From 6062c76cf8b2e5fd0a3d9d3f424f991112d8c843 Mon Sep 17 00:00:00 2001
From: BadStorm Developer <developer@badstorm.xyz>
Date: Sat, 5 Sep 2026 15:18:42 +0200
Subject: [PATCH] fix(talkbot): retry failed voice message processing

---
 containers/talkbot/server.py | 77 +++++++++++++++++++++---------------
 1 file changed, 46 insertions(+), 31 deletions(-)

diff --git a/containers/talkbot/server.py b/containers/talkbot/server.py
index 95a3b95..d18d2b4 100644
--- a/containers/talkbot/server.py
+++ b/containers/talkbot/server.py
@@ -4,6 +4,7 @@ import os
 import re
 import sqlite3
 import subprocess
+import tempfile
 import threading
 from datetime import datetime, timedelta, timezone
 from typing import Optional
@@ -492,40 +493,53 @@ def is_blank_transcription(text: str) -> bool:
 
 def convert_audio_to_wav(audio_bytes: bytes) -> bytes:
     """Convert any FFmpeg-supported audio format to Whisper's WAV format."""
-    try:
-        result = subprocess.run(
-            [
-                "ffmpeg",
-                "-hide_banner",
-                "-loglevel",
-                "error",
-                "-i",
-                "pipe:0",
-                "-vn",
-                "-ac",
-                "1",
-                "-ar",
-                "16000",
-                "-c:a",
-                "pcm_s16le",
-                "-f",
-                "wav",
-                "pipe:1",
-            ],
-            input=audio_bytes,
-            capture_output=True,
-            check=False,
-            timeout=120,
-        )
-    except FileNotFoundError as exc:
-        raise RuntimeError("ffmpeg is not installed in the talkbot container") from exc
-    except subprocess.TimeoutExpired as exc:
-        raise RuntimeError("Audio conversion with ffmpeg timed out") from exc
+    # Use a temporary file instead of stdin: MP4-based audio often has its
+    # metadata (moov atom) at the end and requires seeking during probing.
+    with tempfile.NamedTemporaryFile(
+        prefix="talkbot-input-",
+        suffix=".audio",
+    ) as input_file:
+        input_file.write(audio_bytes)
+        input_file.flush()
+
+        try:
+            result = subprocess.run(
+                [
+                    "ffmpeg",
+                    "-hide_banner",
+                    "-loglevel",
+                    "error",
+                    "-i",
+                    input_file.name,
+                    "-vn",
+                    "-ac",
+                    "1",
+                    "-ar",
+                    "16000",
+                    "-c:a",
+                    "pcm_s16le",
+                    "-f",
+                    "wav",
+                    "pipe:1",
+                ],
+                capture_output=True,
+                check=False,
+                timeout=120,
+            )
+        except FileNotFoundError as exc:
+            raise RuntimeError(
+                "ffmpeg is not installed in the talkbot container"
+            ) from exc
+        except subprocess.TimeoutExpired as exc:
+            raise RuntimeError("Audio conversion with ffmpeg timed out") from exc
 
     if result.returncode != 0 or not result.stdout:
         error = result.stderr.decode("utf-8", errors="replace").strip()
         raise RuntimeError(f"Unable to convert audio to WAV: {error[:500]}")
 
+    if len(result.stdout) <= 78:
+        raise RuntimeError("Audio conversion produced an empty WAV file")
+
     log.info(
         "Audio converted to WAV: %s -> %s bytes",
         len(audio_bytes),
@@ -1198,8 +1212,9 @@ def process_event(job: dict) -> None:
                 message_id,
                 exc,
             )
-
-        return
+            # Let event_worker() reschedule the queued job. If we return here,
+            # the worker considers the event successful and removes it.
+            raise
 
     if message_text:
         log.info(