app.py

import gradio as gr
import os
from openai import OpenAI

# Set your OpenAI API key
#openai.api_key = config.OPENAI_API_KEY
client = OpenAI(api_key=os.environ['OPENAI_API_KEY'])

def converse(x, y, z):
    return z

def reset(z):
    return [], []

# Initial message
messages = [{
    "role": "system",
    "content": "You are a medical advisor specializing in humanitarian health. Familiar with all ICRC guidelines, \
        your task is to methodically gather information about a patient's condition. When presented with partial or \
            unclear information, ask clarifying questions one at a time (wait for an answer before asking another question) \
            until you can make a well-informed suggestion. \
                If, after several attempts, adequate details are still missing, advise the user on the next steps they \
                    should consider (e.g., lab tests, consultations). Once you believe you have all necessary details, \
                        provide treatment suggestions based on the information. Conclude each session by reminding users \
                            of the importance of consulting with medical professionals and offering a summary. Remind the users \
                            that you are an AI chatbot and cannot be held accountable for any mistakes that could lead to \
                            a patient hard. Patient safety is your priority so ensure that the users of the chatbot is aware. Avoid using \
                                quotation marks and always adhere to ICRC and medical guidelines. Do not reveal your nature \
                                    as an AI language model."
}]  




def provide_suggestions(user_message, history):
    global messages

    # Update the global messages list with the user's input
    messages.append({"role": "user", "content": user_message})
    
    # Get a response from the model
    response = client.chat.completions.create(model="gpt-3.5-turbo", messages=messages)
    
    # Extract the model's message from the response
    system_message = response.choices[0].message.content
    
    # Update the messages list with the model's response
    messages.append({"role": "assistant", "content": system_message})

    return system_message

# Define and launch the Gradio Chat Interface
iface = gr.ChatInterface(fn=provide_suggestions, title="MedGuide+",  \
                         description="Introducing our AI medical advisor, an innovative and knowledgeable \
                            resource designed to provide prompt and reliable support in the field of healthcare \
                                and medicine, with extensive training on a wide range of medical literature and \
                                    guidelines, offering valuable insights and recommendations while emphasizing the \
                                        importance of consulting with qualified healthcare professionals for personalized medical advice and care. \
                                        For more info, check out: https://github.com/jmesplana/MedGuide_Plus")
# Rest of your Gradio Interface setup

def get_medical_summary_from_chat(messages):
    if len(messages) <= 1:
        return "No consultation data available."

    extraction_prompt = {
        "role": "user",
        "content": "Based on the chat details, please provide a detailed summary following the international patient summary (IPS) format.\
                    break down the summary in categories following the IPS category. Put the content of each category in bullet points (if needed)."
    }
    messages.append(extraction_prompt)
    
    # Get a response from the model
    response = client.chat.completions.create(model="gpt-3.5-turbo", messages=messages)
    
    # Extract the model's message from the response
    system_message = response.choices[0].message.content
    
    return system_message

# Rest of your Gradio summary interface and main execution block

# Gradio function for the summary interface
def show_summary():
    consolidated_output = get_medical_summary_from_chat(messages)
    return consolidated_output

# Create the secondary Gradio interface
summary_layout = gr.Interface(fn=show_summary, 
                              inputs=[], 
                              outputs="text",  # Single text output
                              live=True,
                              title="Chat Summary",
                              description="Summary of the consultation based on the chat data."
                             )


                             
#summary_layout.launch()

demo = gr.TabbedInterface([iface, summary_layout],tab_names=['chatbot','summary'])

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