app.py

import os
import whisper
import requests
from flask import Flask, request, jsonify, render_template
from dotenv import load_dotenv
from deepgram import DeepgramClient, PrerecordedOptions
import tempfile
import json
import subprocess
from youtube_transcript_api import YouTubeTranscriptApi


import warnings
warnings.filterwarnings("ignore", message="FP16 is not supported on CPU; using FP32 instead")

app = Flask(__name__)
print("APP IS RUNNING, ANIKET")

# Load the .env file
load_dotenv()

print("ENV LOADED, ANIKET")

# Fetch the API key from the .env file
API_KEY = os.getenv("FIRST_API_KEY")
DEEPGRAM_API_KEY = os.getenv("SECOND_API_KEY")

# Ensure the API key is loaded correctly
if not API_KEY:
    raise ValueError("API Key not found. Make sure it is set in the .env file.")

if not DEEPGRAM_API_KEY:
    raise ValueError("DEEPGRAM_API_KEY not found. Make sure it is set in the .env file.")

GEMINI_API_ENDPOINT = "https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-latest:generateContent"
GEMINI_API_KEY = API_KEY

@app.route("/", methods=["GET"])
def health_check():
    return jsonify({"status": "success", "message": "API is running successfully!"}), 200



def download_audio(url, temp_video_path):
    """Download video (MP4 format) from the given URL and save it to temp_video_path."""
    response = requests.get(url, stream=True)
    if response.status_code == 200:
        with open(temp_video_path, 'wb') as f:
            for chunk in response.iter_content(chunk_size=1024):
                f.write(chunk)
        print(f"Audio downloaded successfully to {temp_video_path}")
    else:
        raise Exception(f"Failed to download audio, status code: {response.status_code}")



@app.route('/process-video', methods=['POST'])
def process_video():
    if 'videoUrl' not in request.json:
        return jsonify({"error": "No video URL provided"}), 400

    video_url = request.json['videoUrl']
    temp_video_path = None

    try:
        # Step 1: Download the WAV file from the provided URL
        with tempfile.NamedTemporaryFile(delete=False, suffix=".mp4") as temp_video_file:
            temp_video_path = temp_video_file.name
            download_audio(video_url, temp_video_path)

        # Step 2: get the information from the downloaded MP4 file synchronously
        video_info = get_information_from_video_using_OCR(temp_video_path)

        if not video_info:
            return jsonify({"error": "video information extraction failed"}), 500

        # Step 3: Generate structured recipe information using Gemini API synchronously
        structured_data = query_gemini_api(video_info)

        return jsonify(structured_data)

    except Exception as e:
        return jsonify({"error": str(e)}), 500

    finally:
        # Clean up temporary audio file
        if temp_video_path and os.path.exists(temp_video_path):
            os.remove(temp_video_path)
            print(f"Temporary audio file deleted: {temp_video_path}")






def query_gemini_api(transcription):
    """
    Send transcription text to Gemini API and fetch structured recipe information synchronously.
    """
    try:
        # Define the structured prompt
        prompt = (
            "Analyze the provided cooking video transcription and extract the following structured information:\n"
            "1. Recipe Name: Identify the name of the dish being prepared.\n"
            "2. Ingredients List: Extract a detailed list of ingredients with their respective quantities (if mentioned).\n"
            "3. Steps for Preparation: Provide a step-by-step breakdown of the recipe's preparation process, organized and numbered sequentially.\n"
            "4. Cooking Techniques Used: Highlight the cooking techniques demonstrated in the video, such as searing, blitzing, wrapping, etc.\n"
            "5. Equipment Needed: List all tools, appliances, or utensils mentioned, e.g., blender, hot pan, cling film, etc.\n"
            "6. Nutritional Information (if inferred): Provide an approximate calorie count or nutritional breakdown based on the ingredients used.\n"
            "7. Serving size: In count of people or portion size.\n"
            "8. Special Notes or Variations: Include any specific tips, variations, or alternatives mentioned.\n"
            "9. Festive or Thematic Relevance: Note if the recipe has any special relevance to holidays, events, or seasons.\n"
            "Also, make sure not to provide anything else or any other information or warning or text apart from the above things mentioned."
            f"Text: {transcription}\n"
        )

        # Prepare the payload and headers
        payload = {
            "contents": [
                {
                    "parts": [
                        {"text": prompt}
                    ]
                }
            ]
        }
        headers = {"Content-Type": "application/json"}

        # Send request to Gemini API synchronously
        response = requests.post(
            f"{GEMINI_API_ENDPOINT}?key={GEMINI_API_KEY}",
            json=payload,
            headers=headers,
        )

        # Raise error if response code is not 200
        response.raise_for_status()

        data = response.json()

        return data.get("candidates", [{}])[0].get("content", {}).get("parts", [{}])[0].get("text", "No result found")

    except requests.exceptions.RequestException as e:
        print(f"Error querying Gemini API: {e}")
        return {"error": str(e)}


if __name__ == '__main__':
    app.run(debug=True)