Prioritize external transcription endpoint over browser Web Speech API

Previously the on-device Web Speech API (Chrome/Edge/Safari) was always
preferred, and the configured external transcription endpoint was only
used as a fallback when the browser had no speech API at all (Firefox).

Flip the priority in Composer.tsx: if a transcription endpoint is
configured in Settings, use it for dictation regardless of browser
support, since it's generally far more accurate. Fall back to the
browser's on-device SpeechRecognition only when no endpoint is set.

Update the Settings hint text and README to describe the new behavior.
Rebuilt and redeployed to the repo root.

app/src/components/Composer.tsx

@@ -45,8 +45,10 @@ export function Composer() {
   const speech = useSpeechRecognition();
   const recorder = useRecorder();
 
-  const serverDictation = !speech.supported && !!transcribeUrl.trim();
-  const dictationAvailable = speech.supported || serverDictation;
+  // Prefer an explicitly-configured external transcription endpoint (usually
+  // far more accurate) over the browser's on-device Web Speech API.
+  const serverDictation = !!transcribeUrl.trim();
+  const dictationAvailable = serverDictation || speech.supported;
   const active = speech.listening || recorder.recording;
 
   useEffect(() => {
@@ -129,6 +131,13 @@ export function Composer() {
       await stopDictation();
       return;
     }
+    // External transcription endpoint (if configured) takes priority over the
+    // browser's on-device Web Speech API, which tends to be lower quality.
+    if (serverDictation) {
+      const ok = await recorder.start();
+      if (!ok) setToast("Could not access the microphone.");
+      return;
+    }
     if (speech.supported) {
       if (!window.isSecureContext) {
         setToast("Dictation needs a secure context (HTTPS or localhost).");
@@ -137,11 +146,6 @@ export function Composer() {
       speech.start(append);
       return;
     }
-    if (serverDictation) {
-      const ok = await recorder.start();
-      if (!ok) setToast("Could not access the microphone.");
-      return;
-    }
     setToast(
       "Dictation isn't available in this browser. Use Chrome, Edge, or Safari — or set a transcription endpoint in Settings.",
     );

app/src/components/SettingsPanel.tsx

@@ -130,8 +130,9 @@ export function SettingsPanel() {
         <div className="section">
           <h3>Dictation</h3>
           <div className="hint" style={{ marginTop: 0, marginBottom: 10 }}>
-            Chrome, Edge, and Safari dictate on-device via the Web Speech API. Firefox has no speech
-            API, so set an OpenAI-compatible transcription endpoint below to dictate there.
+            If a transcription endpoint is set below, it's used for dictation (usually far more
+            accurate). Otherwise Chrome, Edge, and Safari fall back to on-device dictation via the
+            Web Speech API — Firefox has no speech API, so it needs an endpoint set here.
           </div>
           <div className="field">
             <label>Transcription endpoint</label>