app.py

from pathlib import Path
import google.generativeai as genai
import re
from PIL import Image
import os
#from google.colab import userdata
#os.environ['GOOGLE_API_KEY'] = userdata.get('GOOGLE_API_KEY')
#GOOGLE_API_KEY = os.environ['GOOGLE_API_KEY']
#genai.configure(api_key=os.environ["GOOGLE_API_KEY"])
#or use this for personal notebook genai.configure(api_key="AIzaSyD----")

# Configuration for our Gemini Models

textgeneration_config = {
  "temperature": 0.9,
  "top_p": 1,
  "top_k": 1,
  "max_output_tokens": 2048,}


visiongeneration_config = {
  "temperature": 0.9,
  "top_p": 1,
  "top_k": 10,
  "max_output_tokens": 1024,
}

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

# Two models - vision and text
textmodel = genai.GenerativeModel('gemini-1.0-pro',
                              generation_config=textgeneration_config,
                              safety_settings=safety_settings)

#imagemodel = genai.GenerativeModel('gemini-pro-vision')

visionmodel = genai.GenerativeModel(model_name="gemini-1.0-pro-vision-latest",
                              generation_config=visiongeneration_config,
                              safety_settings=safety_settings)



# Utility Functions
# Convert an image to base64 string format
import base64
def img2base64(image):
    with open(image, 'rb') as img:
        encoded_string = base64.b64encode(img.read())
    return encoded_string.decode('utf-8')


# Check image format and display user messages in GUI
def user_inputs(history, txt, img):
    if not img:
        history += [(txt, None)]
        return history

    # Open the image for format verification
    try:
        with Image.open(img) as image:
            # Get image format (e.g., PNG, JPEG)
            image_format = image.format.upper()
    except (IOError, OSError):
        return history

    if image_format not in ('JPEG','JPG','PNG'):
        print(f"Warning: Unsupported image format: {image_format}")
        return history

    base64 = img2base64(img)
    data_url = f"data:image/{image_format.lower()};base64,{base64}"
    history += [(f"{txt} ![]({data_url})", None)]



import gradio as gr

TITLE = """<h1 align="center">Your Personal Health Coach</h1>"""
SUBTITLE = """<h2 align="center">Upload an image of your food to knows its calories, macronutrients or ask questions about heath and exercise.</h2>"""
DES = """
<div style="text-align: center; display: flex; justify-content: center; align-items: center;">
    <span>You need to enter your FREE GEMINI KEY in the first text box to connect to Gemini Models. You can find your key here:
        <a href="https://makersuite.google.com/app/apikey">GOOGLE API KEY</a>. <br><br>
        <b> If you wish to ask a question unrelated to the image you have uploaded, just cross the image (top right corner of image) and then submit your question in the textbox.
    </span>
</div>
"""

def generate_model_response(api_key, history, text, img):
  genai.configure(api_key=api_key)
  if not img:
        text = "You are an expert nutritionist and fitness coach. You are accurate, you always stick to the facts, and never make up new facts. \
            For the questions asked by the user, answer accurately and to the point, in a friendly tone." + text
        response = textmodel.generate_content(text)

  else:
        text = "From the image uploaded by the user answer with following information: Food items in the image, \
            percentage of each macronutrient in the food in image and approximate number of calories in the food in image. If there is any additional question, answer that too." + text
        img = Image.open(img)
        response = visionmodel.generate_content([text,img])
  history += [(None, response.text)]
  return history


with gr.Blocks() as app:

  gr.HTML(TITLE)
  gr.HTML(SUBTITLE)
  gr.HTML(DES)

  api_key_box = gr.Textbox(placeholder = "Enter your GEMINI API KEY", label="Your GEMINI API KEY", type="password")

  with gr.Row():
    image_box = gr.Image(type="filepath")
    chatbot = gr.Chatbot(
      scale=3,
      height=750
    )

  text_box = gr.Textbox(
    placeholder="Ask something about the image your uploaded or ask for any health and fitness advice without uploading an image too",
    container=False,
  )

  btn = gr.Button("Submit")

  btn_clicked = btn.click(user_inputs,
                     [chatbot, text_box, image_box],
                     chatbot).then(generate_model_response,[api_key_box, chatbot, text_box, image_box], chatbot)

app.queue()
app.launch(debug=True)