app.py

import streamlit as st
import pandas as pd
import plotly.graph_objects as go
import numpy as np
import openai
from openai import OpenAI
from dotenv import load_dotenv
import os
from scipy.stats import pointbiserialr

load_dotenv()
OPEN_API_KEY = os.getenv("OPEN_API_KEY")
openai.api_key = OPEN_API_KEY

# Function to process the uploaded file and interact with OpenAI's API
def analyze_uploaded_file(text):
    # Assuming the file is a text file. Adjust accordingly for other types.
    try:

        scrap_column_name = 'IsScrap'
        correlations = []
        all_columns = df.columns.to_list()
        for column in all_columns:
            if column == scrap_column_name:
                all_columns.remove(column)
                break
        for i in range(len(all_columns)):
            if all_columns[i] == scrap_column_name:
                continue
            scatter_x = df[all_columns[i]].values.astype(float)
            scatter_y = df[scrap_column_name].values.astype(float)
            corr, p = pointbiserialr(scatter_x, scatter_y)
            correlations.append(round(corr * 100, 2))

        combined = list(zip(all_columns, correlations))
        sorted_combined = sorted(combined, key=lambda x: x[1])
        all_columns, correlations = zip(*sorted_combined)
        input_columns = ""
        for i in range(len(all_columns)):
            input_columns += str(all_columns[i]) + ":" + str(correlations[i]) + ","

        # Initialize OpenAI API (ensure you've set your API key in your environment)
        openai.api_key = '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';

        # Crafting a prompt for the OpenAI model to analyze the causes of scrap from the content
        prompt_text = f"Take the following list of names of columns and their corresponding coefficients and find the top 3." + input_columns + " Give actionable steps on how to minimize scrap because of these columns in a manufacturing facility in this format: Column 1: 'name of column'\n 1. 'actionable step 1'\n 2. 'actionable step 2'\n 3. 'actionable step 3'\n complete these for the next 3 columns"
        system_prompt = f"You give suggestions for how to reduce scrap in a manufacturing facility given the factors."

        response = openai.chat.completions.create(
            model="gpt-3.5-turbo",
            messages=[
                {"role": "system", "content": system_prompt},
                {"role": "user", "content": prompt_text}
            ]
        )
        return response.choices[0].message.content
    except Exception as e:
        return str(e)  # For debugging purposes


def csv_to_text(df):
    # Initialize a list to hold the text representation of each column
    text_columns = []

    # Iterate over the DataFrame columns
    for column in df.columns:
        # Convert each column to a string, joining cells with " | " as the delimiter
        column_text = " | ".join(map(str, df[column].values))
        text_columns.append(column + ":\n" + column_text + "\n")

    # Join all column texts into a single text string, using double newlines as separators
    full_text = "\n".join(text_columns)

    return full_text


# Set page configuration to wide layout
st.set_page_config(layout='wide')

# Set page title
st.title('Sweep AI')
st.subheader('Upload a CSV file to analyze its contents')

# File upload widget
file = st.file_uploader('Upload CSV', type=['csv'])

if file is not None:
    df = pd.read_csv(file)

    parsed_text = csv_to_text(df)
    final_text = analyze_uploaded_file(parsed_text)

    # Display the contents of the CSV file
    st.write('**CSV file contents:**')
    st.write(df)

    st.markdown('---')  # Add separator

    # Display analytics section
    st.subheader('Analytics')

    # Create columns for plots in a 2x2 grid layout
    col1, col2 = st.columns(2)

    # Line plot
    with col1:
        st.write('**Interactive Line Plot**')
        x = np.linspace(0, 10, 100)
        y = np.sin(x)
        fig_line = go.Figure()
        fig_line.add_trace(go.Scatter(x=x, y=y, mode='lines'))
        fig_line.update_layout(
            xaxis_title='X-axis',
            yaxis_title='Y-axis',
            template='plotly_white',
            hovermode='x',
            margin=dict(l=40, r=40, t=40, b=40),
            width=400,
            height=300,
        )
        st.plotly_chart(fig_line, use_container_width=True, align="center")

    # Bar graph
    with col2:
        st.write('**Bar Graph**')
        categories = ['A', 'B', 'C', 'D']
        values = [20, 30, 15, 35]
        fig_bar = go.Figure()
        fig_bar.add_trace(go.Bar(x=categories, y=values))
        fig_bar.update_layout(
            xaxis_title='Categories',
            yaxis_title='Values',
            template='plotly_white',
            hovermode='x',
            margin=dict(l=40, r=40, t=40, b=40),
            width=400,
            height=300,
        )

        st.plotly_chart(fig_bar, use_container_width=True, align="center")

    # Pie chart
    col3, col4 = st.columns(2)
    with col3:
        st.write('**Pie Chart**')
        fig_pie = go.Figure()
        fig_pie.add_trace(go.Pie(labels=categories, values=values))
        fig_pie.update_layout(
            template='plotly_white',
            margin=dict(l=40, r=40, t=40, b=40),
            width=400,
            height=300,
        )
        st.plotly_chart(fig_pie, use_container_width=True, align="center")

    # Scatter plot
    with col4:
        st.write('**Scatter Plot**')
        scatter_x = np.random.rand(50)
        scatter_y = np.random.rand(50)
        fig_scatter = go.Figure()
        fig_scatter.add_trace(go.Scatter(x=scatter_x, y=scatter_y, mode='markers'))
        fig_scatter.update_layout(
            xaxis_title='X-axis',
            yaxis_title='Y-axis',
            template='plotly_white',
            hovermode='closest',
            margin=dict(l=40, r=40, t=40, b=40),
            width=400,
            height=300,
        )
        st.plotly_chart(fig_scatter, use_container_width=True, align="center")

    st.markdown('---')
    st.header('Suggestions to Reduce Scrap')
    st.write(final_text)
    st.markdown('---')

    st.markdown('---')
    st.header('Premium Tier')
    st.subheader('Hardware Configuration Recommendations')

    system_prompt = "Please take the following columns and their values. Ignore the scrap column. " + parsed_text + "Find the ideal value which is determined by rows that have a 0 in the scrap column and output the results in the following format example: OvenCNT: 79.230944, add a newline for each column and repeat for all columns"
    prompt_text = "You are a hardware configuration recommender."

    response = openai.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": system_prompt},
            {"role": "user", "content": prompt_text}
        ]
    )
    output = response.choices[0].message.content
    st.write(output)
    st.markdown('---')


    # Centered text using HTML and CSS
    st.write('<div style="text-align: center;">Thank you for using <span style="font-size: 22px; font-weight: 800;">SweepAI</span>!</div>', unsafe_allow_html=True)