.voiceConversation.setPreference).toHaveBeenCalledWith(
+ 'bot', { providerId: 'router', agent: 'codex', modelId: router.id, voice: 'custom-voice-id' },
+ ));
+ const previewButton = await screen.findByRole('button', { name: 'bots.voiceMode.preview' });
+ await waitFor(() => expect((previewButton as HTMLButtonElement).disabled).toBe(false));
+ fireEvent.click(previewButton);
+ await waitFor(() => expect(window.electronAPI.voiceConversation.preview).toHaveBeenCalled());
+ expect(vi.mocked(window.electronAPI.voiceConversation.preview).mock.calls[0][0]).toMatchObject({ voice: 'custom-voice-id', modelId: router.id });
+ expect(vi.mocked(window.electronAPI.voiceConversation.preview).mock.calls[0][0]).not.toHaveProperty('speed');
+ });
+
it('defaults to sentence playback, persists a playback-only change, and restores the default', async () => {
models = [{ id: 'tts', name: 'Speech', defaultVoice: 'alloy' }];
const view = renderSettings();
diff --cc apps/desktop/src/shared/speechProtocols.ts
index 9ed4c75d72,2d5511379b..ecc9e7db45
--- a/apps/desktop/src/shared/speechProtocols.ts
+++ b/apps/desktop/src/shared/speechProtocols.ts
@@@ -7,12 -7,18 +7,18 @@@ export const SPEECH_PROTOCOLS =
'dashscope-tts',
] as const;
export type SpeechProtocol = (typeof SPEECH_PROTOCOLS)[number];
+ export type SpeechSpeedRange = { min: number; max: number } | null;
-/** Native synthesis limits, shared by settings and Main. Compatible proxies must
- * implement their selected protocol; this is not a claim about gateway forwarding.
+/** Synthesis limits, shared by settings and Main. Compatible proxies must
+ * implement their selected protocol, subject to verified model-specific limits.
* Eleven v3 currently excludes speed: https://elevenlabs.io/docs/eleven-creative/playground/text-to-speech#speed
*/
- export function speechSpeedRange(modelId: string, protocol?: SpeechProtocol) {
+ export function speechSpeedRange(
+ modelId: string,
+ protocol?: SpeechProtocol,
+ routeRange?: SpeechSpeedRange,
+ ): SpeechSpeedRange {
+ if (routeRange !== undefined) return routeRange;
switch (protocol) {
case 'elevenlabs':
return /(?:^|\/)eleven_v3(?:_|$)/i.test(modelId) ? null : { min: 0.7, max: 1.2 };