@ducci/jarvis 1.0.87 → 1.0.89

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ducci/jarvis",
-  "version": "1.0.87",
+  "version": "1.0.89",
   "description": "A fully automated agent system that lives on a server.",
   "main": "./src/index.js",
   "type": "module",

package/src/channels/telegram/index.js

@@ -699,7 +699,13 @@ export async function startTelegramChannel(config) {
           // TTS: send audio summary if voice is enabled (config.voiceEnabled checked live, updated by /voice toggle)
           if (config.voiceEnabled && config.fishAudioApiKey) {
             try {
-              const plain = toPlainText(displayText);
+              // If the response is a raw JSON blob (format_error recovery), extract the actual text
+              let ttsSource = displayText;
+              try {
+                const parsed = JSON.parse(displayText);
+                if (parsed?.response) ttsSource = parsed.response;
+              } catch { /* not JSON, use as-is */ }
+              const plain = toPlainText(ttsSource);
               if (plain) {
                 const ttsText = await generateTtsSummary(plain, config);
                 if (ttsText) {

package/src/server/fish-audio.js

@@ -8,6 +8,7 @@ import { promisify } from 'util';
 import { writeFile, readFile, unlink } from 'fs/promises';
 import { tmpdir } from 'os';
 import { join } from 'path';
+import { Readable } from 'stream';
 const execAsync = promisify(exec);
 
 // System prompt for TTS summary generation.
@@ -26,7 +27,7 @@ Rules:
 - Choose emotions that fit the content: use (confident) or (calm) for informational answers,
   (excited) or (satisfied) for completed tasks, (curious) for questions, etc.
 - You may combine two tags on one sentence: (excited)(soft tone) Great news!
-- No markdown, no code blocks, no bullet points — plain speech only.
+- Plain text only — no emojis, no markdown, no code blocks, no bullet points, no special characters.
 - Keep technical jargon minimal; explain concepts simply as you would speak them.
 - Match the language of the original response.
 
@@ -89,7 +90,13 @@ export async function textToSpeech(text, config) {
     throw new Error(`fish.audio TTS ${response.status}: ${errText.slice(0, 200)}`);
   }
 
-  return Buffer.from(await response.arrayBuffer());
+  // fish.audio streams audio via chunked transfer encoding — arrayBuffer() only
+  // returns the first chunk. Collect all chunks via a Node.js stream.
+  const chunks = [];
+  for await (const chunk of Readable.fromWeb(response.body)) {
+    chunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
+  }
+  return Buffer.concat(chunks);
 }
 
 /**