app.py

import gradio as gr
import os
import time
import openai
import pandas as pd

openai_api_key_textbox = ""
model = None
tokenizer = None
generator = None
csv_name = "disease_database_mini.csv"
df = pd.read_csv(csv_name)
openai.api_key = "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"

def csv_prompter(question,csv_name):
    
    

    fulltext = "A question is provided below. Given the question, extract " + \
               "keywords from the text. Focus on extracting the keywords that we can use " + \
               "to best lookup answers to the question. \n" + \
               "---------------------\n" + \
               "{}\n".format(question) + \
               "---------------------\n" + \
               "Provide keywords in the following comma-separated format.\nKeywords: "

    messages = [
        {"role": "system", "content": ""},
    ]
    messages.append(
        {"role": "user", "content": f"{fulltext}"}
    )
    rsp = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=messages
    )
    keyword_list = rsp.get("choices")[0]["message"]["content"]
    keyword_list = keyword_list.replace(",","").split(" ")

    print(keyword_list)
    divided_text = []
    csvdata = df.to_dict('records')
    step_length = 15
    for csv_item in range(0,len(csvdata),step_length):
        csv_text = str(csvdata[csv_item:csv_item+step_length]).replace("}, {", "\n\n").replace("\"", "")#.replace("[", "").replace("]", "")
        divided_text.append(csv_text)

    answer_llm = ""

    score_textlist = [0] * len(divided_text)

    for i, chunk in enumerate(divided_text):
        for t, keyw in enumerate(keyword_list):
            if keyw.lower() in chunk.lower():
                score_textlist[i] = score_textlist[i] + 1

    answer_list = []
    divided_text = [item for _, item in sorted(zip(score_textlist, divided_text), reverse=True)]

    for i, chunk in enumerate(divided_text):

        if i>4:
            continue

        fulltext = "{}".format(chunk) + \
                   "\n---------------------\n" + \
                   "Based on the Table above and not prior knowledge, " + \
                   "Select the Table Entries that will help to answer the question: {}\n Output in the format of \" Disease: <>; Symptom: <>; Medical Test: <>; Medications: <>;\". If there is no useful form entries, output: 'No Entry'".format(question)

        print(fulltext)
        messages = [
            {"role": "system", "content": ""},
        ]
        messages.append(
            {"role": "user", "content": f"{fulltext}"}
        )
        rsp = openai.ChatCompletion.create(
            model="gpt-3.5-turbo",
            messages=messages
        )
        answer_llm = rsp.get("choices")[0]["message"]["content"]

        print("\nAnswer: " + answer_llm)
        print()
        if not "No Entry" in answer_llm:
            answer_list.append(answer_llm)



    fulltext = "The original question is as follows: {}\n".format(question) + \
               "Based on this Table:\n" + \
               "------------\n" + \
               "{}\n".format(str("\n\n".join(answer_list))) + \
               "------------\n" + \
               "Answer: "
    print(fulltext)
    messages = [
        {"role": "system", "content": ""},
    ]
    messages.append(
        {"role": "user", "content": f"{fulltext}"}
    )
    rsp = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=messages
    )
    answer_llm = rsp.get("choices")[0]["message"]["content"]

    print("\nFinal Answer: " + answer_llm)
    print()

    return answer_llm



with gr.Blocks() as demo:
    gr.Markdown("# Autonomous ChatDoctor (openai version), based on disease database knowledge")
    gr.Markdown("## Example: If I have frontal headache, fever, and painful sinuses, what disease should I have, and what medical test should I take?")
    gr.Markdown("Our model will answer based on the content of the excel below, so please try to ask questions based on the table content.")
    
    chatbot = gr.Chatbot()
    msg = gr.Textbox()
    clear = gr.Button("Clear")
    Initialization = gr.Button("Initialization")

    def restart(history):
        invitation = "ChatDoctor: "
        human_invitation = "Patient: "
        return [[" \n",invitation+" I am ChatDoctor, what medical questions do you have?"]]
            
    def user(user_message, history):
        invitation = "ChatDoctor: "
        human_invitation = "Patient: "
        return "", history +[[human_invitation+user_message, None]]

    def bot(history):
        invitation = "ChatDoctor: "
        human_invitation = "Patient: "
        print(history)

        
        question = ""
        for each_ques in history:
            question = question+ each_ques[0].replace("Patient: ","")+" \n"

        response = csv_prompter(question,csv_name)
        
        response = invitation+ response
        history[-1][1] = response

        return history

    msg.submit(user, [msg, chatbot], [msg, chatbot], queue=False).then(
        bot, chatbot, chatbot
    )
    clear.click(lambda: None, None, chatbot, queue=False).then(restart, chatbot, chatbot)
    Initialization.click(lambda: None, None, chatbot, queue=False).then(restart, chatbot, chatbot)
    gr.Dataframe(df)

    
if __name__ == "__main__":
    demo.launch()