Spaces:
Runtime error
Runtime error
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,156 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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 google.generativeai import upload_file, get_file
|
| 6 |
+
import google.generativeai as genai
|
| 7 |
+
import time
|
| 8 |
+
from pathlib import Path
|
| 9 |
+
import tempfile
|
| 10 |
+
from dotenv import load_dotenv
|
| 11 |
+
import os
|
| 12 |
+
from phi.model.groq import Groq
|
| 13 |
+
from phi.tools.youtube_tools import YouTubeTools
|
| 14 |
+
|
| 15 |
+
# Load environment variables
|
| 16 |
+
load_dotenv()
|
| 17 |
+
|
| 18 |
+
# Configure API keys
|
| 19 |
+
API_KEY = os.getenv("GOOGLE_API_KEY")
|
| 20 |
+
groq_api_key = os.getenv("GROQ_API_KEY")
|
| 21 |
+
if API_KEY:
|
| 22 |
+
genai.configure(api_key=API_KEY)
|
| 23 |
+
|
| 24 |
+
# Page configuration
|
| 25 |
+
st.set_page_config(
|
| 26 |
+
page_title="Multimodal AI Applications",
|
| 27 |
+
page_icon="🌐",
|
| 28 |
+
layout="wide"
|
| 29 |
+
)
|
| 30 |
+
|
| 31 |
+
# Tabs for the two applications
|
| 32 |
+
tab1, tab2 = st.tabs(["Video Summarizer", "YouTube Video Analyzer"])
|
| 33 |
+
|
| 34 |
+
# Tab 1: Video Summarizer with Gemini
|
| 35 |
+
with tab1:
|
| 36 |
+
st.title("Phidata Video AI Summarizer Agent 🎥🎤🖬")
|
| 37 |
+
st.header("Powered by Gemini 2.0 Flash Exp")
|
| 38 |
+
|
| 39 |
+
@st.cache_resource
|
| 40 |
+
def initialize_agent():
|
| 41 |
+
return Agent(
|
| 42 |
+
name="Video AI Summarizer",
|
| 43 |
+
model=Gemini(id="gemini-2.0-flash-exp"),
|
| 44 |
+
tools=[DuckDuckGo()],
|
| 45 |
+
markdown=True,
|
| 46 |
+
)
|
| 47 |
+
|
| 48 |
+
multimodal_Agent = initialize_agent()
|
| 49 |
+
|
| 50 |
+
video_file = st.file_uploader(
|
| 51 |
+
"Upload a video file",
|
| 52 |
+
type=['mp4'],
|
| 53 |
+
help="Upload a video for AI analysis"
|
| 54 |
+
)
|
| 55 |
+
|
| 56 |
+
if video_file:
|
| 57 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
|
| 58 |
+
temp_video.write(video_file.read())
|
| 59 |
+
video_path = temp_video.name
|
| 60 |
+
|
| 61 |
+
st.video(video_path, format="video/mp4", start_time=0)
|
| 62 |
+
|
| 63 |
+
user_query = st.text_area(
|
| 64 |
+
"What insights are you seeking from the video?",
|
| 65 |
+
placeholder="Ask anything about the video content. The AI agent will analyze and gather additional context if needed. This AI agent can search the internet.",
|
| 66 |
+
help="Provide specific questions or insights you want from the video."
|
| 67 |
+
)
|
| 68 |
+
|
| 69 |
+
if st.button("🔍 Analyze Video", key="analyze_video_button"):
|
| 70 |
+
if not user_query:
|
| 71 |
+
st.warning("Please enter a question or insight to analyze the video.")
|
| 72 |
+
else:
|
| 73 |
+
try:
|
| 74 |
+
with st.spinner("Processing video and gathering insights..."):
|
| 75 |
+
processed_video = upload_file(video_path)
|
| 76 |
+
|
| 77 |
+
while processed_video.state.name == "PROCESSING":
|
| 78 |
+
time.sleep(1)
|
| 79 |
+
processed_video = get_file(processed_video.name)
|
| 80 |
+
|
| 81 |
+
analysis_prompt = (
|
| 82 |
+
f"""
|
| 83 |
+
Analyze the uploaded video for content and context. Provide a detailed summary. Search the internet for similar topics discussed in the uploaded video.
