From 5f24fd4db7f409c48ce1b1e0363e1516e31f7388 Mon Sep 17 00:00:00 2001
From: Shantur Rathore <i@shantur.com>
Date: Tue, 24 Mar 2026 22:42:27 +0000
Subject: [PATCH] feat(speech): make prompt input push to talk

---
 packages/ui/src/components/prompt-input.tsx   | 62 ++++++++++++++++---
 .../prompt-input/usePromptVoiceInput.ts       | 60 +++++++++---------
 .../ui/src/styles/messaging/prompt-input.css  |  8 ++-
 3 files changed, 90 insertions(+), 40 deletions(-)

diff --git a/packages/ui/src/components/prompt-input.tsx b/packages/ui/src/components/prompt-input.tsx
index f128791b..3b56c468 100644
--- a/packages/ui/src/components/prompt-input.tsx
+++ b/packages/ui/src/components/prompt-input.tsx
@@ -1,5 +1,5 @@
 import { Suspense, createEffect, createSignal, lazy, on, onCleanup, Show } from "solid-js"
-import { ArrowBigUp, ArrowBigDown, Loader2, Mic, Square } from "lucide-solid"
+import { ArrowBigUp, ArrowBigDown, Loader2, Mic } from "lucide-solid"
 import ExpandButton from "./expand-button"
 import { clearAttachments, removeAttachment } from "../stores/attachments"
 import { resolvePastedPlaceholders } from "../lib/prompt-placeholders"
@@ -464,6 +464,32 @@ export default function PromptInput(props: PromptInputProps) {
 
   const instance = () => getActiveInstance()
 
+  let voiceButtonPressed = false
+
+  const beginVoicePress = (event?: PointerEvent | KeyboardEvent) => {
+    if (voiceButtonPressed || props.disabled || voiceInput.isTranscribing() || !voiceInput.canUseVoiceInput()) return
+    voiceButtonPressed = true
+
+    if (event instanceof PointerEvent) {
+      const target = event.currentTarget
+      if (target instanceof HTMLElement) {
+        try {
+          target.setPointerCapture(event.pointerId)
+        } catch {
+          // no-op
+        }
+      }
+    }
+
+    void voiceInput.startRecording()
+  }
+
+  const endVoicePress = () => {
+    if (!voiceButtonPressed) return
+    voiceButtonPressed = false
+    voiceInput.stopRecording()
+  }
+
   return (
     <div class="prompt-input-container">
       <div
@@ -612,25 +638,43 @@ export default function PromptInput(props: PromptInputProps) {
             <button
               type="button"
               class={`prompt-voice-button ${voiceInput.isRecording() ? "is-recording" : ""}`}
-              onClick={() => void voiceInput.toggleRecording()}
+              onPointerDown={(event) => {
+                event.preventDefault()
+                beginVoicePress(event)
+              }}
+              onPointerUp={(event) => {
+                event.preventDefault()
+                endVoicePress()
+              }}
+              onPointerCancel={() => endVoicePress()}
+              onLostPointerCapture={() => endVoicePress()}
+              onKeyDown={(event) => {
+                if (event.repeat) return
+                if (event.key !== " " && event.key !== "Enter") return
+                event.preventDefault()
+                beginVoicePress(event)
+              }}
+              onKeyUp={(event) => {
+                if (event.key !== " " && event.key !== "Enter") return
+                event.preventDefault()
+                endVoicePress()
+              }}
+              onBlur={() => endVoicePress()}
               disabled={!voiceInput.isRecording() && (props.disabled || voiceInput.isTranscribing() || !voiceInput.canUseVoiceInput())}
               aria-label={voiceInput.buttonTitle()}
               title={voiceInput.buttonTitle()}
             >
               <Show
-                when={voiceInput.isTranscribing()}
+                when={voiceInput.isRecording()}
                 fallback={
-                  <Show when={voiceInput.isRecording()} fallback={<Mic class="h-4 w-4" aria-hidden="true" />}>
-                    <Square class="h-4 w-4" aria-hidden="true" />
+                  <Show when={voiceInput.isTranscribing()} fallback={<Mic class="h-4 w-4" aria-hidden="true" />}>
+                    <Loader2 class="h-4 w-4 animate-spin" aria-hidden="true" />
                   </Show>
                 }
               >
-                <Loader2 class="h-4 w-4 animate-spin" aria-hidden="true" />
+                <span class="prompt-voice-timer">{formatVoiceTimer(voiceInput.elapsedMs())}</span>
               </Show>
             </button>
-            <Show when={voiceInput.isRecording()}>
-              <span class="prompt-voice-timer">{formatVoiceTimer(voiceInput.elapsedMs())}</span>
-            </Show>
           </Show>
           <button
             type="button"
diff --git a/packages/ui/src/components/prompt-input/usePromptVoiceInput.ts b/packages/ui/src/components/prompt-input/usePromptVoiceInput.ts
index f05c4e05..9a2713b8 100644
--- a/packages/ui/src/components/prompt-input/usePromptVoiceInput.ts
+++ b/packages/ui/src/components/prompt-input/usePromptVoiceInput.ts
@@ -56,18 +56,7 @@ export function usePromptVoiceInput(options: UsePromptVoiceInputOptions) {
       return
     }
 
