pi-web-voice 0.1.2 → 0.1.4

package/bin/pi-web-voice.js

@@ -21,7 +21,16 @@ const hook = path.join(__dirname, "..", "hook.cjs");
 // `pi-web-voice doctor [file.wav]` checks the speech backend and exits.
 if (args[0] === "doctor") {
   const { doctor } = require("../lib/doctor.cjs");
-  doctor(args.slice(1)).then((code) => process.exit(code));
+  doctor(args.slice(1)).then((code) => {
+    // Setting the code rather than calling process.exit lets the HTTP
+    // connection finish closing. Forcing an exit mid-teardown trips a libuv
+    // assertion on Windows: !(handle->flags & UV_HANDLE_CLOSING).
+    process.exitCode = code;
+    // Keep-alive sockets can still hold the loop open for a few seconds after
+    // the answer is in hand. This timer does not itself keep the process
+    // alive, and by the time it fires nothing is mid-close.
+    setTimeout(() => process.exit(code), 750).unref();
+  });
   return;
 }
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "pi-web-voice",
-  "version": "0.1.2",
+  "version": "0.1.4",
   "description": "Voice input for pi-web. A NODE_OPTIONS hook that injects a microphone button into the chat composer and transcribes speech with Azure AI Speech, Azure OpenAI, or any OpenAI-compatible endpoint. No fork, no patching, no rebuild.",
   "keywords": [
     "pi",

package/public/inject.js

@@ -97,6 +97,7 @@
         denied: "麦克风权限被拒绝",
         empty: "没有识别到语音",
         failed: "转写失败",
+        noComposer: "找不到输入框，转写结果",
       }
     : {
         idle: "Voice input — click, or press and hold",
@@ -107,6 +108,7 @@
         denied: "Microphone permission denied",
         empty: "No speech detected",
         failed: "Transcription failed",
+        noComposer: "No composer found; transcript",
       };
 
   // ── audio ────────────────────────────────────────────────────────────────
@@ -224,10 +226,26 @@
 
   // ── composer ─────────────────────────────────────────────────────────────
 
+  // "The last visible textarea" is not good enough: pi-web's workspace
+  // terminal is xterm.js, which keeps a hidden IME helper textarea mounted
+  // after the composer. Writing there drops the transcript at best, and at
+  // worst hands it to the shell.
+  function isComposerCandidate(area) {
+    if (area.offsetParent === null) return false;
+    if (area.classList.contains("xterm-helper-textarea")) return false;
+    return !area.closest(".xterm");
+  }
+
   function findComposer() {
-    const areas = Array.from(document.querySelectorAll("textarea")).filter(
-      (area) => area.offsetParent !== null,
-    );
+    // The button was mounted next to the composer's own toolbar, so walking up
+    // from it finds the right textarea even when the page holds several.
+    const button = document.getElementById(BUTTON_ID);
+    for (let node = button?.parentElement; node; node = node.parentElement) {
+      const nearby = Array.from(node.querySelectorAll("textarea")).filter(isComposerCandidate);
+      if (nearby.length) return nearby[nearby.length - 1];
+    }
+
+    const areas = Array.from(document.querySelectorAll("textarea")).filter(isComposerCandidate);
     return areas[areas.length - 1] || null;
   }
 
@@ -453,7 +471,10 @@
           // Always inserted, never sent: a wrong term is one keystroke from
           // being fixed, and Enter is right there when it is correct.
           const textarea = findComposer();
+          // Never fail silently: a transcript with nowhere to go is shown
+          // rather than dropped, so it can still be copied by hand.
           if (textarea) insertAtCaret(textarea, text);
+          else this.toast(`${T.noComposer}: ${text}`);
         }
       } catch (error) {
         this.toast(`${T.failed}: ${error.message}`);
@@ -516,6 +537,7 @@
     ui,
     recorder,
     config: CONFIG,
+    findComposer,
     get sessionId() {
       return sessionId;
     },