Samir433 commited on
Commit
fbdeeab
·
verified ·
1 Parent(s): 56b248b

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +30 -268
app.py CHANGED
@@ -1,296 +1,58 @@
1
  # main.py
2
  import streamlit as st
3
- import os
4
- import tempfile
5
- import whisper
6
- import yt_dlp
7
  from dotenv import load_dotenv
8
  from config import Config
9
  from transcript_utils import extract_video_id, fetch_transcript
10
  from vector_utils import build_vector_store, get_retriever
11
  from qa_pipeline import run_qa
12
- from youtube_transcript_api import YouTubeTranscriptApi
13
- from youtube_transcript_api._errors import TranscriptsDisabled, NoTranscriptFound, VideoUnavailable
14
 
15
  # Load environment variables
16
  load_dotenv()
17
 
18
  st.set_page_config(page_title="YouTube Video Chat-QA", layout="centered")
19
- st.title("🎥 YouTube Video Chat-QA App")
20
 
21
  # Initialize session state
22
  if 'ready' not in st.session_state:
23
  st.session_state.ready = False
24
  if 'messages' not in st.session_state:
25
  st.session_state.messages = []
26
- if 'transcript_method' not in st.session_state:
27
- st.session_state.transcript_method = None
28
-
29
- @st.cache_resource
30
- def load_whisper_model():
31
- """Load Whisper model with caching for better performance"""
32
- try:
33
- return whisper.load_model("base")
34
- except Exception as e:
35
- st.error(f"Failed to load Whisper model: {str(e)}")
36
- return None
37
-
38
- def download_audio_from_youtube(video_id):
39
- """Download audio from YouTube video using yt-dlp with better error handling"""
40
- try:
41
- # Create temporary directory
42
- temp_dir = tempfile.mkdtemp()
43
- output_path = os.path.join(temp_dir, f"{video_id}.%(ext)s")
44
-
45
- ydl_opts = {
46
- 'format': 'bestaudio/best',
47
- 'outtmpl': output_path,
48
- 'extractaudio': True,
49
- 'audioformat': 'mp3',
50
- 'noplaylist': True,
51
- 'no_warnings': True,
52
- 'quiet': True,
53
- # Add user agent to avoid bot detection
54
- 'http_headers': {
55
- 'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.124 Safari/537.36'
56
- },
57
- # Add retries for network issues
58
- 'retries': 3,
59
- 'socket_timeout': 30,
60
- }
61
-
62
- with yt_dlp.YoutubeDL(ydl_opts) as ydl:
63
- url = f"https://www.youtube.com/watch?v={video_id}"
64
- info = ydl.extract_info(url, download=True)
65
-
66
- # Find the downloaded file
67
- for file in os.listdir(temp_dir):
68
- if file.startswith(video_id):
69
- return os.path.join(temp_dir, file)
70
-
71
- return None
72
- except Exception as e:
73
- st.error(f"Audio download failed: {str(e)}")
74
- return None
75
-
76
- def transcribe_with_whisper(video_id):
77
- """Fallback transcription using Whisper when YouTube transcript fails"""
78
- try:
79
- # Load Whisper model
80
- model = load_whisper_model()
81
- if not model:
82
- return None
83
-
84
- # Download audio
85
- audio_file = download_audio_from_youtube(video_id)
86
- if not audio_file:
87
- return None
88
-
89
- # Transcribe using Whisper
90
- with st.spinner("🤖 Transcribing audio using AI (this may take a few minutes)..."):
91
- result = model.transcribe(audio_file)
92
-
93
- # Clean up audio file
94
- try:
95
- os.remove(audio_file)
96
- os.rmdir(os.path.dirname(audio_file))
97
- except:
98
- pass
99
-
100
- # Convert to transcript format compatible with your existing code
101
- if result and result.get("text"):
102
- transcript = [{
103
- "text": result["text"],
104
- "start": 0,
105
- "duration": len(result["text"].split()) * 0.5
106
- }]
107
- return transcript
108
-
109
- return None
110
-
111
- except Exception as e:
112
- st.error(f"Whisper transcription failed: {str(e)}")
113
- return None
114
-
115
- def enhanced_fetch_transcript(video_id, languages=None):
116
- """Enhanced transcript fetching with multiple fallback methods"""
117
- transcript = None
118
- method_used = None
119
-
120
- # Method 1: Try YouTube Transcript API first (CORRECTED)
121
- try:
122
- # CORRECT: Call static method directly, not as instance
123
- transcript = YouTubeTranscriptApi.get_transcript(
124
- video_id,
125
- languages=languages or ['en', 'hi', 'auto-generated']
126
- )
127
- method_used = "YouTube API"
128
- st.success("✅ Transcript fetched from YouTube")
129
-
130
- except (TranscriptsDisabled, NoTranscriptFound, VideoUnavailable) as e:
131
- st.warning("⚠️ YouTube transcript not available. Trying alternative method...")
132
-
133
- # Method 2: Try with different language codes
134
- try:
135
- transcript = YouTubeTranscriptApi.get_transcript(video_id, languages=['en'])
136
- method_used = "YouTube API (English only)"
137
- st.success("✅ English transcript found")
138
- except:
139
- pass
140
-
141
- except Exception as e:
142
- st.warning(f"⚠️ YouTube API error: {str(e)}")
143
-
144
- # Method 3: Fallback to Whisper transcription (only if network allows)
145
- if not transcript:
146
- st.info("🔄 Attempting AI transcription as fallback...")
147
- transcript = transcribe_with_whisper(video_id)
148
- if transcript:
149
- method_used = "Whisper AI"
150
- st.success("✅ Transcript generated using AI")
151
-
152
- return transcript, method_used
153
 
