@agent-native/toolkit 0.9.0 → 0.10.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (190) hide show
  1. package/README.md +12 -3
  2. package/agent-native.eject.json +71 -0
  3. package/dist/chat-history/ChatHistoryRail.d.ts +33 -0
  4. package/dist/chat-history/ChatHistoryRail.d.ts.map +1 -0
  5. package/dist/chat-history/ChatHistoryRail.js +35 -0
  6. package/dist/chat-history/ChatHistoryRail.js.map +1 -0
  7. package/dist/chat-history/ChatHistoryRail.spec.d.ts +2 -0
  8. package/dist/chat-history/ChatHistoryRail.spec.d.ts.map +1 -0
  9. package/dist/chat-history/ChatHistoryRail.spec.js +157 -0
  10. package/dist/chat-history/ChatHistoryRail.spec.js.map +1 -0
  11. package/dist/chat-history/index.d.ts +2 -0
  12. package/dist/chat-history/index.d.ts.map +1 -1
  13. package/dist/chat-history/index.js +2 -0
  14. package/dist/chat-history/index.js.map +1 -1
  15. package/dist/chat-history/useChatHistoryRailController.d.ts +26 -0
  16. package/dist/chat-history/useChatHistoryRailController.d.ts.map +1 -0
  17. package/dist/chat-history/useChatHistoryRailController.js +36 -0
  18. package/dist/chat-history/useChatHistoryRailController.js.map +1 -0
  19. package/dist/chat-history/useChatHistoryRailController.spec.d.ts +2 -0
  20. package/dist/chat-history/useChatHistoryRailController.spec.d.ts.map +1 -0
  21. package/dist/chat-history/useChatHistoryRailController.spec.js +101 -0
  22. package/dist/chat-history/useChatHistoryRailController.spec.js.map +1 -0
  23. package/dist/chat-history.css +61 -0
  24. package/dist/composer/ComposerPlusMenu.js +1 -1
  25. package/dist/composer/ComposerPlusMenu.js.map +1 -1
  26. package/dist/composer/PromptComposer.d.ts.map +1 -1
  27. package/dist/composer/PromptComposer.js +22 -4
  28. package/dist/composer/PromptComposer.js.map +1 -1
  29. package/dist/composer/RealtimeVoiceMode.d.ts +5 -1
  30. package/dist/composer/RealtimeVoiceMode.d.ts.map +1 -1
  31. package/dist/composer/RealtimeVoiceMode.js +27 -8
  32. package/dist/composer/RealtimeVoiceMode.js.map +1 -1
  33. package/dist/composer/RealtimeVoiceMode.spec.js +51 -11
  34. package/dist/composer/RealtimeVoiceMode.spec.js.map +1 -1
  35. package/dist/composer/VoiceButton.d.ts +2 -0
  36. package/dist/composer/VoiceButton.d.ts.map +1 -1
  37. package/dist/composer/VoiceButton.js +33 -1
  38. package/dist/composer/VoiceButton.js.map +1 -1
  39. package/dist/composer/VoiceButton.spec.js +9 -1
  40. package/dist/composer/VoiceButton.spec.js.map +1 -1
  41. package/dist/composer/realtime-voice-audio-level.d.ts.map +1 -1
  42. package/dist/composer/realtime-voice-audio-level.js +2 -1
  43. package/dist/composer/realtime-voice-audio-level.js.map +1 -1
  44. package/dist/composer/realtime-voice-audio-level.spec.js +7 -0
  45. package/dist/composer/realtime-voice-audio-level.spec.js.map +1 -1
  46. package/dist/composer/useRealtimeVoiceMode.d.ts +7 -0
  47. package/dist/composer/useRealtimeVoiceMode.d.ts.map +1 -1
  48. package/dist/composer/useRealtimeVoiceMode.js +95 -23
  49. package/dist/composer/useRealtimeVoiceMode.js.map +1 -1
  50. package/dist/composer/useRealtimeVoiceMode.spec.js +58 -9
  51. package/dist/composer/useRealtimeVoiceMode.spec.js.map +1 -1
  52. package/dist/conformance/__fixtures__/css-in-js-adapter.d.ts +7 -0
  53. package/dist/conformance/__fixtures__/css-in-js-adapter.d.ts.map +1 -0
  54. package/dist/conformance/__fixtures__/css-in-js-adapter.js +84 -0
  55. package/dist/conformance/__fixtures__/css-in-js-adapter.js.map +1 -0
  56. package/dist/conformance/conformance.spec.d.ts +2 -0
  57. package/dist/conformance/conformance.spec.d.ts.map +1 -0
  58. package/dist/conformance/conformance.spec.js +58 -0
  59. package/dist/conformance/conformance.spec.js.map +1 -0
  60. package/dist/conformance/index.d.ts +4 -0
  61. package/dist/conformance/index.d.ts.map +1 -0
