app.py

from flask import Flask, render_template, request, jsonify, Response
import google.generativeai as genai
import os
from PIL import Image
import tempfile
import io
import uuid
import time

app = Flask(__name__)

# Configuration Gemini
token = os.environ.get("TOKEN")
genai.configure(api_key=token)

safety_settings = [
    {"category": "HARM_CATEGORY_HARASSMENT", "threshold": "BLOCK_NONE"},
    {"category": "HARM_CATEGORY_HATE_SPEECH", "threshold": "BLOCK_NONE"},
    {"category": "HARM_CATEGORY_SEXUALLY_EXPLICIT", "threshold": "BLOCK_NONE"},
    {"category": "HARM_CATEGORY_DANGEROUS_CONTENT", "threshold": "BLOCK_NONE"},
]

mm = """ resous cet exercice. tu répondras en détaillant au maximum ton procédé de calcul. réponse attendue uniquement en Latex"""

model = genai.GenerativeModel(
    model_name="gemini-exp-1206",
    safety_settings=safety_settings
)

# Dictionnaire pour stocker les réponses en cours de génération
pending_responses = {}

@app.route('/')
def home():
    return render_template('index.html')

@app.route('/generate', methods=['POST'])
def generate():
    if 'image' not in request.files:
        return jsonify({'error': 'No image uploaded'}), 400

    image_file = request.files['image']
    request_id = str(uuid.uuid4())  # Générer un identifiant unique

    # Sauvegarder temporairement l'image
    with tempfile.NamedTemporaryFile(delete=False, suffix='.png') as temp_file:
        image_file.save(temp_file.name)
        try:
            image = Image.open(temp_file.name)
            # Convertir l'image en bytes pour le streaming
            img_byte_arr = io.BytesIO()
            image.save(img_byte_arr, format='PNG')
            img_byte_arr = img_byte_arr.getvalue()

            # Stocker la tâche de génération dans le dictionnaire
            pending_responses[request_id] = {
                'status': 'processing',
                'response': model.generate_content([mm, {"mime_type": "image/png", "data": img_byte_arr}], stream=True)
            }

            return jsonify({'request_id': request_id})
        except Exception as e:
            return jsonify({'error': str(e)}), 500
        finally:
            # Nettoyer le fichier temporaire
            os.unlink(temp_file.name)
    
@app.route('/stream/<request_id>')
def stream(request_id):
    def generate_stream():
        while request_id in pending_responses and pending_responses[request_id]['status'] == 'processing':
            try:
                chunk = next(pending_responses[request_id]['response'])
                yield f"data: {chunk.text}\n\n"
            except StopIteration:
                pending_responses[request_id]['status'] = 'completed'
            except Exception as e:
                yield f"data: Error: {str(e)}\n\n"
                pending_responses[request_id]['status'] = 'error'
            time.sleep(0.1)  # Attendre un peu avant de vérifier à nouveau

        if request_id in pending_responses:
            del pending_responses[request_id]

    return Response(generate_stream(), mimetype='text/event-stream')

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