Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -11,38 +11,54 @@ def get_api_key():
|
|
| 11 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY", "")
|
| 12 |
if not GOOGLE_API_KEY:
|
| 13 |
GOOGLE_API_KEY = st.text_input("Enter your Gemini API key", type="password")
|
| 14 |
-
return GOOGLE_API_KEY or "AIzaSyA8TTu9s6fJDG9RlMwOyHFxg270xLgpiyE"
|
| 15 |
|
| 16 |
# VideoProcessor class
|
| 17 |
class VideoProcessor:
|
| 18 |
def __init__(self, api_key):
|
| 19 |
genai.configure(api_key=api_key)
|
| 20 |
-
self.model = genai.GenerativeModel("gemini-2.0-flash")
|
| 21 |
|
| 22 |
def upload_video(self, video_path, display_name="uploaded_video"):
|
| 23 |
-
|
|
|
|
|
|
|
|
|
|
| 24 |
|
| 25 |
def wait_for_processing(self, video_file):
|
| 26 |
-
|
|
|
|
|
|
|
| 27 |
time.sleep(2)
|
| 28 |
video_file = genai.get_file(video_file.name)
|
|
|
|
| 29 |
if video_file.state.name == "FAILED":
|
| 30 |
raise RuntimeError("Video processing failed")
|
|
|
|
|
|
|
| 31 |
return video_file
|
| 32 |
|
| 33 |
def chat_with_video(self, video_file, prompt):
|
| 34 |
-
|
| 35 |
-
|
| 36 |
-
|
| 37 |
-
|
| 38 |
-
|
| 39 |
-
|
| 40 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 41 |
|
| 42 |
# Main app function
|
| 43 |
def main():
|
| 44 |
st.set_page_config(page_title="Video Retrieval-Augmented Generation", page_icon="π¬", layout="wide")
|
| 45 |
-
st.header("Video Retrieval-Augmented Generation - Gemini 2.0")
|
| 46 |
st.markdown("---")
|
| 47 |
|
| 48 |
# Step 1: API Key input
|
|
@@ -58,59 +74,92 @@ def main():
|
|
| 58 |
uploaded_file = st.file_uploader("Upload a video", type=['mp4', 'mov', 'avi', 'mkv', 'webm'])
|
| 59 |
|
| 60 |
if uploaded_file:
|
| 61 |
-
|
|
|
|
|
|
|
| 62 |
file_size = len(uploaded_file.getvalue()) / (1024**2)
|
| 63 |
-
|
| 64 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 65 |
if st.session_state.video_name != uploaded_file.name:
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
|
| 80 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
st.video(uploaded_file.getvalue())
|
| 82 |
else:
|
| 83 |
-
st.error("
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 90 |
|
| 91 |
# Step 3: Chat about Video
|
| 92 |
st.subheader("Step 3: Chat with your video")
|
| 93 |
|
| 94 |
if st.session_state.video_file:
|
|
|
|
| 95 |
for msg in st.session_state.messages:
|
| 96 |
with st.chat_message(msg["role"]):
|
| 97 |
st.markdown(msg["content"])
|
| 98 |
|
| 99 |
-
|
| 100 |
-
|
| 101 |
-
st.session_state.messages.append({"role": "user", "content": user_question})
|
| 102 |
-
with st.chat_message("user"):
|
| 103 |
-
st.markdown(user_question)
|
| 104 |
-
|
| 105 |
-
with st.chat_message("assistant"):
|
| 106 |
-
placeholder = st.empty()
|
| 107 |
-
with st.spinner("Generating response..."):
|
| 108 |
-
response = st.session_state.video_processor.chat_with_video(st.session_state.video_file, user_question)
|
| 109 |
-
|
| 110 |
-
placeholder.markdown(response)
|
| 111 |
-
st.session_state.messages.append({"role": "assistant", "content": response})
|
| 112 |
-
else:
|
| 113 |
-
st.info("Please upload a video in step 2 to start chatting.")
