Update app.py

app.py

@@ -134,7 +134,10 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
         if job_id:
             _update_job(job_id, status="processing", step=step, message=message)
 
-    out_dir = tempfile.mkdtemp()
+    out_dir       = tempfile.mkdtemp()
+    # FIX: persistent output file outside out_dir so it survives cleanup
+    fd, final_audio_path = tempfile.mkstemp(suffix="_clearwave.wav")
+    os.close(fd)
     try:
         progress(1, "⏳ Step 1/5 — Denoising...")
         denoise1 = denoiser.process(
@@ -182,7 +185,11 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
         progress(5, "⏳ Step 5/5 — Summarizing...")
         summary = translator.summarize(transcript)
 
-        with open(clean1, "rb") as f:
+        # FIX: copy audio out of out_dir before we delete it in finally
+        import shutil as _shutil
+        _shutil.copy2(clean1, final_audio_path)
+
+        with open(final_audio_path, "rb") as f:
             enhanced_b64 = base64.b64encode(f.read()).decode("utf-8")
 
         result = {
@@ -192,7 +199,7 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
             "transcript":    transcript,
             "translation":   translation,
             "summary":       summary,
-            "audioPath":     clean1,
+            "audioPath":     final_audio_path,
             "enhancedAudio": enhanced_b64,
             "stats": {
                 "language":             detected_lang.upper(),
@@ -217,10 +224,19 @@ def run_pipeline(audio_path, src_lang="auto", tgt_lang="te",
 
     except Exception as e:
         logger.error(f"Pipeline failed: {e}", exc_info=True)
+        # FIX: clean up the persistent output file on error
+        try:
+            os.unlink(final_audio_path)
+        except Exception:
+            pass
         err = {"status": "error", "message": f"❌ Error: {str(e)}"}
         if job_id:
             _update_job(job_id, **err)
         return err
+    finally:
+        # FIX: always delete the temp working directory
+        import shutil as _shutil
+        _shutil.rmtree(out_dir, ignore_errors=True)
 
 
 # ══════════════════════════════════════════════════════════════════════
@@ -465,17 +481,24 @@ async def api_process_url(request: _Request):
     _update_job(job_id, status="downloading", message="Downloading audio...")
 
     def _download_and_run():
-        tmp_path   = None
-        audio_path = None
+        tmp_path       = None
+        converted_path = None   # FIX: track convert_to_wav output separately
         try:
             resp = requests.get(audio_url, timeout=300, stream=True)
             resp.raise_for_status()
-            url_lower = audio_url.lower()
-            if   "wav"  in url_lower: suffix = ".wav"
-            elif "mpeg" in url_lower: suffix = ".mpeg"
-            elif "mp4"  in url_lower: suffix = ".mp4"
-            elif "m4a"  in url_lower: suffix = ".m4a"
-            else:                     suffix = ".mp3"
+            # FIX: detect suffix from Content-Type header first, fall back to URL
+            content_type = resp.headers.get("content-type", "")
+            if   "wav"  in content_type: suffix = ".wav"
+            elif "mpeg" in content_type: suffix = ".mp3"
+            elif "mp4"  in content_type: suffix = ".mp4"
+            elif "aac"  in content_type: suffix = ".aac"
+            else:
+                url_lower = audio_url.lower().split("?")[0]  # strip query params
+                if   url_lower.endswith(".wav"):  suffix = ".wav"
+                elif url_lower.endswith(".mp4"):  suffix = ".mp4"
+                elif url_lower.endswith(".m4a"):  suffix = ".m4a"
+                elif url_lower.endswith(".mpeg"): suffix = ".mpeg"
+                else:                             suffix = ".mp3"
 
             fd, tmp_path = tempfile.mkstemp(suffix=suffix)
             os.close(fd)
@@ -491,11 +514,12 @@ async def api_process_url(request: _Request):
                             _update_job(job_id, status="downloading",
                                         message=f"Downloading... {pct}%")
 
-            audio_path = convert_to_wav(tmp_path)
+            # FIX: convert_to_wav may produce a *different* file; track it
+            converted_path = convert_to_wav(tmp_path)
 
             # FIX L11: plain function call
             run_pipeline(
-                audio_path, src_lang, tgt_lang,
+                converted_path, src_lang, tgt_lang,
                 opt_fillers, opt_stutters, opt_silences,
                 opt_breaths, opt_mouth, job_id=job_id
             )
@@ -508,7 +532,8 @@ async def api_process_url(request: _Request):
             logger.error(f"Job {job_id} failed: {e}", exc_info=True)
             _update_job(job_id, status="error", message=f"❌ Error: {str(e)}")
         finally:
-            for p in [tmp_path, audio_path]:
+            # FIX: delete both the download and the converted copy
+            for p in {tmp_path, converted_path}:   # set deduplicates if equal
                 try:
                     if p and os.path.exists(p):
                         os.unlink(p)