Update app.py

app.py

@@ -1,6 +1,6 @@
 """
-Video Transcription APP
-Production-Safe Version with Retry, Chunk Handling & Progress Bar
+AI Graduate Job-Matching Platform
+Production-Safe Version with Retry & Chunk Handling
 Developer: Najaf Ali Sharqi
 MoviePy-free (uses ffmpeg + tempfile)
 """
@@ -20,7 +20,7 @@ client = Groq(api_key=os.getenv("GROQ_API_KEY"))
 
 TRANSCRIPTION_MODEL = "whisper-large-v3"
 REWRITE_MODEL = "openai/gpt-oss-120b"
-CHUNK_SIZE = 300  # 5 min chunks for transcription
+CHUNK_SIZE = 300  # 5 min chunks for transcription to avoid rate limit
 REWRITE_CHUNK_SIZE = 4000  # words per rewrite chunk
 MAX_RETRIES = 5  # retries for 429 errors
 
@@ -28,6 +28,7 @@ MAX_RETRIES = 5  # retries for 429 errors
 # UTILITY FUNCTIONS
 # ============================================================================
 def get_video_duration(video_path):
+    """Get video duration using ffprobe"""
     try:
         result = subprocess.run(
             ["ffprobe", "-v", "error", "-show_entries",
@@ -37,43 +38,49 @@ def get_video_duration(video_path):
             stderr=subprocess.STDOUT,
             text=True
         )
-        return float(result.stdout.strip())
+        duration = float(result.stdout.strip())
+        return duration
     except Exception as e:
         raise RuntimeError(f"Failed to get video duration: {str(e)}")
 
 def extract_audio_chunks(video_path, chunk_size=CHUNK_SIZE):
-    duration = get_video_duration(video_path)
-    num_chunks = math.ceil(duration / chunk_size)
-    audio_chunks = []
-
-    temp_audio_full = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
-    audio_full_path = temp_audio_full.name
-    temp_audio_full.close()
-
-    os.system(f"ffmpeg -y -i \"{video_path}\" -q:a 0 -map a \"{audio_full_path}\"")
-
-    for i in range(num_chunks):
-        start = i * chunk_size
-        end = min((i + 1) * chunk_size, duration)
-        temp_chunk = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
-        chunk_path = temp_chunk.name
-        temp_chunk.close()
-        os.system(f"ffmpeg -y -i \"{audio_full_path}\" -ss {start} -to {end} -c copy \"{chunk_path}\"")
-        audio_chunks.append(chunk_path)
-
-    if os.path.exists(audio_full_path):
-        os.unlink(audio_full_path)
-
-    return audio_chunks
+    """Extract audio from video and split into chunks"""
+    try:
+        duration = get_video_duration(video_path)
+        num_chunks = math.ceil(duration / chunk_size)
+        audio_chunks = []
+
+        # extract full audio first
+        temp_audio_full = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
+        audio_full_path = temp_audio_full.name
+        temp_audio_full.close()
+
+        os.system(f"ffmpeg -y -i \"{video_path}\" -q:a 0 -map a \"{audio_full_path}\"")
+
+        # split audio into chunks
+        for i in range(num_chunks):
+            start = i * chunk_size
+            end = min((i + 1) * chunk_size, duration)
+            temp_chunk = tempfile.NamedTemporaryFile(suffix=".mp3", delete=False)
+            chunk_path = temp_chunk.name
+            temp_chunk.close()
+            os.system(f"ffmpeg -y -i \"{audio_full_path}\" -ss {start} -to {end} -c copy \"{chunk_path}\"")
+            audio_chunks.append(chunk_path)
+
+        # remove full audio file
+        if os.path.exists(audio_full_path):
+            os.unlink(audio_full_path)
+
+        return audio_chunks
+    except Exception as e:
+        raise RuntimeError(f"Audio extraction failed: {str(e)}")
 
 # ============================================================================
-# TRANSCRIPTION WITH RETRY & PROGRESS
+# TRANSCRIPTION WITH RETRY
 # ============================================================================
-def transcribe_audio_chunks(audio_chunks, language="English", progress_fn=None):
+def transcribe_audio_chunks(audio_chunks, language="English"):
     transcript_text = ""
-    total_chunks = len(audio_chunks)
-
-    for idx, chunk in enumerate(audio_chunks, start=1):
+    for chunk in audio_chunks:
         for attempt in range(MAX_RETRIES):
             try:
                 with open(chunk, "rb") as audio_file:
@@ -85,7 +92,7 @@ def transcribe_audio_chunks(audio_chunks, language="English", progress_fn=None):
                         language="en" if language=="English" else "ur"
                     )
                 transcript_text += transcription.text + "\n\n"
-                break
+                break  # success
             except Exception as e:
                 err_str = str(e)
                 if "Rate limit reached" in err_str or "429" in err_str:
@@ -95,67 +102,60 @@ def transcribe_audio_chunks(audio_chunks, language="English", progress_fn=None):
                 else:
                     transcript_text += f"[Error in chunk transcription: {err_str}]\n\n"
                     break
-
-        if progress_fn:
-            progress_fn(idx / total_chunks, f"Transcribing audio chunk {idx}/{total_chunks}...")
-
         if os.path.exists(chunk):
             os.unlink(chunk)
-
     return transcript_text.strip()
 
