app.py

import streamlit as st
import pandas as pd
import json
import http.client
from io import StringIO
from openai import AzureOpenAI

class ContentFormatter:
    @staticmethod
    def chat_completions(text, settings_params):
        message = [
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": text}
        ]
        data = {"messages": message, **settings_params}
        return json.dumps(data)

class AzureAgent:
    def __init__(self, api_key, azure_uri, deployment_name, api_version):
        self.azure_uri = azure_uri
        self.headers = {
            'Authorization': f"Bearer {api_key}",
            'Content-Type': 'application/json'
        }
        self.deployment_name = deployment_name
        self.api_version = api_version
        self.chat_formatter = ContentFormatter

    def invoke(self, text, **kwargs):
        body = self.chat_formatter.chat_completions(text, {**kwargs})
        conn = http.client.HTTPSConnection(self.azure_uri)
        conn.request("POST", f'/v{self.api_version}/chat/completions', body=body, headers=self.headers)
        response = conn.getresponse()
        data = response.read()
        conn.close()
        decoded_data = data.decode("utf-8")
        parsed_data = json.loads(decoded_data)
        content = parsed_data["choices"][0]["message"]["content"]
        return content

class GPTAgent:
    def __init__(self, api_key, azure_endpoint, deployment_name, api_version):
        self.client = AzureOpenAI(
            api_key=api_key,
            api_version=api_version,
            azure_endpoint=azure_endpoint
        )
        self.deployment_name = deployment_name

    def invoke(self, text, **kwargs):
        response = self.client.chat.completions.create(
            model=self.deployment_name,
            messages=[
                {"role": "system", "content": "You are a helpful assistant."},
                {"role": "user", "content": text}
            ],
            **kwargs
        )
        return response.choices[0].message.content

# Streamlit app interface
st.title('JobFair: A Benchmark for Fairness in LLM Employment Decision')

model_type = st.radio("Select the type of agent", ('AzureAgent', 'GPTAgent'))
api_key = st.text_input("API Key", type="password")
endpoint_url = st.text_input("Endpoint URL")
deployment_name = st.text_input("Model Name")
api_version = st.text_input("API Version")  # Default API version

temperature = st.slider("Temperature", min_value=0.0, max_value=1.0, value=0.5, step=0.01)
max_tokens = st.number_input("Max Tokens", min_value=1, max_value=1000, value=150)
stop_sequences = st.text_input("Stop Sequences", "")
parameters = {"temperature": temperature, "max_tokens": max_tokens, "stop": [stop_sequences] if stop_sequences else []}


# File upload and data display
uploaded_file = st.file_uploader("Choose a file")
if uploaded_file is not None:
    # Read data
    data = StringIO(uploaded_file.getvalue().decode("utf-8"))
    df = pd.read_csv(data)
    st.write('Uploaded Data:', df.head())

    # Process data button
    if st.button('Process Data'):
        if model_type == 'AzureAgent':
            agent = AzureAgent(api_key, endpoint_url, deployment_name, api_version)
        else:
            agent = GPTAgent(api_key, endpoint_url, deployment_name, api_version)

        # Example processing step (adapt as needed)
        df['Response'] = df['prompt'].apply(lambda x: agent.invoke(x, parameters))

        st.write('Processed Data:', df.head())

        # Generate download link
        st.download_button(
            label="Download processed data",
            data=df.to_csv().encode('utf-8'),
            file_name='processed_data.csv',
            mime='text/csv',
        )