154
  # Setup section for video link
155
- with st.expander("🔗 Setup Video", expanded=not st.session_state.ready):
156
- st.markdown("Enter a YouTube video URL to start chatting about its content:")
157
- url = st.text_input("YouTube Video Link", key='input_url', placeholder="https://www.youtube.com/watch?v=...")
158
-
159
- col1, col2 = st.columns(2)
160
- with col1:
161
- if st.button("🚀 Submit Link", type="primary"):
162
- if not url:
163
- st.error("Please enter a YouTube URL")
164
- else:
165
- vid = extract_video_id(url)
166
- if not vid:
167
- st.error("❌ Invalid YouTube URL. Please check and try again.")
168
- else:
169
- st.session_state.video_id = vid
170
- st.session_state.ready = True
171
- st.session_state.messages = []
172
- st.rerun()
173
-
174
- with col2:
175
- if st.button("🔄 Reset"):
176
- for key in ['ready', 'video_id', 'messages', 'input_url', 'transcript_method']:
177
- if key in st.session_state:
178
- del st.session_state[key]
179
- st.rerun()
180
 
181
  # Chat interface after video is ready
182
  def chat_interface():
183
- st.markdown("---")
184
-
185
- # Display video
186
- try:
187
- st.video(f"https://www.youtube.com/watch?v={st.session_state.video_id}")
188
- except Exception as e:
189
- st.error(f"Error displaying video: {str(e)}")
190
-
191
- # Show transcript method if available
192
- if st.session_state.get('transcript_method'):
193
- st.info(f"📝 Transcript source: {st.session_state.transcript_method}")
194
-
195
- # Display chat history first
196
- for msg in st.session_state.messages:
197
- with st.chat_message(msg['role']):
198
- st.write(msg['content'])
199
-
200
  # Input new message
201
- new_q = st.chat_input("💬 Ask a question about this video...")
202
-
203
  if new_q:
204
- # Add user message
205
  st.session_state.messages.append({'role': 'user', 'content': new_q})
206
-
207
- # Display user message immediately
208
- with st.chat_message("user"):
209
- st.write(new_q)
210
-
211
- # Process the question
212
- with st.chat_message("assistant"):
213
- with st.spinner("🔍 Analyzing video content..."):
214
- # Get transcript using enhanced method
215
- transcript, method_used = enhanced_fetch_transcript(
216
- st.session_state.video_id,
217
- languages=["en", "hi"]
218
- )
219
-
220
- # Store the method used
221
- if method_used:
222
- st.session_state.transcript_method = method_used
223
-
224
- if not transcript:
225
- error_msg = """
226
- ❌ **Unable to obtain transcript for this video.**
227
-
228
- This could be due to:
229
- - Network connectivity issues in the deployment environment
230
- - YouTube's restrictions on cloud platforms (common on Hugging Face Spaces)
231
- - Video has no available transcripts/captions
232
- - The video is private or restricted
233
-
234
- **Suggestions:**
235
- - Try a different video with verified captions
236
- - Use videos from educational channels (usually have transcripts)
237
- - Test locally if possible
238
- """
239
- st.markdown(error_msg)
240
- st.session_state.messages.append({
241
- 'role': 'assistant',
242
- 'content': error_msg
243
- })
244
- else:
245
- try:
246
- # Build vector store and get answer
247
- vector_store = build_vector_store(transcript, None, Config.EMBEDDING_MODEL)
248
- retriever = get_retriever(vector_store)
249
- answer = run_qa(retriever, new_q)
250
-
251
- # Display answer
252
- st.write(answer)
253
- st.session_state.messages.append({'role': 'assistant', 'content': answer})
254
-
255
- except Exception as e:
256
- error_msg = f"❌ Error processing your question: {str(e)}"
257
- st.error(error_msg)
258
- st.session_state.messages.append({
259
- 'role': 'assistant',
260
- 'content': error_msg
261
- })
262
 
