Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,20 @@
|
|
1 |
-
import streamlit as st
|
2 |
from phi.agent import Agent
|
3 |
from phi.model.google import Gemini
|
4 |
from phi.tools.duckduckgo import DuckDuckGo
|
5 |
-
from
|
|
|
6 |
import google.generativeai as genai
|
7 |
|
8 |
import time
|
9 |
from pathlib import Path
|
10 |
-
|
11 |
import tempfile
|
12 |
-
|
13 |
from dotenv import load_dotenv
|
14 |
load_dotenv()
|
15 |
|
16 |
import os
|
17 |
|
18 |
-
API_KEY=os.getenv("GOOGLE_API_KEY")
|
19 |
if API_KEY:
|
20 |
genai.configure(api_key=API_KEY)
|
21 |
|
@@ -29,24 +28,44 @@ st.set_page_config(
|
|
29 |
st.title("Phidata Video AI Summarizer Agent π₯π€π¬")
|
30 |
st.header("Powered by Gemini 2.0 Flash Exp")
|
31 |
|
32 |
-
|
33 |
@st.cache_resource
|
34 |
def initialize_agent():
|
35 |
return Agent(
|
36 |
name="Video AI Summarizer",
|
37 |
model=Gemini(id="gemini-2.0-flash-exp"),
|
38 |
-
tools=[DuckDuckGo()],
|
39 |
markdown=True,
|
40 |
)
|
41 |
|
42 |
-
|
43 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
|
45 |
-
#
|
46 |
video_file = st.file_uploader(
|
47 |
"Upload a video file", type=['mp4', 'mov', 'avi'], help="Upload a video for AI analysis"
|
48 |
)
|
49 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
50 |
if video_file:
|
51 |
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
|
52 |
temp_video.write(video_file.read())
|
@@ -54,13 +73,7 @@ if video_file:
|
|
54 |
|
55 |
st.video(video_path, format="video/mp4", start_time=0)
|
56 |
|
57 |
-
|
58 |
-
"What insights are you seeking from the video?",
|
59 |
-
placeholder="Ask anything about the video content. The AI agent will analyze and gather additional context if needed.",
|
60 |
-
help="Provide specific questions or insights you want from the video."
|
61 |
-
)
|
62 |
-
|
63 |
-
if st.button("π Analyze Video", key="analyze_video_button"):
|
64 |
if not user_query:
|
65 |
st.warning("Please enter a question or insight to analyze the video.")
|
66 |
else:
|
@@ -72,16 +85,8 @@ if video_file:
|
|
72 |
time.sleep(1)
|
73 |
processed_video = get_file(processed_video.name)
|
74 |
|
75 |
-
#
|
76 |
-
analysis_prompt = (
|
77 |
-
f"""
|
78 |
-
Analyze the uploaded video for content and context.
|
79 |
-
Respond to the following query using video insights and supplementary web research:
|
80 |
-
{user_query}
|
81 |
-
|
82 |
-
Provide a detailed, user-friendly, and actionable response.
|
83 |
-
"""
|
84 |
-
)
|
85 |
|
86 |
# AI agent processing
|
87 |
response = multimodal_Agent.run(analysis_prompt, videos=[processed_video])
|
@@ -95,8 +100,30 @@ if video_file:
|
|
95 |
finally:
|
96 |
# Clean up temporary video file
|
97 |
Path(video_path).unlink(missing_ok=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
98 |
else:
|
99 |
-
st.info("Upload a video file to begin analysis.")
|
100 |
|
101 |
# Customize text area height
|
102 |
st.markdown(
|
@@ -109,5 +136,3 @@ st.markdown(
|
|
109 |
""",
|
110 |
unsafe_allow_html=True
|
111 |
)
|
112 |
-
|
113 |
-
|
|
|
1 |
+
import streamlit as st
|
2 |
from phi.agent import Agent
|
3 |
from phi.model.google import Gemini
|
4 |
from phi.tools.duckduckgo import DuckDuckGo
|
5 |
+
from phi.tools.youtube_tools import YouTubeTools # Import YouTubeTools for YouTube summarization
|
6 |
+
from google.generativeai import upload_file, get_file
|
7 |
import google.generativeai as genai
|
8 |
|
9 |
import time
|
10 |
from pathlib import Path
|
|
|
11 |
import tempfile
|
|
|
12 |
from dotenv import load_dotenv
|
13 |
load_dotenv()
|
14 |
|
15 |
import os
|
16 |
|
17 |
+
API_KEY = os.getenv("GOOGLE_API_KEY")
|
18 |
if API_KEY:
|
19 |
genai.configure(api_key=API_KEY)
|
20 |
|
|
|
28 |
st.title("Phidata Video AI Summarizer Agent π₯π€π¬")
|
29 |
st.header("Powered by Gemini 2.0 Flash Exp")
|
30 |
|
31 |
+
# Initialize the agent
|
32 |
@st.cache_resource
|
33 |
def initialize_agent():
|
34 |
return Agent(
|
35 |
name="Video AI Summarizer",
|
36 |
model=Gemini(id="gemini-2.0-flash-exp"),
|
37 |
+
tools=[DuckDuckGo(), YouTubeTools()], # Add YouTubeTools here for YouTube video summarization
|
38 |
markdown=True,
|
39 |
)
|
40 |
|
41 |
+
multimodal_Agent = initialize_agent()
|
42 |
+
|
43 |
+
# Input field for user query
|
44 |
+
user_query = st.text_area(
|
45 |
+
"What insights are you seeking from the video?",
|
46 |
+
placeholder="Ask anything about the video content. The AI agent will analyze and gather additional context if needed.",
|
47 |
+
help="Provide specific questions or insights you want from the video."
