Update app.py

app.py

@@ -1,6 +1,6 @@
 """
-AI Graduate Job-Matching Platform
-Complete Hugging Face Spaces Deployment Version
+Video Transcription APP
+Production-Safe Version with Retry, Chunk Handling & Progress Bar
 Developer: Najaf Ali Sharqi
 MoviePy-free (uses ffmpeg + tempfile)
 """
@@ -10,23 +10,23 @@ import gradio as gr
 import tempfile
 import subprocess
 import math
+import time
 from groq import Groq
 
 # ============================================================================
 # CONFIGURATION
 # ============================================================================
-
 client = Groq(api_key=os.getenv("GROQ_API_KEY"))
 
-TRANSCRIPTION_MODEL = "whisper-large-v3"  # transcription model
+TRANSCRIPTION_MODEL = "whisper-large-v3"
 REWRITE_MODEL = "openai/gpt-oss-120b"
-CHUNK_SIZE = 600  # 10 min in seconds for audio splitting
-REWRITE_CHUNK_WORDS = 4000  # number of words per GPT rewrite chunk
+CHUNK_SIZE = 300  # 5 min chunks for transcription
+REWRITE_CHUNK_SIZE = 4000  # words per rewrite chunk
+MAX_RETRIES = 5  # retries for 429 errors
 
 # ============================================================================
 # UTILITY FUNCTIONS
 # ============================================================================
-
 def get_video_duration(video_path):
     try:
         result = subprocess.run(
@@ -37,108 +37,125 @@ def get_video_duration(video_path):
             stderr=subprocess.STDOUT,
             text=True
         )
-        duration = float(result.stdout.strip())
-        return duration
+        return float(result.stdout.strip())
     except Exception as e:
         raise RuntimeError(f"Failed to get video duration: {str(e)}")
 
 def extract_audio_chunks(video_path, chunk_size=CHUNK_SIZE):
-    try:
-        duration = get_video_duration(video_path)
-        num_chunks = math.ceil(duration / chunk_size)
-        audio_chunks = []
+    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()
+    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}\"")
+    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)
+    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)
+    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)}")
+    return audio_chunks
 
 # ============================================================================
-# TRANSCRIPTION
+# TRANSCRIPTION WITH RETRY & PROGRESS
 # ============================================================================
-
-def transcribe_audio_chunks(audio_chunks, language="English"):
+def transcribe_audio_chunks(audio_chunks, language="English", progress_fn=None):
     transcript_text = ""
-    for chunk in audio_chunks:
-        try:
-            with open(chunk, "rb") as audio_file:
-                transcription = client.audio.transcriptions.create(
-                    file=(os.path.basename(chunk), audio_file.read()),
-                    model=TRANSCRIPTION_MODEL,
-                    temperature=0,
-                    response_format="verbose_json",
-                    language="en" if language=="English" else "ur"
-                )
-            transcript_text += transcription.text + "\n\n"
-        except Exception as e:
-            transcript_text += f"[Error in chunk transcription: {str(e)}]\n\n"
-        finally:
-            if os.path.exists(chunk):
-                os.unlink(chunk)
+    total_chunks = len(audio_chunks)
+
+    for idx, chunk in enumerate(audio_chunks, start=1):
+        for attempt in range(MAX_RETRIES):
+            try:
+                with open(chunk, "rb") as audio_file:
+                    transcription = client.audio.transcriptions.create(
+                        file=(os.path.basename(chunk), audio_file.read()),
+                        model=TRANSCRIPTION_MODEL,
+                        temperature=0,
+                        response_format="verbose_json",
+                        language="en" if language=="English" else "ur"
+                    )
+                transcript_text += transcription.text + "\n\n"
+                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"Rate limit hit. Waiting {wait_time}s before retrying...")
+                    time.sleep(wait_time)
+                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)
+# GRAMMAR & PUNCTUATION CORRECTION (Chunked) WITH PROGRESS
 # ============================================================================
+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)
 
-def rewrite_transcript_chunked(text, language="English", chunk_words=REWRITE_CHUNK_WORDS):
-    """
-    Rewrite transcript in smaller word-based chunks to avoid token limits
-    """
-    try:
-        words = text.split()
-        rewritten_chunks = []
-
-        for i in range(0, len(words), chunk_words):
-            chunk_text = " ".join(words[i:i+chunk_words])
-            prompt = f"Correct grammar, punctuation, and make it readable. Language: {language}\n\n{chunk_text}"
-            response = client.chat.completions.create(
-                model=REWRITE_MODEL,
-                messages=[{"role": "user", "content": prompt}],
-                temperature=0
-            )
-            rewritten_chunks.append(response.choices[0].message.content)
-
-        rewritten_text = "\n\n".join(rewritten_chunks)
-        return rewritten_text
-    except Exception as e:
-        return f"[Error rewriting transcript: {str(e)}]"
-
-def correct_transcript(transcript_text, language):
+    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}]")
+                    break
+
+        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()):
     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)
+    rewritten_text = rewrite_transcript_chunked(transcript_text, language=lang, progress_fn=progress.update)
     return rewritten_text, "✅ Grammar & punctuation corrected / گرامر اور پنکچویشن درست کی گئی"
 
 # ============================================================================
 # PROCESS VIDEO
 # ============================================================================
-
-def generate_transcript(video_file, language):
+def generate_transcript(video_file, language, progress=gr.Progress()):
     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)
+        transcript = transcribe_audio_chunks(audio_chunks, language=lang, progress_fn=progress.update)
         return transcript, "✅ Transcription completed / ٹرانسکرپشن مکمل"
     except Exception as e:
         error_msg = str(e)
@@ -147,7 +164,6 @@ def generate_transcript(video_file, language):
 # ============================================================================
 # DOWNLOAD FUNCTION
 # ============================================================================
-
 def download_transcript(text, language):
     lang_suffix = "en" if "English" in language else "ur"
     temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=f"_{lang_suffix}.txt")
@@ -158,7 +174,6 @@ def download_transcript(text, language):
 # ============================================================================
 # GRADIO UI
 # ============================================================================
-
 def create_ui():
     with gr.Blocks(title="AI Graduate Job Matcher") as demo:
         gr.HTML("<h1>🎓 AI Graduate Job-Matching Platform</h1><p>Developer: Najaf Ali Sharqi</p>")
@@ -200,7 +215,6 @@ def create_ui():
 # ============================================================================
 # MAIN
 # ============================================================================
-
 if __name__ == "__main__":
     if not os.getenv("GROQ_API_KEY"):
         print("⚠️ WARNING: GROQ_API_KEY not found!")