app.py

import streamlit as st
import pytesseract
from io import StringIO
from PIL import Image
import json
import openai
import pandas as pd

openai.api_key = st.secrets["open_ai_key"]

uploaded_file = st.file_uploader("Choose a file")
if uploaded_file is not None:
    
    image = Image.open(uploaded_file)
    
    with st.expander('Uploaded image', expanded=False):
        st.image(image)
    
    with st.expander('Raw OCR', expanded=False):
        extractedInformation = pytesseract.image_to_string(image)
        st.write(extractedInformation)
    
    with st.expander('GPT cleaned output', expanded=True):
        context = '''parse the unstructured input of a cooking recipe into a valid JSON array of objects in the following format:
        [{
        "recipe_name": "the name of the recipe expressed as a string",
        "recipe_ingredients": "list of ingredients specified in the recipe expressed as string",
        "recipe_steps": "list of steps to cook the recipe expressed as a string"
        }]'''
    
        gpt_response = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=[
                {"role": "system", "content": context},
                {"role": "user", "content": extractedInformation}
            ],
            temperature=0.2,
            max_tokens=1000
        )
    
        st.json((json.loads(gpt_response['choices'][0]['message']['content'])))
        cost = gpt_response['usage']["prompt_tokens"]*(0.0015/1000) + gpt_response['usage']["completion_tokens"]*(0.002/1000)
        st.write(f'Cost for query was approx ${cost}')