app.py

import gradio as gr
from phi.agent import Agent
from phi.model.google import Gemini
from phi.model.groq import Groq
from phi.tools.duckduckgo import DuckDuckGo
from phi.tools.youtube_tools import YouTubeTools
from google.generativeai import upload_file, get_file
import google.generativeai as genai
from dotenv import load_dotenv
import time
import tempfile
import os
from pathlib import Path

# Load environment variables
load_dotenv()

# Google API Key
API_KEY = os.getenv("GOOGLE_API_KEY")
if API_KEY:
    genai.configure(api_key=API_KEY)

# Groq API Key
groq_api_key = os.getenv("Groq_Api_Key")

# Initialize Multimodal Agent
def initialize_multimodal_agent():
    return Agent(
        name="Video AI Summarizer",
        model=Gemini(id="gemini-2.0-flash-exp"),
        tools=[DuckDuckGo()],
        markdown=True,
    )

# Video Analysis Function
def analyze_video(video_file, user_query):
    if not user_query:
        return "Please enter a question or insight to analyze the video."
    
    with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video:
        temp_video.write(video_file.read())
        video_path = temp_video.name
    
    # Upload and process video
    try:
        processed_video = upload_file(video_path)
        while processed_video.state.name == "PROCESSING":
            time.sleep(1)
            processed_video = get_file(processed_video.name)

        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.
        """

        multimodal_agent = initialize_multimodal_agent()
        response = multimodal_agent.run(analysis_prompt, videos=[processed_video])

        # Clean up temporary video file
        Path(video_path).unlink(missing_ok=True)

        return response.content

    except Exception as error:
        Path(video_path).unlink(missing_ok=True)
        return f"An error occurred during analysis: {error}"

# YouTube Summarization Function
def summarize_youtube(video_url):
    if not video_url.strip():
        return "Please enter a valid YouTube video URL."
    
    try:
        youtube_agent = Agent(
            tools=[YouTubeTools()],
            model=Groq(id="llama-3.3-70b-versatile", api_key=groq_api_key),
            show_tool_calls=True,
            description="You are a YouTube agent. Obtain the captions of a YouTube video and answer questions.",
        )

        response = youtube_agent.print_response(f"Summarize this video {video_url}", markdown=True)
        return response

    except Exception as e:
        return f"An error occurred: {e}"

# Gradio Interface
with gr.Blocks() as demo:
    gr.Markdown("# Phidata Video AI Summarizer Agent 🎥🎤🖬")

    with gr.Tab("🎥 Video Upload"):
        gr.Markdown("### 🎥 Analyze Uploaded Video")
        video_file = gr.File(label="Upload a video file", type="filepath", file_types=["mp4", "mov", "avi"])
        user_query = gr.Textbox(label="What insights are you seeking from the video?", placeholder="Ask anything about the video content.")
        analyze_button = gr.Button("🔍 Analyze Video")
        analysis_result = gr.Textbox(label="Analysis Result", interactive=False)

        analyze_button.click(fn=analyze_video, inputs=[video_file, user_query], outputs=analysis_result)

    with gr.Tab("🌐 YouTube Summarizer"):
        gr.Markdown("### 🌐 Summarize YouTube Videos")
        video_url = gr.Textbox(label="Enter YouTube video URL:")
        summarize_button = gr.Button("Summarize Video")
        youtube_summary = gr.Textbox(label="Summary", interactive=False)

        summarize_button.click(fn=summarize_youtube, inputs=video_url, outputs=youtube_summary)

# Launch the interface
demo.launch()