  62. package/dist/conformance/index.js +4 -0
  63. package/dist/conformance/index.js.map +1 -0
  64. package/dist/conformance/runner.d.ts +12 -0
  65. package/dist/conformance/runner.d.ts.map +1 -0
  66. package/dist/conformance/runner.js +441 -0
  67. package/dist/conformance/runner.js.map +1 -0
  68. package/dist/conformance/types.d.ts +26 -0
  69. package/dist/conformance/types.d.ts.map +1 -0
  70. package/dist/conformance/types.js +13 -0
  71. package/dist/conformance/types.js.map +1 -0
  72. package/dist/conformance/version.d.ts +6 -0
  73. package/dist/conformance/version.d.ts.map +1 -0
  74. package/dist/conformance/version.js +14 -0
  75. package/dist/conformance/version.js.map +1 -0
  76. package/dist/design-system/components.d.ts +64 -0
  77. package/dist/design-system/components.d.ts.map +1 -0
  78. package/dist/design-system/components.js +50 -0
  79. package/dist/design-system/components.js.map +1 -0
  80. package/dist/design-system/context.d.ts +11 -0
  81. package/dist/design-system/context.d.ts.map +1 -0
  82. package/dist/design-system/context.js +17 -0
  83. package/dist/design-system/context.js.map +1 -0
  84. package/dist/design-system/default-adapter.d.ts +3 -0
  85. package/dist/design-system/default-adapter.d.ts.map +1 -0
  86. package/dist/design-system/default-adapter.js +166 -0
  87. package/dist/design-system/default-adapter.js.map +1 -0
  88. package/dist/design-system/default-adapter.spec.d.ts +2 -0
  89. package/dist/design-system/default-adapter.spec.d.ts.map +1 -0
  90. package/dist/design-system/default-adapter.spec.js +96 -0
  91. package/dist/design-system/default-adapter.spec.js.map +1 -0
  92. package/dist/design-system/definition.d.ts +9 -0
  93. package/dist/design-system/definition.d.ts.map +1 -0
  94. package/dist/design-system/definition.js +4 -0
  95. package/dist/design-system/definition.js.map +1 -0
  96. package/dist/design-system/design-system.spec.d.ts +2 -0
  97. package/dist/design-system/design-system.spec.d.ts.map +1 -0
  98. package/dist/design-system/design-system.spec.js +121 -0
  99. package/dist/design-system/design-system.spec.js.map +1 -0
  100. package/dist/design-system/error-boundary.d.ts +18 -0
  101. package/dist/design-system/error-boundary.d.ts.map +1 -0
  102. package/dist/design-system/error-boundary.js +16 -0
  103. package/dist/design-system/error-boundary.js.map +1 -0
  104. package/dist/design-system/index.d.ts +8 -0
  105. package/dist/design-system/index.d.ts.map +1 -0
  106. package/dist/design-system/index.js +8 -0
  107. package/dist/design-system/index.js.map +1 -0
  108. package/dist/design-system/theme.d.ts +30 -0
  109. package/dist/design-system/theme.d.ts.map +1 -0
  110. package/dist/design-system/theme.js +95 -0
  111. package/dist/design-system/theme.js.map +1 -0
  112. package/dist/design-system/theme.spec.d.ts +2 -0
  113. package/dist/design-system/theme.spec.d.ts.map +1 -0
  114. package/dist/design-system/theme.spec.js +43 -0
  115. package/dist/design-system/theme.spec.js.map +1 -0
  116. package/dist/design-system/types.d.ts +288 -0
  117. package/dist/design-system/types.d.ts.map +1 -0
  118. package/dist/design-system/types.js +2 -0
  119. package/dist/design-system/types.js.map +1 -0
  120. package/dist/provider.d.ts +4 -3
  121. package/dist/provider.d.ts.map +1 -1
  122. package/dist/provider.js +46 -4
  123. package/dist/provider.js.map +1 -1
  124. package/dist/styles.css +35 -0
  125. package/dist/ui/button.d.ts +5 -0
  126. package/dist/ui/button.d.ts.map +1 -1
  127. package/dist/ui/button.js +25 -4
  128. package/dist/ui/button.js.map +1 -1
  129. package/dist/ui/dialog.d.ts +4 -0
  130. package/dist/ui/dialog.d.ts.map +1 -1
  131. package/dist/ui/dialog.js +2 -2
  132. package/dist/ui/dialog.js.map +1 -1
  133. package/dist/ui/popover.d.ts +1 -0
  134. package/dist/ui/popover.d.ts.map +1 -1
