AI Meal Planner

How I Built a Telegram Bot That Combines GPT-4 and GPT-3

How It Works

1. Users send a photo of their ingredients or manually enter them as text.

2. The bot processes the image using GPT-4 to extract ingredients.

3. GPT-3 generates three high-protein, healthy recipes based on the identified ingredients.

4. Users can choose a recipe, get nutritional info, or subscribe for premium features.

Tech Stack

• Telegram Bot API: Handles user interactions and commands.

• GPT-4: Used for advanced image recognition to extract ingredient lists from photos.

• GPT-3: Generates recipes and handles conversational interactions.

• Stripe Integration: Provides subscription-based premium access.

• Flask: Used for webhooks or additional server-side logic.

Features of the Bot

1. Image-Based Ingredient Detection: Users send a photo of their fridge or pantry, and the bot identifies the ingredients using GPT-4.

2. Text-Based Entry: Users can manually list their ingredients if they prefer.

3. Recipe Suggestions: GPT-3 generates three concise recipe suggestions tailored to the ingredients.

4. Premium Access: Offers extended features such as unlimited usage and additional recipe details for subscribers.

5. Referrals and Coupons: Users can refer friends or use coupon codes for free trials.

6. Interactive Menu: Includes options for tutorials, feedback, and support.

Code Highlights

GPT-4 for Image Recognition

def generate_feedback_img(image_path):
    base64_image = encode_image(image_path)
    headers = {
        "Content-Type": "application/json",
        "Authorization": f"Bearer {api_key}"
    }
    payload = {
        "model": "gpt-4-vision-preview",
        "messages": [
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "List all food items in this image, separated by commas. Be concise."
                    },
                    {
                        "type": "image_url",
                        "image_url": {"url": f"data:image/jpeg;base64,{base64_image}"}
                    }
                ]
            }
        ],
        "max_tokens": 500
    }
    response = requests.post("https://api.openai.com/v1/chat/completions", headers=headers, json=payload)
    return response.json()['choices'][0]['message']['content']

GPT-3 for Recipe Generation

def generate_response_gpt3(new_text, user_id):
    history = user_histories.get(user_id, [])
    history.append({"role": "user", "content": new_text})
    openai.api_key = api_key
    chat_completion = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=history,
        max_tokens=500,
        temperature=0.7
    )
    response = chat_completion.choices[0].message['content']
    history.append({"role": "assistant", "content": response})
    user_histories[user_id] = history
    return response

Telegram Bot Integration

def MyTelegramBot():
    TOKEN = config.BOT_TOKEN
    updater = Updater(TOKEN, use_context=True)
    dp = updater.dispatcher
    dp.add_handler(CommandHandler("start", start))
    dp.add_handler(CommandHandler("tutorial", tutorial))
    dp.add_handler(MessageHandler(Filters.photo, photo_handler))
    dp.add_handler(CallbackQueryHandler(button_handler))
    updater.start_polling()
    updater.idle()

What I Learned

1. Combining APIs: GPT-4 for vision and GPT-3 for language generation is a powerful duo.

2. User Experience Matters: Adding interactive features like tutorials and feedback creates a user-friendly experience.

3. Scalability: Handling subscriptions and usage limits ensures the bot is sustainable for long-term use.