|
48 |
+
)
|
49 |
|
50 |
+
# Check for either video file upload or YouTube URL
|
51 |
video_file = st.file_uploader(
|
52 |
"Upload a video file", type=['mp4', 'mov', 'avi'], help="Upload a video for AI analysis"
|
53 |
)
|
54 |
|
55 |
+
youtube_url = st.text_input("Or, paste a YouTube URL to summarize:")
|
56 |
+
|
57 |
+
# Unified analysis prompt
|
58 |
+
def generate_analysis_prompt(user_query, video_source):
|
59 |
+
return (
|
60 |
+
f"""
|
61 |
+
Analyze the {video_source} for content and context.
|
62 |
+
Respond to the following query using video insights and supplementary web research:
|
63 |
+
{user_query}
|
64 |
+
Provide a detailed, user-friendly, and actionable response.
|
65 |
+
"""
|
66 |
+
)
|
67 |
+
|
68 |
+
# Process uploaded video
|
69 |
if video_file:
|
70 |
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
|
71 |
temp_video.write(video_file.read())
|
|
|
73 |
|
74 |
st.video(video_path, format="video/mp4", start_time=0)
|
75 |
|
76 |
+
if st.button("π Analyze Uploaded Video", key="analyze_uploaded_video_button"):
|
|
|
|
|
|
|
|
|
|
|
|
|
77 |
if not user_query:
|
78 |
st.warning("Please enter a question or insight to analyze the video.")
|
79 |
else:
|
|
|
85 |
time.sleep(1)
|
86 |
processed_video = get_file(processed_video.name)
|
87 |
|
88 |
+
# Use the unified prompt for analysis (Uploaded video)
|
89 |
+
analysis_prompt = generate_analysis_prompt(user_query, "uploaded video")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
|
91 |
# AI agent processing
|
92 |
response = multimodal_Agent.run(analysis_prompt, videos=[processed_video])
|
|
|
100 |
finally:
|
101 |
# Clean up temporary video file
|
102 |
Path(video_path).unlink(missing_ok=True)
|
103 |
+
|
104 |
+
# Process YouTube URL
|
105 |
+
elif youtube_url:
|
106 |
+
if st.button("π Summarize YouTube Video", key="summarize_youtube_button"):
|
107 |
+
if not user_query:
|
108 |
+
st.warning("Please enter a question or insight to analyze the YouTube video.")
|
109 |
+
else:
|
110 |
+
try:
|
111 |
+
with st.spinner("Fetching video captions and summarizing..."):
|
112 |
+
# Use the unified prompt for summarization (YouTube URL)
|
113 |
+
analysis_prompt = generate_analysis_prompt(user_query, "YouTube URL")
|
114 |
+
|
115 |
+
# Summarize the YouTube video using the agent
|
116 |
+
response = multimodal_Agent.get_response(f"{analysis_prompt} {youtube_url}")
|
117 |
+
|
118 |
+
# Display the result
|
119 |
+
st.subheader("Video Summary")
|
120 |
+
st.markdown(response["content"], unsafe_allow_html=True)
|
121 |
+
|
122 |
+
except Exception as error:
|
123 |
+
st.error(f"An error occurred: {error}")
|
124 |
+
|
125 |
else:
|
126 |
+
st.info("Upload a video file or paste a YouTube URL to begin analysis.")
|
127 |
|
128 |
# Customize text area height
|
129 |
st.markdown(
|
|
|
136 |
""",
|
137 |
unsafe_allow_html=True
|
138 |
)
|
|
|
|