  135. package/dist/ui/popover.js +2 -2
  136. package/dist/ui/popover.js.map +1 -1
  137. package/dist/ui/select.d.ts +3 -1
  138. package/dist/ui/select.d.ts.map +1 -1
  139. package/dist/ui/select.js +1 -1
  140. package/dist/ui/select.js.map +1 -1
  141. package/dist/ui/sidebar.d.ts +6 -2
  142. package/dist/ui/sidebar.d.ts.map +1 -1
  143. package/dist/ui/sidebar.js +12 -15
  144. package/dist/ui/sidebar.js.map +1 -1
  145. package/dist/ui/tooltip.d.ts +3 -1
  146. package/dist/ui/tooltip.d.ts.map +1 -1
  147. package/dist/ui/tooltip.js +2 -2
  148. package/dist/ui/tooltip.js.map +1 -1
  149. package/package.json +31 -1
  150. package/src/chat-history/ChatHistoryRail.spec.tsx +312 -0
  151. package/src/chat-history/ChatHistoryRail.tsx +150 -0
  152. package/src/chat-history/index.ts +14 -0
  153. package/src/chat-history/useChatHistoryRailController.spec.tsx +131 -0
  154. package/src/chat-history/useChatHistoryRailController.ts +76 -0
  155. package/src/chat-history.css +61 -0
  156. package/src/composer/ComposerPlusMenu.tsx +11 -8
  157. package/src/composer/PromptComposer.tsx +27 -6
  158. package/src/composer/RealtimeVoiceMode.spec.tsx +116 -11
  159. package/src/composer/RealtimeVoiceMode.tsx +106 -36
  160. package/src/composer/VoiceButton.spec.ts +17 -1
  161. package/src/composer/VoiceButton.tsx +49 -0
  162. package/src/composer/realtime-voice-audio-level.spec.ts +15 -0
  163. package/src/composer/realtime-voice-audio-level.ts +2 -1
  164. package/src/composer/useRealtimeVoiceMode.spec.ts +82 -8
  165. package/src/composer/useRealtimeVoiceMode.tsx +112 -25
  166. package/src/conformance/__fixtures__/css-in-js-adapter.tsx +401 -0
  167. package/src/conformance/conformance.spec.tsx +83 -0
  168. package/src/conformance/index.ts +3 -0
  169. package/src/conformance/runner.tsx +710 -0
  170. package/src/conformance/types.ts +45 -0
  171. package/src/conformance/version.ts +22 -0
  172. package/src/design-system/components.tsx +146 -0
  173. package/src/design-system/context.tsx +31 -0
  174. package/src/design-system/default-adapter.spec.tsx +143 -0
  175. package/src/design-system/default-adapter.tsx +915 -0
  176. package/src/design-system/definition.ts +14 -0
  177. package/src/design-system/design-system.spec.tsx +183 -0
  178. package/src/design-system/error-boundary.tsx +37 -0
  179. package/src/design-system/index.ts +7 -0
  180. package/src/design-system/theme.spec.ts +61 -0
  181. package/src/design-system/theme.ts +145 -0
  182. package/src/design-system/types.ts +363 -0
  183. package/src/provider.tsx +98 -5
  184. package/src/styles.css +35 -0
  185. package/src/ui/button.tsx +95 -6
  186. package/src/ui/dialog.tsx +48 -23
  187. package/src/ui/popover.tsx +16 -2
  188. package/src/ui/select.tsx +7 -3
  189. package/src/ui/sidebar.tsx +47 -48
  190. package/src/ui/tooltip.tsx +7 -3
@@ -1,4 +1,5 @@
1
1
  import { Button } from "@agent-native/toolkit/ui/button";
2
+ import { Checkbox } from "@agent-native/toolkit/ui/checkbox";
2
3
  import {
3
4
  Select,
4
5
  SelectContent,
@@ -62,6 +63,7 @@ export interface RealtimeVoiceModeCopy {
62
63
  startWithOpenAiKey: string;
63
64
  startVoiceMode: string;
64
65
  keepDictating: string;
66
+ rememberPreference: string;
65
67
  showChat: string;
66
68
  hideChat: string;
67
69
  endVoiceMode: string;
@@ -111,6 +113,8 @@ export interface RealtimeVoiceModeEntryProps {
111
113
  onOpenChange?: (open: boolean) => void;
112
114
  onStartVoiceMode: () => void;
113
115
  onKeepDictating: () => void;
116
+ preferredMode?: RealtimeVoiceInputMode | null;
117
+ onRememberPreference?: (mode: RealtimeVoiceInputMode) => void;
114
118
  setupRequired?: boolean;
115
119
  openAiConfigured?: boolean;
116
120
  connectingBuilder?: boolean;
@@ -119,6 +123,8 @@ export interface RealtimeVoiceModeEntryProps {
119
123
  className?: string;
120
124
  }
121
125
 
