app.py

import gradio as gr 
import os 
import pandas as pd 
import matplotlib.pyplot as plt 
import numpy as np 
from langchain_community.chat_models import ChatAnyscale
from Hypothesis_Test_Agent.Hypothesis_Agent import infer_hypothesis
from EDA_Agent.EDA import infer_EDA
from Basic_Inf_Agent.Basic_Inference_Agent import infer_basic

print(os.getcwd())

def run_agent(data_path:str = '', provider:str = 'Mistral', agent_type:str = 'EDA', query:str = '', layers:str ='', temp:float = 0.1):
    
    df = pd.read_csv(data_path)
    df.to_csv('./df.csv', index=False)
    
    if provider.lower() == 'mistral':
        
        llm = ChatAnyscale(model_name='mistralai/Mixtral-8x7B-Instruct-v0.1', temperature=temp)
        
        if agent_type == 'Data Explorer':
            EDA_response, EDA_image = infer_EDA(user_input=query, llm=llm, df=df)
            return EDA_response, EDA_image
        
        if agent_type == 'Hypothesis Tester':
            hypothesis_response = infer_hypothesis(user_input=query, llm=llm, df=df)
            return hypothesis_response, None
        
        if agent_type == 'Basic Inference':
            basic_response = infer_basic(user_input=query, df=df, llm=llm)
            return basic_response, None
        

    return None


demo = gr.Interface (
    run_agent,
    [
        gr.UploadButton(label="Upload your CSV!"),
        gr.Radio(["Mistral","GPT"],label="Select Your LLM"),
        gr.Radio(["Data Explorer", "Hypothesis Tester", "Basic Inference", "Super Inference"], label="Choose Your Agent"),
        gr.Text(label="Query", info="Your input to the Agent. Be descriptive!"),
        gr.Text(label="Architecture", info="Specify the layer by layer architecture only for Super Inference Agent"),
        gr.Slider(label="Model Temperature", info="Slide right to make your model more creative!")
    ],
    [
        gr.Text(label="Agent Output", info="This might take a while to generate since agent debugs its errors too ..."),
        gr.Image(label="Graph")
    ]
)

demo.launch(share=True)