dataset_describer.py

import os
import openai
import streamlit as st
import pandas as pd
import re
import uuid




def read_csv_to_string(file_path):
    with open(file_path, 'r', newline='', encoding='utf-8') as file:
        content = file.read()
    return content

def extract_string_between_triple_backticks(text):
    pattern = r'```(.*?)```'
    matches = re.findall(pattern, text, re.DOTALL)
    return matches

def save_to_py_file(strings, file_path):
    with open(file_path, 'w', encoding='utf-8') as file:
        for string in strings:
            file.write(string)
            file.write('\n')

#@st.cache_data
def load_data(uploaded_file,file_name):
    dataframe_version = pd.read_csv(uploaded_file, encoding='utf-8',delimiter=csv_delimiter)
    dataframe_version.to_csv('datasets/'+file_name+'.csv')
    #string_version = uploaded_file.read().decode("utf-8")
    string_version = read_csv_to_string('datasets/'+file_name+'.csv')
    return dataframe_version,string_version


max_token = 3000
summ_temprature = 0.5

# upload a csv file to the datasets folder using st.fileuploader
with st.sidebar:
    uploaded_file = st.file_uploader("Choose a CSV file from your computer", accept_multiple_files=False)
    csv_delimiter = st.radio("choose delimiter", (';', ','))

if uploaded_file is not None:
    unique_filename = str(uuid.uuid4())

    df,st_version = load_data(uploaded_file,unique_filename)
    st.dataframe(df.head(5))

    # Set up OpenAI API key
    openai.api_key = st.secrets['open_ai_key']['OPENAI_API_KEY']

    with st.form(key='my_form'):
        plot_description = st.text_area("Describe the plot you want to see.")
        submit_button = st.form_submit_button(label='Submit')

    #plot_description = st.text_area("Describe the plot you want to see.")

    if plot_description:

        response_dataset_description = openai.ChatCompletion.create(
                model="gpt-3.5-turbo",
                messages=[
                        {"role": "user", "content": st.secrets['chatgpt_queries']['datadescribe_query1'] +
                                                    st_version[0:200] + "\n" +
                                                    st.secrets['chatgpt_queries']['datadescribe_query2']
                                                    }
            ],
            max_tokens = max_token,
            temperature = summ_temprature,
            )
        
        
        
        response_dataset_summary = openai.ChatCompletion.create(
                model="gpt-3.5-turbo",
                messages=[
                        {"role": "user", "content": st.secrets['chatgpt_queries']['datadescribe_query1'] +
                                                    st_version[0:200] + "\n" +
                                                    plot_description + "\n" +
                                                    st.secrets['chatgpt_queries']['datadescribe_query3'] +
                                                    st.secrets['chatgpt_queries']['datadescribe_query4'] +
                                                    st.secrets['chatgpt_queries']['datadescribe_query5']+
                                                    st.secrets['chatgpt_queries']['datadescribe_query6']+
                                                    f'datasets/{unique_filename}.csv'+ 
                                                    st.secrets['chatgpt_queries']['datadescribe_query7']
                                                    }
            ],
            max_tokens = max_token,
            temperature = summ_temprature,
            )

        with st.sidebar:
            st.write(response_dataset_description.choices[0].message.content)

        created_code = extract_string_between_triple_backticks(response_dataset_summary.choices[0].message.content)

        save_to_py_file(created_code, 'plotter.py')


        with open("plotter.py") as f:
            exec(f.read())
       
        os.remove('datasets/'+unique_filename+'.csv')