|
| 114 |
-
|
| 115 |
-
if __name__ == "__main__":
|
| 116 |
-
main()
|
|
|
|
| 11 |
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY", "")
|
| 12 |
if not GOOGLE_API_KEY:
|
| 13 |
GOOGLE_API_KEY = st.text_input("Enter your Gemini API key", type="password")
|
| 14 |
+
return GOOGLE_API_KEY or "AIzaSyA8TTu9s6fJDG9RlMwOyHFxg270xLgpiyE" # Warning: Hardcoded key
|
| 15 |
|
| 16 |
# VideoProcessor class
|
| 17 |
class VideoProcessor:
|
| 18 |
def __init__(self, api_key):
|
| 19 |
genai.configure(api_key=api_key)
|
| 20 |
+
self.model = genai.GenerativeModel("gemini-2.0-flash-exp")
|
| 21 |
|
| 22 |
def upload_video(self, video_path, display_name="uploaded_video"):
|
| 23 |
+
try:
|
| 24 |
+
return genai.upload_file(path=video_path, display_name=display_name)
|
| 25 |
+
except Exception as e:
|
| 26 |
+
raise RuntimeError(f"Failed to upload video: {str(e)}")
|
| 27 |
|
| 28 |
def wait_for_processing(self, video_file):
|
| 29 |
+
max_attempts = 60 # Maximum wait time: 2 minutes
|
| 30 |
+
attempts = 0
|
| 31 |
+
while video_file.state.name == "PROCESSING" and attempts < max_attempts:
|
| 32 |
time.sleep(2)
|
| 33 |
video_file = genai.get_file(video_file.name)
|
| 34 |
+
attempts += 1
|
| 35 |
if video_file.state.name == "FAILED":
|
| 36 |
raise RuntimeError("Video processing failed")
|
| 37 |
+
if attempts >= max_attempts:
|
| 38 |
+
raise RuntimeError("Video processing timeout")
|
| 39 |
return video_file
|
| 40 |
|
| 41 |
def chat_with_video(self, video_file, prompt):
|
| 42 |
+
try:
|
| 43 |
+
response = self.model.generate_content([video_file, prompt])
|
| 44 |
+
return response.text
|
| 45 |
+
except Exception as e:
|
| 46 |
+
return f"Error generating response: {str(e)}"
|
| 47 |
+
|
| 48 |
+
# Initialize session state properly
|
| 49 |
+
if "messages" not in st.session_state:
|
| 50 |
+
st.session_state.messages = []
|
| 51 |
+
if "video_processor" not in st.session_state:
|
| 52 |
+
st.session_state.video_processor = None
|
| 53 |
+
if "video_file" not in st.session_state:
|
| 54 |
+
st.session_state.video_file = None
|
| 55 |
+
if "video_name" not in st.session_state:
|
| 56 |
+
st.session_state.video_name = None
|
| 57 |
|
| 58 |
# Main app function
|
| 59 |
def main():
|
| 60 |
st.set_page_config(page_title="Video Retrieval-Augmented Generation", page_icon="π¬", layout="wide")
|
| 61 |
+
st.header("π¬ Video Retrieval-Augmented Generation - Gemini 2.0")
|
| 62 |
st.markdown("---")
|
| 63 |
|
| 64 |
# Step 1: API Key input
|
|
|
|
| 74 |
uploaded_file = st.file_uploader("Upload a video", type=['mp4', 'mov', 'avi', 'mkv', 'webm'])
|
| 75 |
|
| 76 |
if uploaded_file:
|
| 77 |
+
# Validate video file
|
| 78 |
+
mime_type = mimetypes.guess_type(uploaded_file.name)[0]
|
| 79 |
+
if mime_type and mime_type.startswith("video"):
|
| 80 |
file_size = len(uploaded_file.getvalue()) / (1024**2)
|
| 81 |
+
|
| 82 |
+
# Display file info
|
| 83 |
+
col1, col2 = st.columns(2)
|
| 84 |
+
with col1:
|
| 85 |
+
st.info(f"π File: {uploaded_file.name}")
|
| 86 |
+
with col2:
|
| 87 |
+
st.info(f"πΎ Size: {file_size:.2f} MB")
|
| 88 |
+
|
| 89 |
+
# Check file size limit (200MB for Gemini)
|
| 90 |
+
if file_size > 200:
|
| 91 |
+
st.error("β File too large! Maximum size is 200MB")
|
| 92 |
+
st.stop()
|
| 93 |
+
|
| 94 |
+
# Process new video if different from current
|
| 95 |
if st.session_state.video_name != uploaded_file.name:
|
| 96 |
+
tmp_path = None
|
| 97 |
+
try:
|
| 98 |
+
# Initialize processor if needed
|
| 99 |
+
if not st.session_state.video_processor:
|
| 100 |
+
st.session_state.video_processor = VideoProcessor(api_key)
|
| 101 |
+
|
| 102 |
+
# Create temporary file
|
| 103 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=Path(uploaded_file.name).suffix) as tmp:
|
| 104 |
+
tmp.write(uploaded_file.getvalue())
|
| 105 |
+
tmp_path = tmp.name
|
| 106 |
+
|
| 107 |
+
# Upload and process with progress indication
|
| 108 |
+
with st.spinner("π€ Uploading and processing video..."):
|
| 109 |
+
progress_bar = st.progress(0)
|
| 110 |
+
progress_bar.progress(25, text="Uploading video...")
