app.py

# ======================== app.py ========================

import gradio as gr
import os
import json
import shutil # Used for file operations

# Import the core functions from your existing files
from modules.video_analyzer import analyze_video_for_ppe
from modules.rag_indexer import index_analysis_data
from modules.rag_query import run_query

# Configuration (Keep consistent with rag_indexer.py and rag_query.py)
VIDEO_FILENAME = "uploaded_video.mp4" # Temp name for the uploaded file
RAW_ANALYSIS_FILE = 'raw_analysis.json'
DB_PATH = "./chroma_db"
COLLECTION_NAME = 'video_analysis_data' # Use the general collection name

def pipeline_fn(video_file, user_query):
    """
    The main function connecting the Gradio inputs to the RAG pipeline.
    
    Args:
        video_file: The temporary file object from Gradio (gr.File).
        user_query: The text question from Gradio (gr.Textbox).
    
    Returns:
        The text response from the RAG query.
    """
    if video_file is None:
        return "Error: Please upload a video file first."
    if not user_query:
        return "Error: Please enter a question to query the video analysis."

    # 1. Handle File Upload and Naming
    # Gradio passes a temporary file path, we need to copy it 
    # and rename it so the analyzer can find it consistently.
    try:
        # We copy the file to the expected working directory
        temp_video_path = os.path.join(os.getcwd(), VIDEO_FILENAME)
        shutil.copy(video_file.name, temp_video_path)
        print(f"Copied uploaded file to: {temp_video_path}")
    except Exception as e:
        return f"File handling error: {e}"

    # 2. Analyze Video
    print("\n--- STAGE 1: Analyzing Video ---")
    # frames_per_sec=0.5 is a sensible default for a quick demo
    analysis_results = analyze_video_for_ppe(
        video_path=temp_video_path, 
        frames_per_sec= 2
    )
    
    # Save the raw analysis for the indexer to pick up
    with open(RAW_ANALYSIS_FILE, 'w') as f:
        json.dump(analysis_results, f, indent=4)
        
    # 3. Index Data
    print("\n--- STAGE 2: Indexing Analysis Data ---")
    # This must be run to create/update the ChromaDB with the new analysis
    index_analysis_data(json_file=RAW_ANALYSIS_FILE, collection_name=COLLECTION_NAME)

    # 4. Execute RAG Query
    print("\n--- STAGE 3: Executing RAG Query ---")
    rag_answer = run_query(user_query)
    
    # 5. Cleanup (Optional but Recommended for Demo)
    os.remove(temp_video_path) # Remove the copied video file
    os.remove(RAW_ANALYSIS_FILE) # Remove the temporary JSON file
    
    return rag_answer

# --- Gradio Interface Definition ---

# Define the input components
video_input = gr.File(
    label="Upload Video File (.mp4, .mov, etc.)",
    file_types=["video"], # Restrict to video files
    type="filepath"
)
query_input = gr.Textbox(
    label="Ask a Question about the Video Content", 
    placeholder="e.g., What are people doing in the video?",
    lines=2
)

# Define the output component
output_textbox = gr.Textbox(
    label="RAG Analysis Result",
    lines=10,
    interactive=False
)

# Create the Gradio Interface
demo = gr.Interface(
    fn=pipeline_fn,
    inputs=[video_input, query_input],
    outputs=output_textbox,
    title="🚀 Video Content RAG Pipeline",
    description="Upload a video, and ask a question. The pipeline runs object detection, indexes the data, and uses Gemini to answer your question based on the analysis.",
)

if __name__ == "__main__":
    print("Launching Gradio App...")
    # This will open the app in your browser at http://127.0.0.1:7860/
    demo.launch()