126
+ export type RealtimeVoiceInputMode = "realtime" | "dictation";
127
+
122
128
  /**
123
129
  * Composer mic entry point for apps that support a full-duplex voice session.
124
130
  * The first click offers voice mode without silently changing the existing
@@ -132,6 +138,8 @@ export function RealtimeVoiceModeEntry({
132
138
  onOpenChange,
133
139
  onStartVoiceMode,
134
140
  onKeepDictating,
141
+ preferredMode = null,
142
+ onRememberPreference,
135
143
  setupRequired = false,
136
144
  openAiConfigured = false,
137
145
  connectingBuilder = false,
@@ -143,6 +151,8 @@ export function RealtimeVoiceModeEntry({
143
151
  const open = controlledOpen ?? uncontrolledOpen;
144
152
  const titleId = useId();
145
153
  const descriptionId = useId();
154
+ const rememberPreferenceId = useId();
155
+ const [rememberPreference, setRememberPreference] = useState(false);
146
156
  const [collisionBoundary, setCollisionBoundary] =
147
157
  useState<HTMLElement | null>(null);
148
158
 
@@ -159,8 +169,9 @@ export function RealtimeVoiceModeEntry({
159
169
  onOpenChange?.(nextOpen);
160
170
  };
161
171
 
162
- const choose = (callback: () => void) => {
172
+ const choose = (mode: RealtimeVoiceInputMode, callback: () => void) => {
163
173
  setOpen(false);
174
+ if (rememberPreference) onRememberPreference?.(mode);
164
175
  callback();
165
176
  };
166
177
 
@@ -177,6 +188,23 @@ export function RealtimeVoiceModeEntry({
177
188
  disabled={disabled}
178
189
  aria-label={copy.entryButtonLabel}
179
190
  aria-expanded={open}
191
+ onClick={(event) => {
192
+ if (!preferredMode) return;
193
+ if (
194
+ preferredMode === "realtime" &&
195
+ (providerStatusPending || setupRequired)
196
+ ) {
197
+ return;
198
+ }
199
+ event.preventDefault();
200
+ event.stopPropagation();
201
+ choose(
202
+ preferredMode,
203
+ preferredMode === "realtime"
204
+ ? onStartVoiceMode
205
+ : onKeepDictating,
206
+ );
207
+ }}
180
208
  className={cn(
181
209
  "size-7 shrink-0 text-muted-foreground hover:text-foreground",
182
210
  className,
@@ -197,43 +225,54 @@ export function RealtimeVoiceModeEntry({
197
225
  collisionPadding={16}
198
226
  data-collision-boundary={collisionBoundary ? "agent-panel" : "viewport"}
199
227
  className={cn(
200
- "p-4",
228
+ setupRequired ? "p-4" : "p-3",
201
229
  setupRequired
202
230
  ? "w-[min(calc(100vw-2rem),var(--radix-popover-content-available-width,30rem),30rem)]"
203
- : "w-[min(calc(100vw-2rem),var(--radix-popover-content-available-width,22rem),22rem)]",
231
+ : "w-[min(calc(100vw-2rem),var(--radix-popover-content-available-width,18rem),18rem)]",
204
232
  )}
205
233
  aria-labelledby={titleId}
206
- aria-describedby={descriptionId}
234
+ aria-describedby={setupRequired ? descriptionId : undefined}
207
235
  >
208
- <div className="grid gap-3">
236
+ <div className={cn("grid", setupRequired ? "gap-3" : "gap-2.5")}>
209
237
  <div className="grid gap-1">
210
238
  <h2 id={titleId} className="text-sm font-semibold text-foreground">
211
239
  {setupRequired ? copy.setupTitle : copy.promptTitle}
212
240
  </h2>
213
- <p
214
- id={descriptionId}
215
- className="text-sm leading-relaxed text-muted-foreground"
216
- >
217
- {setupRequired ? copy.setupDescription : copy.promptDescription}
218
- </p>
241
+ {setupRequired ? (
242
+ <p
243
+ id={descriptionId}
244
+ className="text-sm leading-relaxed text-muted-foreground"
245
+ >
246
+ {copy.setupDescription}
247
+ </p>
248
+ ) : null}
219
249
  </div>
220
250
 