|
| 84 |
+
Respond to the following query using video insights and supplementary web research: {user_query}
|
| 85 |
+
Provide a detailed, user-friendly, and actionable response.
|
| 86 |
+
"""
|
| 87 |
+
)
|
| 88 |
+
|
| 89 |
+
response = multimodal_Agent.run(analysis_prompt, videos=[processed_video])
|
| 90 |
+
|
| 91 |
+
st.subheader("Analysis Result")
|
| 92 |
+
st.markdown(response.content)
|
| 93 |
+
|
| 94 |
+
except Exception as error:
|
| 95 |
+
st.error(f"An error occurred during analysis: {error}")
|
| 96 |
+
|
| 97 |
+
finally:
|
| 98 |
+
Path(video_path).unlink(missing_ok=True)
|
| 99 |
+
else:
|
| 100 |
+
st.info("Upload a video file to begin analysis.")
|
| 101 |
+
|
| 102 |
+
# Tab 2: YouTube Video Analyzer with Groq
|
| 103 |
+
with tab2:
|
| 104 |
+
st.title("YouTube Video Analyzer")
|
| 105 |
+
st.header("Analyze Youtube video using video link 🎬 🧷")
|
| 106 |
+
|
| 107 |
+
try:
|
| 108 |
+
youtube_agent = Agent(
|
| 109 |
+
model=Groq(id="llama3-8b-8192", api_key=groq_api_key),
|
| 110 |
+
tools=[YouTubeTools(), DuckDuckGo()],
|
| 111 |
+
show_tool_calls=True,
|
| 112 |
+
get_video_captions=True,
|
| 113 |
+
get_video_data = True,
|
| 114 |
+
description="You are a YouTube agent. Obtain the captions of a YouTube video, analyze its content and context, provide detailed summaries with key points, conduct web research on similar topics, and answer user questions.",
|
| 115 |
+
)
|
| 116 |
+
except Exception as e:
|
| 117 |
+
st.error(f"Error initializing the agent: {e}")
|
| 118 |
+
st.stop()
|
| 119 |
+
|
| 120 |
+
video_url = st.text_input("Enter YouTube Video URL:", "")
|
| 121 |
+
user_query = st.text_area("Enter your question about the video (optional):", "")
|
| 122 |
+
|
| 123 |
+
if st.button("✨ Analyze the Video", key="Analyze_Video_Button"):
|
| 124 |
+
if video_url:
|
| 125 |
+
with st.spinner("Analyzing..."):
|
| 126 |
+
try:
|
| 127 |
+
prompt_prefix = """Analyze the YouTube video provided by the URL. Provide a detailed summary of the video, including key points.
|
| 128 |
+
Structure the output as follows:
|
| 129 |
+
**Detailed Summary:**
|
| 130 |
+
[Detailed summary of the video content]
|
| 131 |
+
**Key Points:**
|
| 132 |
+
* [Key point 1]
|
| 133 |
+
* [Key point 2]
|
| 134 |
+
* [Key point 3]
|
| 135 |
+
... (and so on)
|
| 136 |
+
Conduct web research on similar topics discussed in the video to provide additional context.
|
| 137 |
+
"""
|
| 138 |
+
|
| 139 |
+
if user_query:
|
| 140 |
+
prompt = f"""{prompt_prefix}
|
| 141 |
+
Respond to the following query using video insights and supplementary web research:
|
| 142 |
+
{user_query}
|
| 143 |
+
Provide a detailed, user-friendly, and actionable response.
|
| 144 |
+
Video URL: {video_url}"""
|
| 145 |
+
else:
|
| 146 |
+
prompt = f"""{prompt_prefix}
|
| 147 |
+
Video URL: {video_url}"""
|
| 148 |
+
|
| 149 |
+
output = youtube_agent.run(prompt)
|
| 150 |
+
|
| 151 |
+
st.markdown(output.content)
|
| 152 |
+
|
| 153 |
+
except Exception as e:
|
| 154 |
+
st.error(f"Error analyzing the video: {e}")
|
| 155 |
+
else:
|
| 156 |
+
st.warning("Please enter a YouTube video URL.")
|