 # ============================================================================
-# GRAMMAR & PUNCTUATION CORRECTION (Chunked) WITH PROGRESS
+# GRAMMAR & PUNCTUATION CORRECTION (Chunked)
 # ============================================================================
-def rewrite_transcript_chunked(text, language="English", chunk_size=REWRITE_CHUNK_SIZE, progress_fn=None):
-    words = text.split()
-    rewritten_chunks = []
-    total_chunks = math.ceil(len(words) / chunk_size)
-
-    for idx in range(0, len(words), chunk_size):
-        chunk_text = " ".join(words[idx:idx+chunk_size])
-        prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{chunk_text}"
-        for attempt in range(MAX_RETRIES):
-            try:
-                response = client.chat.completions.create(
-                    model=REWRITE_MODEL,
-                    messages=[{"role": "user", "content": prompt}],
-                    temperature=0
-                )
-                rewritten_chunks.append(response.choices[0].message.content)
-                break
-            except Exception as e:
-                err_str = str(e)
-                if "Rate limit reached" in err_str or "429" in err_str:
-                    wait_time = 5 + attempt*5
-                    print(f"Rewrite rate limit hit. Waiting {wait_time}s before retrying...")
-                    time.sleep(wait_time)
-                else:
-                    rewritten_chunks.append(f"[Error rewriting chunk: {err_str}]")
+def rewrite_transcript_chunked(text, language="English", chunk_size=REWRITE_CHUNK_SIZE):
+    """Rewrite transcript in smaller chunks to avoid token limits"""
+    try:
+        words = text.split()
+        rewritten_chunks = []
+        for i in range(0, len(words), chunk_size):
+            chunk_text = " ".join(words[i:i+chunk_size])
+            prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{chunk_text}"
+            for attempt in range(MAX_RETRIES):
+                try:
+                    response = client.chat.completions.create(
+                        model=REWRITE_MODEL,
+                        messages=[{"role": "user", "content": prompt}],
+                        temperature=0
+                    )
+                    rewritten_chunks.append(response.choices[0].message.content)
                     break
+                except Exception as e:
+                    err_str = str(e)
+                    if "Rate limit reached" in err_str or "429" in err_str:
+                        wait_time = 5 + attempt*5
+                        print(f"Rewrite rate limit hit. Waiting {wait_time}s before retrying...")
+                        time.sleep(wait_time)
+                    else:
+                        rewritten_chunks.append(f"[Error rewriting chunk: {err_str}]")
+                        break
+        return "\n\n".join(rewritten_chunks)
+    except Exception as e:
+        return f"[Error rewriting transcript: {str(e)}]"
 
-        if progress_fn:
-            progress_fn((idx // chunk_size + 1) / total_chunks, f"Rewriting text chunk {(idx // chunk_size + 1)}/{total_chunks}...")
-
-    return "\n\n".join(rewritten_chunks)
-
-def correct_transcript(transcript_text, language, progress=gr.Progress()):
+def correct_transcript(transcript_text, language):
     lang = "English" if "English" in language else "Urdu"
     if not transcript_text:
         return "", "Please generate transcript first." if lang=="English" else "پہلے ٹرانسکرپٹ بنائیں۔"
-    rewritten_text = rewrite_transcript_chunked(transcript_text, language=lang, progress_fn=progress.update)
+    rewritten_text = rewrite_transcript_chunked(transcript_text, language=lang)
     return rewritten_text, "✅ Grammar & punctuation corrected / گرامر اور پنکچویشن درست کی گئی"
 
 # ============================================================================
 # PROCESS VIDEO
 # ============================================================================
-def generate_transcript(video_file, language, progress=gr.Progress()):
+def generate_transcript(video_file, language):
     lang = "English" if "English" in language else "Urdu"
     if not video_file:
         return "", "Please upload a video file." if lang=="English" else "براہ کرم ایک ویڈیو فائل اپ لوڈ کریں۔"
     try:
         audio_chunks = extract_audio_chunks(video_file.name)
-        transcript = transcribe_audio_chunks(audio_chunks, language=lang, progress_fn=progress.update)
+        transcript = transcribe_audio_chunks(audio_chunks, language=lang)
         return transcript, "✅ Transcription completed / ٹرانسکرپشن مکمل"
     except Exception as e:
         error_msg = str(e)