221
- <div className="flex flex-col-reverse gap-2 sm:flex-row sm:flex-wrap sm:justify-end">
222
- <Button
223
- type="button"
224
- variant="ghost"
225
- size="sm"
226
- onClick={() => choose(onKeepDictating)}
227
- >
228
- {copy.keepDictating}
229
- </Button>
251
+ <div
252
+ className={cn(
253
+ "gap-2",
254
+ setupRequired
255
+ ? "flex flex-col-reverse sm:flex-row sm:flex-wrap sm:justify-end"
256
+ : "grid",
257
+ )}
258
+ >
230
259
  {setupRequired ? (
231
260
  <>
261
+ <Button
262
+ type="button"
263
+ variant="ghost"
264
+ size="sm"
265
+ onClick={() => choose("dictation", onKeepDictating)}
266
+ >
267
+ {copy.keepDictating}
268
+ </Button>
232
269
  <Button
233
270
  type="button"
234
271
  variant="outline"
235
272
  size="sm"
236
- onClick={() => choose(onUseOpenAiKey ?? onStartVoiceMode)}
273
+ onClick={() =>
274
+ choose("realtime", onUseOpenAiKey ?? onStartVoiceMode)
275
+ }
237
276
  >
238
277
  {openAiConfigured
239
278
  ? copy.startWithOpenAiKey
@@ -244,7 +283,9 @@ export function RealtimeVoiceModeEntry({
244
283
  size="sm"
245
284
  className="whitespace-nowrap"
246
285
  disabled={connectingBuilder}
247
- onClick={() => choose(onConnectBuilder ?? onStartVoiceMode)}
286
+ onClick={() =>
287
+ choose("realtime", onConnectBuilder ?? onStartVoiceMode)
288
+ }
248
289
  >
249
290
  {connectingBuilder ? (
250
291
  <IconLoader2 className="animate-spin" />
@@ -253,21 +294,49 @@ export function RealtimeVoiceModeEntry({
253
294
  </Button>
254
295
  </>
255
296
  ) : (
256
- <Button
257
- type="button"
258
- size="sm"
259
- disabled={providerStatusPending}
260
- onClick={() => choose(onStartVoiceMode)}
261
- >
262
- {providerStatusPending ? (
263
- <IconLoader2 className="animate-spin" />
264
- ) : (
265
- <IconMicrophone />
266
- )}
267
- {copy.startVoiceMode}
268
- </Button>
297
+ <>
298
+ <Button
299
+ type="button"
300
+ size="sm"
301
+ className="w-full"
302
+ disabled={providerStatusPending}
303
+ onClick={() => choose("realtime", onStartVoiceMode)}
304
+ >
305
+ {providerStatusPending ? (
306
+ <IconLoader2 className="animate-spin" />
307
+ ) : (
308
+ <IconMicrophone />
309
+ )}
310
+ {copy.startVoiceMode}
311
+ </Button>
312
+ <Button
313
+ type="button"
314
+ variant="outline"
315
+ size="sm"
316
+ className="w-full"
317
+ onClick={() => choose("dictation", onKeepDictating)}
318
+ >
319
+ {copy.keepDictating}
320
+ </Button>
321
+ </>
269
322
  )}
270
323
  </div>
324
+
325
+ {!setupRequired ? (
326
+ <label
327
+ htmlFor={rememberPreferenceId}
328
+ className="flex cursor-pointer items-center gap-2 text-xs text-muted-foreground"
329
+ >
330
+ <Checkbox
331
+ id={rememberPreferenceId}
332
+ checked={rememberPreference}
333
+ onCheckedChange={(checked) =>
334
+ setRememberPreference(checked === true)
335
+ }
336
+ />
337
+ {copy.rememberPreference}
338
+ </label>
339
+ ) : null}
271
340
  </div>
272
341
  </PopoverContent>
273
342
  </Popover>
@@ -312,7 +381,7 @@ export interface RealtimeVoiceModeInlineSettings {
312
381
 
