new-backend.py

from flask import Flask, request, jsonify
from flask_cors import CORS
import os
from openai import OpenAI

client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

app = Flask(__name__)
CORS(app)  # Enable CORS

@app.route('/process_transcript', methods=['POST'])
def process_transcript():
    data = request.json
    print(request.json)
    transcript = data.get('transcript', '')  # Extract the transcript from the incoming request

    # Define the system and user messages 
    messages = [
        {
            "role": "system",
            "content": """You are a helpful medical data assistant. Your job is to understand the doctor-patient interaction an array of two data objects to extract important health data relating to a patients condition
            Example1 input: "Patient looks pale and is experiencing shortness of breath"
            Example1 output:
            [
                {
                    reasoning: "Mentioned muscle pain which can be caused by vitamin d deficiency",
                    category: "Time In Daylight",
                    data1_time: "6 month",
                    data1_type: "mean",
                    data2_time: "3 month",
                    data2_type: "trend"
                },
                {
                    reasoning: "Shortness of breath might indicate oxygen issues",
                    category: "Oxygen Saturation",
                    data1_time: "1 week",
                    data1_type: "min",
                    data2_time: "6 month",
                    data2_type: "mean"
                }
            ]

            Example2 Input: "Patient is experiencing rapid weight gain and feels constantly tired"
            Example2 Output:
            [
                {
                    "reasoning": "Constant fatigue could be related to poor sleep quality",
                    "category": "Sleep Time",
                    "data1_time": "1 month",
                    "data1_type": "min",
                    "data2_time": "1 week",
                    "data2_type": "mean"
                },
                {
                    "reasoning": "Weight gain might also affect heart health",
                    "category": "Resting Heart Rate",
                    "data1_time": "3 month",
                    "data1_type": "max",
                    "data2_time": "1 month",
                    "data2_type": "mean"
                }
            ]

            Example3 Input: "Patient has been feeling very anxious and has trouble sleeping"
            Example3 Output:
            [
                {
                    "reasoning": "Anxiety can elevate heart rate",
                    "category": "Heart Rate",
                    "data1_time": "1 week",
                    "data1_type": "mean",
                    "data2_time": "1 month",
                    "data2_type": "trend"
                },
                {
                    "reasoning": "Trouble sleeping might indicate irregular sleep patterns",
                    "category": "Sleep Time",
                    "data1_time": "1 week",
                    "data1_type": "min",
                    "data2_time": "6 month",
                    "data2_type": "mean"
                }
            ]

            Format options of each object:
            {
                "reasoning": "short sentence explaining reasoning for ",
                "category":  
                    "Body Mass Index",
                    "Height",
                    "Body Mass",
                    "Heart Rate",
                    "Oxygen Saturation",
                    "Blood Pressure Systolic",
                    "Blood Pressure Diastolic",
                    "Respiratory Rate",
                    "Step Count",
                    "Distance Walking Running",
                    "Basal Energy Burned",
                    "Active Energy Burned",
                    "Flights Climbed",
                    "Number Of Times Fallen",
                    "Apple Exercise Time",
                    "Dietary Caffeine",
                    "Distance Cycling",
                    "Resting Heart Rate",
                    "VO2 Max",
                    "Walking Heart Rate Average",
                    "Environmental Audio Exposure",
                    "Headphone Audio Exposure",
                    "Walking Double Support Percentage",
                    "Six Minute Walk Test Distance",
                    "Apple Stand Time",
                    "Walking Speed",
                    "Walking Step Length",
                    "Walking Asymmetry Percentage",
                    "Stair Ascent Speed",
                    "Stair Descent Speed",
                    "Sleep Duration Goal",
                    "Apple Walking Steadiness",
                    "Running Stride Length",
                    "Running Vertical Oscillation",
                    "Running Ground Contact Time",
                    "Heart Rate Recovery One Minute",
                    "Running Power",
                    "Environmental Sound Reduction",
                    "Running Speed",
                    "Time In Daylight",
                    "Physical Effort",
                    "Core Sleep",
                    "Sleep Time",
                    "Sleep Awake Time",
                    "Deep Sleep",
                    "REM Sleep",
                    "Heart Rate Variability SDNN"
                "data1_time": "1 week", "1 month", "3 month", "6 month", "All time"
                "data1_type": "max", "min", "mean", "median", "trend" (if time is "all time" then type CANNOT be "trend")
                "data2_time": same options as data1_type
                "data2_type": same options as data2_type
            }
            Remember you can ONLY select data categories from the list above, and you should output an array of two, three, or four objects of relevent data.
            ONLY respond in this format. If there is not enough information, return health information that is generally helpful in many contexts in the same format as normal output
            """
        },
        {
            "role": "user",
            "content": transcript  # The transcript received from the client
        }
    ]

    # Call API
    response = client.chat.completions.create(model="gpt-4-0125-preview",
    messages=messages,
    max_tokens=2000)

    # Extract the response content
    response_content = response.choices[0].message.content

    # Return the response content as JSON
    return jsonify({"result": response_content})

@app.route('/create_report', methods=['POST'])
def create_report():
    data = request.json
    print(request.json)
    transcript = data.get('combinedData', '')  # Extract the transcript from the incoming request

    # Define the system and user messages for OpenAI
    messages = [
        {
            "role": "system",
            "content": """You are a helpful medical data assistant. Your job is to understand the doctor-patient interaction and create a SOAP formatted formal and succinct summary of the interaction, incorporating the important data mentioned in the transcript.
            Do not include information that was not included in the conversation history. If there is not enough information, then leave sections of the soap outline blank. Do not format the different parts of the soap note with '*' or any markdown. Do not include explanations like "there is limited information so.." or anything like that. It should just be the soap notes and any other relevant medical information from the conversation. 
            """
        },
        {
            "role": "user",
            "content": transcript  # The transcript received from the client
        }
    ]

    # Call OpenAI's API
    response = client.chat.completions.create(model="gpt-4-0125-preview",
    messages=messages,
    max_tokens=2000)

    # Extract the response content
    response_content = response.choices[0].message.content

    # Return the response content as JSON
    return jsonify({"result": response_content})

if __name__ == '__main__':
    app.run(debug=True)