app.py

import streamlit as st
from phi.agent import Agent
from phi.model.google import Gemini
from phi.tools.duckduckgo import DuckDuckGo
from google.generativeai import upload_file, get_file
import google.generativeai as genai
import time
from pathlib import Path
import tempfile
from dotenv import load_dotenv
import os
from pytube import YouTube  # Import pytube for downloading YouTube videos

load_dotenv()

API_KEY = os.getenv("GOOGLE_API_KEY")
if API_KEY:
    genai.configure(api_key=API_KEY)

# Page configuration
st.set_page_config(
    page_title="Multimodal AI Agent- Video Summarizer",
    page_icon="🎥",
    layout="wide"
)

st.title("Phidata Video AI Summarizer Agent 🎥🎤🖬")
st.header("Powered by Gemini 2.0 Flash Exp")


@st.cache_resource
def initialize_agent():
    return Agent(
        name="Video AI Summarizer",
        model=Gemini(id="gemini-2.0-flash-exp"),
        tools=[DuckDuckGo()],
        markdown=True,
    )

# Initialize the agent
multimodal_Agent = initialize_agent()

# YouTube video URL input
video_url = st.text_input(
    "Enter a YouTube video URL", 
    help="Paste a YouTube video link for AI analysis"
)

if video_url:
    try:
        # Download the YouTube video
        yt = YouTube(video_url)
        video_stream = yt.streams.filter(progressive=True, file_extension='mp4').first()
        
        # Create a temporary file to save the video
        with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
            video_stream.download(output_path=temp_video.name)
            video_path = temp_video.name
        
        st.video(video_path, format="video/mp4", start_time=0)

        user_query = st.text_area(
            "What insights are you seeking from the video?",
            placeholder="Ask anything about the video content. The AI agent will analyze and gather additional context if needed.",
            help="Provide specific questions or insights you want from the video."
        )

        if st.button("🔍 Analyze Video", key="analyze_video_button"):
            if not user_query:
                st.warning("Please enter a question or insight to analyze the video.")
            else:
                try:
                    with st.spinner("Processing video and gathering insights..."):
                        # Upload and process the video file
                        processed_video = upload_file(video_path)
                        while processed_video.state.name == "PROCESSING":
                            time.sleep(1)
                            processed_video = get_file(processed_video.name)

                        # Prompt generation for analysis
                        analysis_prompt = (
                            f"""
                            Analyze the uploaded video for content and context.
                            Respond to the following query using video insights and supplementary web research:
                            {user_query}
                            Provide a detailed, user-friendly, and actionable response.
                            """
                        )

                        # AI agent processing
                        response = multimodal_Agent.run(analysis_prompt, videos=[processed_video])

                    # Display the result
                    st.subheader("Analysis Result")
                    st.markdown(response.content)

                except Exception as error:
                    st.error(f"An error occurred during analysis: {error}")
                finally:
                    # Clean up temporary video file
                    Path(video_path).unlink(missing_ok=True)
    except Exception as e:
        st.error(f"Error downloading YouTube video: {e}")
else:
    st.info("Paste a YouTube video URL to begin analysis.")