313
382
  const SILENT_AUDIO_LEVELS = createRealtimeVoiceAudioLevelStore();
314
383
  const WAVEFORM_WEIGHTS = [0.55, 0.82, 1, 0.82, 0.55];
315
- const AUDIO_ACTIVITY_THRESHOLD = 0.035;
384
+ const AUDIO_ACTIVITY_THRESHOLD = 0.1;
316
385
 
317
386
  function usePrefersReducedMotion(): boolean {
318
387
  const [reduced, setReduced] = useState(false);
@@ -831,6 +900,7 @@ export function RealtimeVoiceModeDock({
831
900
  className={cn(
832
901
  "relative isolate size-16 overflow-visible rounded-full ring-1 backdrop-blur-xl transition-transform duration-150 ease-out focus-visible:ring-offset-2 active:scale-[0.97] motion-reduce:transition-none",
833
902
  ORB_STATE_CLASSES[state],
903
+ state === "working" && "agent-realtime-voice-working",
834
904
  )}
835
905
  >
836
906
  <span className="relative z-10 flex items-center justify-center">
@@ -1,6 +1,22 @@
1
1
  import { describe, expect, it } from "vitest";
2
2
 
3
- import { isRealtimeVoiceSetupRequired } from "./VoiceButton.js";
3
+ import {
4
+ isRealtimeVoiceSetupRequired,
5
+ normalizeVoiceInputPreference,
6
+ } from "./VoiceButton.js";
7
+
8
+ describe("normalizeVoiceInputPreference", () => {
9
+ it("accepts only persisted realtime and dictation choices", () => {
10
+ expect(normalizeVoiceInputPreference({ mode: "realtime" })).toBe(
11
+ "realtime",
12
+ );
13
+ expect(normalizeVoiceInputPreference({ mode: "dictation" })).toBe(
14
+ "dictation",
15
+ );
16
+ expect(normalizeVoiceInputPreference({ mode: "voice" })).toBeNull();
17
+ expect(normalizeVoiceInputPreference(null)).toBeNull();
18
+ });
19
+ });
4
20
 
5
21
  describe("isRealtimeVoiceSetupRequired", () => {
6
22
  it("waits for the voice-specific provider status before prompting setup", () => {
@@ -14,9 +14,11 @@ import {
14
14
  IconLoader2,
15
15
  IconX,
16
16
  } from "@tabler/icons-react";
17
+ import { useCallback, useEffect, useState } from "react";
17
18
 
18
19
  import { Tooltip, TooltipContent, TooltipTrigger } from "../ui/tooltip.js";
19
20
  import { RealtimeVoiceModeEntry } from "./RealtimeVoiceMode.js";
21
+ import type { RealtimeVoiceInputMode } from "./RealtimeVoiceMode.js";
20
22
  import { useComposerRuntimeAdapters } from "./runtime-adapters.js";
21
23
  import {
22
24
  useRealtimeVoiceModeCopy,
@@ -24,6 +26,16 @@ import {
24
26
  } from "./useRealtimeVoiceMode.js";
25
27
  import type { VoiceDictationApi } from "./useVoiceDictation.js";
26
28
 
