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 phi.model.groq import Groq from phi.tools.youtube_tools import YouTubeTools # Load environment variables load_dotenv() # Configure API keys API_KEY = os.getenv("GOOGLE_API_KEY") groq_api_key = os.getenv("GROQ_API_KEY") if API_KEY: genai.configure(api_key=API_KEY) # Page configuration st.set_page_config( page_title="Multimodal AI Applications", page_icon="๐ŸŒ", layout="wide" ) # Tabs for the two applications tab1, tab2 = st.tabs(["Video Summarizer", "YouTube Video Analyzer"]) # Tab 1: Video Summarizer with Gemini with tab1: st.title("Phidata Video AI Summarizer Agent ๐ŸŽฅ๐ŸŽค๐Ÿ–ฌ") 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, ) multimodal_Agent = initialize_agent() video_file = st.file_uploader( "Upload a video file", type=['mp4'], help="Upload a video for AI analysis" ) if video_file: with tempfile.NamedTemporaryFile(delete=False, suffix='.mp4') as temp_video: temp_video.write(video_file.read()) 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. This AI agent can search the internet.", help="Provide specific questions or insights you want from the video." ) if st.button("๐Ÿ” 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..."): 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. Provide a detailed summary. Search the internet for similar topics discussed in the uploaded video. Respond to the following query using video insights and supplementary web research: {user_query} Provide a detailed, user-friendly, and actionable response. """ ) response = multimodal_Agent.run(analysis_prompt, videos=[processed_video]) st.subheader("Analysis Result") st.markdown(response.content) except Exception as error: st.error(f"An error occurred during analysis: {error}") finally: Path(video_path).unlink(missing_ok=True) else: st.info("Upload a video file to begin analysis.") # Tab 2: YouTube Video Analyzer with Groq with tab2: st.title("YouTube Video Analyzer") st.header("Analyze Youtube video using video link ๐ŸŽฌ ๐Ÿงท") try: youtube_agent = Agent( model=Groq(id="llama3-8b-8192", api_key=groq_api_key), tools=[YouTubeTools(), DuckDuckGo()], show_tool_calls=True, get_video_captions=True, get_video_data = True, 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.", ) except Exception as e: st.error(f"Error initializing the agent: {e}") st.stop() video_url = st.text_input("Enter YouTube Video URL:", "") user_query = st.text_area("Enter your question about the video (optional):", "") if st.button("โœจ Analyze the Video", key="Analyze_Video_Button"): if video_url: with st.spinner("Analyzing..."): try: prompt_prefix = """Analyze the YouTube video provided by the URL. Provide a detailed summary of the video, including key points. Structure the output as follows: **Detailed Summary:** [Detailed summary of the video content] **Key Points:** * [Key point 1] * [Key point 2] * [Key point 3] ... (and so on) Conduct web research on similar topics discussed in the video to provide additional context. """ if user_query: prompt = f"""{prompt_prefix} Respond to the following query using video insights and supplementary web research: {user_query} Provide a detailed, user-friendly, and actionable response. Video URL: {video_url}""" else: prompt = f"""{prompt_prefix} Video URL: {video_url}""" output = youtube_agent.run(prompt) st.markdown(output.content) except Exception as e: st.error(f"Error analyzing the video: {e}") else: st.warning("Please enter a YouTube video URL.")