fix(talkbot): improve text translation language handling

containers/talkbot/server.py

@@ -62,6 +62,18 @@ QUEUE_DB_PATH = os.environ.get(
 MAX_ATTEMPTS = 3
 RETRY_BACKOFF_SECONDS = 20
 
+AUDIO_EXTENSIONS = {
+    ".aac",
+    ".flac",
+    ".m4a",
+    ".mp3",
+    ".oga",
+    ".ogg",
+    ".opus",
+    ".wav",
+    ".webm",
+}
+
 MANAGER_USER = NC_USER
 
 app = FastAPI(title="Nextcloud Talk voice translator")
@@ -314,6 +326,20 @@ def extract_file_param(parameters) -> Optional[dict]:
     return None
 
 
+def is_audio_file(file_param: dict) -> bool:
+    mimetype = str(file_param.get("mimetype", "") or "").lower()
+    if mimetype.startswith("audio/"):
+        return True
+
+    filename = str(
+        file_param.get("name")
+        or file_param.get("path")
+        or ""
+    ).lower()
+
+    return any(filename.endswith(extension) for extension in AUDIO_EXTENSIONS)
+
+
 def actor_user_id(actor: dict) -> str:
     actor_id = str(actor.get("id", "") or "")
 
@@ -1223,6 +1249,17 @@ def process_event(job: dict) -> None:
             sender_user_id,
         )
 
+        sender_language = normalize_language_code(
+            get_user_language(sender_user_id),
+        )
+
+        log.info(
+            "Text sender %s language: %s -> %s",
+            sender_user_id,
+            sender_language,
+            language_name(sender_language),
+        )
+
         translations = []
 
         for recipient in recipients:
@@ -1232,10 +1269,17 @@ def process_event(job: dict) -> None:
 
             translated = translate(
                 message_text,
-                "auto",
+                sender_language,
                 target_language,
             )
 
+            log.info(
+                "Translated text for %s (%s): %s",
+                recipient,
+                language_name(target_language),
+                translated,
+            )
+
             if translated.strip().lower() == message_text.lower():
                 continue
 
@@ -1464,6 +1508,18 @@ async def webhook(
         parameters,
     )
 
+    if file_param and not is_audio_file(file_param):
+        log.info(
+            "Ignoring unsupported attachment in message %s: name=%s mimetype=%s",
+            message_id,
+            file_param.get("name"),
+            file_param.get("mimetype"),
+        )
+
+        # If the message also contains text, process the text and ignore the
+        # unrelated attachment. Otherwise there is nothing for this bot to do.
+        file_param = None
+
     if not file_param and not message_text:
         log.info(
             "Message %s has no text or attachment",