29
+ const VOICE_INPUT_PREFERENCE_KEY = "voice-input-preference";
30
+
31
+ export function normalizeVoiceInputPreference(
32
+ value: unknown,
33
+ ): RealtimeVoiceInputMode | null {
34
+ if (!value || typeof value !== "object" || Array.isArray(value)) return null;
35
+ const mode = (value as { mode?: unknown }).mode;
36
+ return mode === "realtime" || mode === "dictation" ? mode : null;
37
+ }
38
+
27
39
  function openOpenAiKeySettings(): void {
28
40
  window.location.hash = "#secrets:OPENAI_API_KEY";
29
41
  window.dispatchEvent(new Event("agent-panel:open"));
@@ -63,6 +75,41 @@ export function VoiceButton({ voice, isMac, disabled }: VoiceButtonProps) {
63
75
  trackingFlow: "voice_transcription",
64
76
  onConnected: () => voiceProviders.refresh(),
65
77
  });
78
+ const [inputPreference, setInputPreference] = useState<{
79
+ ready: boolean;
80
+ mode: RealtimeVoiceInputMode | null;
81
+ }>({ ready: false, mode: null });
82
+
83
+ useEffect(() => {
84
+ let cancelled = false;
85
+ void Promise.resolve(
86
+ adapters.voice!.readAppState!(VOICE_INPUT_PREFERENCE_KEY),
87
+ )
88
+ .then((value) => {
89
+ if (!cancelled) {
90
+ setInputPreference({
91
+ ready: true,
92
+ mode: normalizeVoiceInputPreference(value),
93
+ });
94
+ }
95
+ })
96
+ .catch(() => {
97
+ if (!cancelled) setInputPreference({ ready: true, mode: null });
98
+ });
99
+ return () => {
100
+ cancelled = true;
101
+ };
102
+ }, [adapters]);
103
+
104
+ const rememberInputPreference = useCallback(
105
+ (mode: RealtimeVoiceInputMode) => {
106
+ setInputPreference({ ready: true, mode });
107
+ void Promise.resolve(
108
+ adapters.voice!.setAppState!(VOICE_INPUT_PREFERENCE_KEY, { mode }),
109
+ ).catch(() => undefined);
110
+ },
111
+ [adapters],
112
+ );
66
113
 
67
114
  if (!supported) return null;
68
115
 
@@ -90,6 +137,8 @@ export function VoiceButton({ voice, isMac, disabled }: VoiceButtonProps) {
90
137
  }}
91
138
  onStartVoiceMode={() => void realtimeVoice.start()}
92
139
  onKeepDictating={() => void start()}
140
+ preferredMode={inputPreference.ready ? inputPreference.mode : null}
141
+ onRememberPreference={rememberInputPreference}
93
142
  />
94
143
  );
95
144
  }
@@ -12,6 +12,21 @@ describe("realtime voice audio levels", () => {
12
12
  expect(normalizeRealtimeVoiceRms(new Uint8Array(32).fill(255))).toBe(1);
13
13
  });
14
14
 
15
+ it("makes ordinary speech visibly responsive above the noise floor", () => {
16
+ const quietSpeech = Uint8Array.from({ length: 32 }, (_, index) =>
17
+ index % 2 === 0 ? 124 : 132,
18
+ );
19
+ const typicalSpeech = Uint8Array.from({ length: 32 }, (_, index) =>
20
+ index % 2 === 0 ? 120 : 136,
21
+ );
22
+
23
+ expect(normalizeRealtimeVoiceRms(quietSpeech)).toBeGreaterThan(0.3);
24
+ expect(normalizeRealtimeVoiceRms(typicalSpeech)).toBeGreaterThan(0.5);
25
+ expect(normalizeRealtimeVoiceRms(typicalSpeech)).toBeGreaterThan(
26
+ normalizeRealtimeVoiceRms(quietSpeech),
27
+ );
28
+ });
29
+
15
30
  it("attacks faster than it decays", () => {
16
31
  expect(smoothRealtimeVoiceLevel(0, 1)).toBe(0.55);
17
32
  expect(smoothRealtimeVoiceLevel(1, 0)).toBe(0.8);
@@ -25,7 +25,8 @@ export function normalizeRealtimeVoiceRms(samples: Uint8Array): number {
25
25
  sumSquares += normalized * normalized;
26
26
  }
27
27
  const rms = Math.sqrt(sumSquares / samples.length);
28
- return clampLevel(Math.max(0, rms - 0.012) * 3.2);
28
+ const activeLevel = clampLevel((rms - 0.012) / 0.18);
29
+ return Math.sqrt(activeLevel);
29
30
  }
30
31
 
31
32
  export function smoothRealtimeVoiceLevel(
@@ -6,6 +6,7 @@ import {
6
6
  createRealtimeVoiceGreetingEvent,
7
7
  createRealtimeVoiceGreetingStarter,
8
8
  createRealtimeVoicePreferenceUpdate,
9
+ createRealtimeVoiceResponseCoordinator,
9
10
  createRealtimeVoiceSession,
10
11
  createRealtimeVoiceSessionWithCapability,
11
12
  createRealtimeVoiceTranscriptSequencer,
@@ -521,7 +522,7 @@ describe("Realtime voice dynamic tool manifests", () => {
521
522
  });
522
523
 