-    if (!canUseVoiceInput() || options.disabled() || state() === "transcribing") return
-
-    try {
-      await startRecording()
-    } catch (error) {
-      cleanupMedia(false)
-      showAlertDialog(t("promptInput.voiceInput.error.permission"), {
-        title: t("promptInput.voiceInput.error.title"),
-        detail: error instanceof Error ? error.message : String(error),
-        variant: "error",
-      })
-    }
+    await startRecording()
   }
 
   function stopRecording() {
@@ -86,6 +75,8 @@ export function usePromptVoiceInput(options: UsePromptVoiceInputOptions) {
   }
 
   async function startRecording() {
+    if (!canUseVoiceInput() || options.disabled() || state() === "transcribing" || state() === "recording") return
+
     if (!isSupported()) {
       showAlertDialog(t("promptInput.voiceInput.error.unsupported"), {
         title: t("promptInput.voiceInput.error.title"),
@@ -94,26 +85,35 @@ export function usePromptVoiceInput(options: UsePromptVoiceInputOptions) {
       return
     }
 
-    recordedChunks = []
-    shouldTranscribe = true
-    mediaStream = await navigator.mediaDevices.getUserMedia({ audio: true })
-    mediaRecorder = createRecorder(mediaStream)
+    try {
+      recordedChunks = []
+      shouldTranscribe = true
+      mediaStream = await navigator.mediaDevices.getUserMedia({ audio: true })
+      mediaRecorder = createRecorder(mediaStream)
 
-    mediaRecorder.addEventListener("dataavailable", (event) => {
-      if (event.data.size > 0) {
-        recordedChunks.push(event.data)
-      }
-    })
+      mediaRecorder.addEventListener("dataavailable", (event) => {
+        if (event.data.size > 0) {
+          recordedChunks.push(event.data)
+        }
+      })
 
-    mediaRecorder.addEventListener("stop", () => {
-      void finalizeRecording()
-    })
+      mediaRecorder.addEventListener("stop", () => {
+        void finalizeRecording()
+      })
 
-    recordingStartedAt = Date.now()
-    setElapsedMs(0)
-    setState("recording")
-    startTimer()
-    mediaRecorder.start()
+      recordingStartedAt = Date.now()
+      setElapsedMs(0)
+      setState("recording")
+      startTimer()
+      mediaRecorder.start()
+    } catch (error) {
+      cleanupMedia(false)
+      showAlertDialog(t("promptInput.voiceInput.error.permission"), {
+        title: t("promptInput.voiceInput.error.title"),
+        detail: error instanceof Error ? error.message : String(error),
+        variant: "error",
+      })
+    }
   }
 
   async function finalizeRecording() {
@@ -209,6 +209,8 @@ export function usePromptVoiceInput(options: UsePromptVoiceInputOptions) {
     state,
     elapsedMs,
     canUseVoiceInput,
+    startRecording,
+    stopRecording,
     toggleRecording,
     cancelRecording,
     isRecording: () => state() === "recording",
diff --git a/packages/ui/src/styles/messaging/prompt-input.css b/packages/ui/src/styles/messaging/prompt-input.css
index 4bc749a7..f737d006 100644
--- a/packages/ui/src/styles/messaging/prompt-input.css
+++ b/packages/ui/src/styles/messaging/prompt-input.css
@@ -180,7 +180,8 @@
 }
 
 .prompt-voice-button {
-  @apply w-10 h-10 rounded-md border-none cursor-pointer flex items-center justify-center transition-all flex-shrink-0;
+  @apply h-10 rounded-md border-none cursor-pointer flex items-center justify-center transition-all flex-shrink-0;
+  min-width: 2.5rem;
   background-color: color-mix(in oklab, var(--surface-secondary) 82%, var(--surface-base));
   color: var(--text-secondary);
 }
@@ -196,6 +197,7 @@
 }
 
 .prompt-voice-button.is-recording {
+  min-width: 3.5rem;
   background-color: color-mix(in oklab, var(--button-danger-bg, rgba(239, 68, 68, 0.85)) 88%, white 12%);
   color: var(--button-danger-text, var(--text-inverted, #ffffff));
 }
@@ -206,8 +208,10 @@
 
 .prompt-voice-timer {
   font-size: 0.68rem;
+  font-variant-numeric: tabular-nums;
+  font-weight: 600;
   line-height: 1;
-  color: var(--text-muted);
+  color: currentColor;
 }
 
 .stop-button:hover:not(:disabled) {