|
| 111 |
+
|
| 112 |
+
video_file = st.session_state.video_processor.upload_video(tmp_path, uploaded_file.name)
|
| 113 |
+
progress_bar.progress(50, text="Processing video...")
|
| 114 |
+
|
| 115 |
+
processed_file = st.session_state.video_processor.wait_for_processing(video_file)
|
| 116 |
+
progress_bar.progress(100, text="Complete!")
|
| 117 |
+
|
| 118 |
+
# Update session state
|
| 119 |
+
st.session_state.video_file = processed_file
|
| 120 |
+
st.session_state.video_name = uploaded_file.name
|
| 121 |
+
st.session_state.messages = [] # Clear previous conversation
|
| 122 |
+
|
| 123 |
+
st.success("β
Video processed successfully!")
|
| 124 |
+
time.sleep(1) # Show success message briefly
|
| 125 |
+
progress_bar.empty() # Clear progress bar
|
| 126 |
+
|
| 127 |
+
except Exception as e:
|
| 128 |
+
st.error(f"β Error processing video: {str(e)}")
|
| 129 |
+
st.session_state.video_file = None
|
| 130 |
+
st.session_state.video_name = None
|
| 131 |
+
|
| 132 |
+
finally:
|
| 133 |
+
# Clean up temporary file
|
| 134 |
+
if tmp_path and os.path.exists(tmp_path):
|
| 135 |
+
os.unlink(tmp_path)
|
| 136 |
+
|
| 137 |
+
# Display video player
|
| 138 |
st.video(uploaded_file.getvalue())
|
| 139 |
else:
|
| 140 |
+
st.error("β Please upload a valid video file")
|
| 141 |
+
|
| 142 |
+
# Control buttons
|
| 143 |
+
col1, col2 = st.columns(2)
|
| 144 |
+
with col1:
|
| 145 |
+
if st.button("π Reset Chat", disabled=not st.session_state.messages):
|
| 146 |
+
st.session_state.messages = []
|
| 147 |
+
st.rerun()
|
| 148 |
+
|
| 149 |
+
with col2:
|
| 150 |
+
if st.button("ποΈ Reset All", disabled=not st.session_state.video_file):
|
| 151 |
+
for key in list(st.session_state.keys()):
|
| 152 |
+
del st.session_state[key]
|
| 153 |
+
st.rerun()
|
| 154 |
|
| 155 |
# Step 3: Chat about Video
|
| 156 |
st.subheader("Step 3: Chat with your video")
|
| 157 |
|
| 158 |
if st.session_state.video_file:
|
| 159 |
+
# Display chat history
|
| 160 |
for msg in st.session_state.messages:
|
| 161 |
with st.chat_message(msg["role"]):
|
| 162 |
st.markdown(msg["content"])
|
| 163 |
|
| 164 |
+
# Chat input
|
| 165 |
+
user_question = st.chat_input("Ask a question ab
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|