Update app.py

app.py

@@ -42,45 +42,37 @@ def extract_main_content(url):
 
 def generate_human_like_audio(text, speaker):
     try:
-        temp_dir = tempfile.mkdtemp()
-        wav_path = os.path.join(temp_dir, "summary.wav")
-        mp3_path = os.path.join(temp_dir, "summary.mp3")
+        # Save in tmp directory accessible by Gradio
+        base = tempfile.gettempdir()
+        wav_path = os.path.join(base, "summary.wav")
+        mp3_path = os.path.join(base, "summary.mp3")
 
         tts_model.tts_to_file(text=text, speaker=speaker, file_path=wav_path)
-
         os.system(f"ffmpeg -y -i {wav_path} -codec:a libmp3lame -qscale:a 4 {mp3_path}")
 
-        if os.path.exists(mp3_path):
-            return wav_path, mp3_path
-        else:
-            return wav_path, None
+        return wav_path if os.path.exists(wav_path) else None, mp3_path if os.path.exists(mp3_path) else None
     except Exception as e:
         print(f"TTS ERROR: {e}")
         return None, None
 
 def url_to_audio_summary(url, speaker):
-    try:
-        article_text = extract_main_content(url)
-        if article_text.startswith("Error"):
-            return article_text, None, None
+    article_text = extract_main_content(url)
+    if article_text.startswith("Error"):
+        return article_text, None, None
 
-        if len(article_text) > 1500:
-            article_text = article_text[:1500] + "..."
+    # Truncate
+    if len(article_text) > 1500:
+        article_text = article_text[:1500] + "..."
 
-        summary = summary_chain.invoke({"text": article_text})
-        summary = summary["text"] if isinstance(summary, dict) and "text" in summary else summary
+    summary = (summary_chain.invoke if hasattr(summary_chain, 'invoke') else summary_chain)({"text": article_text})
+    summary = summary.get("text") if isinstance(summary, dict) else summary
 
-        wav_path, mp3_path = generate_human_like_audio(summary, speaker)
-        return summary, wav_path, mp3_path
-    except Exception as e:
-        return f"Error: {str(e)}", None, None
+    wav_path, mp3_path = generate_human_like_audio(summary, speaker)
+    return summary, wav_path, mp3_path
 
 def interface_wrapper(url, speaker):
     summary, wav_path, mp3_path = url_to_audio_summary(url, speaker)
-    download_html = ""
-    if mp3_path and os.path.exists(mp3_path):
-        download_html = f'<a href="file/{os.path.basename(mp3_path)}" download target="_blank">Click to download MP3</a>'
-    return summary, wav_path, download_html
+    return summary, wav_path, mp3_path
 
 iface = gr.Interface(
     fn=interface_wrapper,
@@ -91,7 +83,7 @@ iface = gr.Interface(
     outputs=[
         gr.Textbox(label="Summary"),
         gr.Audio(label="Preacher-style Audio Summary", type="filepath"),
-        gr.HTML(label="Download MP3")
+        gr.File(label="Download MP3")
     ],
     title="Preaching-Style URL to Audio Agent",
     description="Summarizes article content and reads it aloud in a warm, preacher-style voice using YourTTS. CPU-only."