523
524
  describe("extractRealtimeVoiceFunctionCalls", () => {
524
- it("uses the low-latency completed-arguments event", () => {
525
+ it("waits for the terminal response event before returning a function call", () => {
525
526
  expect(
526
527
  extractRealtimeVoiceFunctionCalls({
527
528
  type: "response.function_call_arguments.done",
@@ -529,13 +530,7 @@ describe("extractRealtimeVoiceFunctionCalls", () => {
529
530
  call_id: "call-1",
530
531
  arguments: '{"path":"/inbox"}',
531
532
  }),
532
- ).toEqual([
533
- {
534
- name: "navigate",
535
- callId: "call-1",
536
- argumentsText: '{"path":"/inbox"}',
537
- },
538
- ]);
533
+ ).toEqual([]);
539
534
  });
540
535
 
541
536
  it("falls back to completed function items on response.done", () => {
@@ -543,6 +538,7 @@ describe("extractRealtimeVoiceFunctionCalls", () => {
543
538
  extractRealtimeVoiceFunctionCalls({
544
539
  type: "response.done",
545
540
  response: {
541
+ status: "completed",
546
542
  output: [
547
543
  { type: "message", role: "assistant" },
548
544
  {
@@ -562,6 +558,29 @@ describe("extractRealtimeVoiceFunctionCalls", () => {
562
558
  },
563
559
  ]);
564
560
  });
561
+
562
+ it("ignores function items from failed or cancelled responses", () => {
563
+ const output = [
564
+ {
565
+ type: "function_call",
566
+ name: "navigate",
567
+ call_id: "call-3",
568
+ arguments: "{}",
569
+ },
570
+ ];
571
+ expect(
572
+ extractRealtimeVoiceFunctionCalls({
573
+ type: "response.done",
574
+ response: { status: "failed", output },
575
+ }),
576
+ ).toEqual([]);
577
+ expect(
578
+ extractRealtimeVoiceFunctionCalls({
579
+ type: "response.done",
580
+ response: { status: "cancelled", output },
581
+ }),
582
+ ).toEqual([]);
583
+ });
565
584
  });
566
585
 
567
586
  describe("extractCompletedRealtimeVoiceTranscript", () => {
@@ -632,6 +651,61 @@ describe("Realtime voice startup and transcript ordering", () => {
632
651
  expect(send).toHaveBeenCalledTimes(2);
633
652
  });
634
653
 
654
+ it("queues response requests behind the active response and retries a rejected request", () => {
655
+ const sent: Record<string, unknown>[] = [];
656
+ const coordinator = createRealtimeVoiceResponseCoordinator((event) =>
657
+ sent.push(event),
658
+ );
659
+
660
+ coordinator.request();
661
+ coordinator.handleEvent({
662
+ type: "response.created",
663
+ response: { id: "resp-active" },
664
+ });
665
+ coordinator.request();
666
+ expect(sent).toHaveLength(1);
667
+
668
+ expect(
669
+ coordinator.handleError(
670
+ "Conversation already has an active response in progress: resp-active",
671
+ ),
672
+ ).toBe(true);
673
+ coordinator.handleEvent({
674
+ type: "response.done",
675
+ response: { id: "resp-active", status: "completed" },
676
+ });
677
+
678
+ expect(sent).toHaveLength(2);
679
+ expect(sent[1]).toEqual({ type: "response.create" });
680
+ });
681
+
682
+ it("drops queued work after a failed response so the next turn can start", () => {
683
+ const sent: Record<string, unknown>[] = [];
684
+ const coordinator = createRealtimeVoiceResponseCoordinator((event) =>
685
+ sent.push(event),
686
+ );
687
+
688
+ coordinator.request();
689
+ coordinator.handleEvent({ type: "response.created" });
690
+ coordinator.request();
691
+ coordinator.handleEvent({
692
+ type: "response.done",
693
+ response: { status: "failed" },
694
+ });
695
+
696
+ expect(sent).toHaveLength(1);
697
+ coordinator.request();
698
+ expect(sent).toHaveLength(2);
699
+ });
700
+
701
+ it("ignores unrelated realtime errors", () => {
702
+ const coordinator = createRealtimeVoiceResponseCoordinator(vi.fn());
703
+
704
+ expect(coordinator.handleError("The microphone was disconnected.")).toBe(
705
+ false,
706
+ );
707
+ });
708
+
635
709
  it("publishes in conversation order when user ASR finishes after the assistant", () => {
636
710
  const published: Array<{ role: string; text: string }> = [];
637
711
  const sequencer = createRealtimeVoiceTranscriptSequencer((transcript) => {