263
- # Main app logic
264
- if st.session_state.ready:
265
- chat_interface()
266
- else:
267
- # Welcome message
268
- st.markdown("""
269
- ### 👋 Welcome to YouTube Video Chat-QA!
270
-
271
- This app allows you to:
272
- - 📺 Load any YouTube video
273
- - 💬 Ask questions about the video content
274
- - 🤖 Get AI-powered answers based on the video transcript
275
-
276
- **⚠️ Note for Hugging Face Spaces users:**
277
- Due to network restrictions, some features may be limited. The app works best with:
278
- - Educational videos with clear captions
279
- - Popular videos from established channels
280
- - Videos with manual (not auto-generated) transcripts
281
- """)
282
 
283
- # Sidebar with information
284
- with st.sidebar:
285
- st.markdown("### ℹ️ App Status")
286
- st.markdown("""
287
- **Network Status:**
288
- - ⚠️ Limited connectivity (Cloud environment)
289
- - ✅ YouTube Transcript API (when available)
290
- - ❌ Whisper AI (Network restricted)
291
- """)
292
-
293
- if st.session_state.ready:
294
- st.markdown(f"**Current Video ID:** `{st.session_state.video_id}`")
295
- if st.session_state.get('transcript_method'):
296
- st.markdown(f"**Transcript Method:** {st.session_state.transcript_method}")
 
1
  # main.py
2
  import streamlit as st
 
 
 
 
3
  from dotenv import load_dotenv
4
  from config import Config
5
  from transcript_utils import extract_video_id, fetch_transcript
6
  from vector_utils import build_vector_store, get_retriever
7
  from qa_pipeline import run_qa
 
 
8
 
9
  # Load environment variables
10
  load_dotenv()
11
 
12
  st.set_page_config(page_title="YouTube Video Chat-QA", layout="centered")
13
+ st.title("YouTube Video Chat-QA App")
14
 
15
  # Initialize session state
16
  if 'ready' not in st.session_state:
17
  st.session_state.ready = False
18
  if 'messages' not in st.session_state:
19
  st.session_state.messages = []
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
20
 
21
  # Setup section for video link
22
+ with st.expander("Setup Video", expanded=not st.session_state.ready):
23
+ url = st.text_input("YouTube Video Link", key='input_url')
24
+ if st.button("Submit Link"):
25
+ vid = extract_video_id(url)
26
+ if not vid:
27
+ st.error("Invalid YouTube URL.")
28
+ else:
29
+ st.session_state.video_id = vid
30
+ st.session_state.ready = True
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
31
 
32
  # Chat interface after video is ready
33
  def chat_interface():
34
+ st.video(f"https://www.youtube.com/watch?v={st.session_state.video_id}")
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
35
  # Input new message
36
+ new_q = st.chat_input("Ask a question about this video...")
 
37
  if new_q:
 
38
  st.session_state.messages.append({'role': 'user', 'content': new_q})
39
+ transcript = fetch_transcript(st.session_state.video_id, languages=["en", "hi"])
40
+ if not transcript:
41
+ st.session_state.messages.append({'role': 'assistant', 'content': "Transcript not available."})
42
+ else:
43
+ vector_store = build_vector_store(transcript, None, Config.EMBEDDING_MODEL)
44
+ retriever = get_retriever(vector_store)
45
+ answer = run_qa(retriever, new_q)
46
+ st.session_state.messages.append({'role': 'assistant', 'content': answer})
47
+
48
+ # Display chat history
49
+ for msg in st.session_state.messages:
50
+ st.chat_message(msg['role']).write(msg['content'])
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
51
 
52
+ if st.button("Exit Chat"):
53
+ for key in ['ready', 'video_id', 'messages', 'input_url']:
54
+ if key in st.session_state:
55
+ del st.session_state[key]
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
56
 
57
+ if st.session_state.ready:
58
+ chat_interface()