Cache or Reuse Mongodb connection in AWS lambda using Python

Question

I'm building a serverless application using Python and Mongodb. In documentation I found that I need to write db connection outside handler function. I have used Mangum python package as adapter to handle API gateway.

from fastapi import FastAPI, Body, status, Depends

from mangum import Mangum

from motor.motor_asyncio import AsyncIOMotorClient

from fastapi.responses import JSONResponse

from app.utility.config import MONGODB_URL, MAX_CONNECTIONS_COUNT, MIN_CONNECTIONS_COUNT, MAX_DB_THREADS_WAIT_COUNT, MAX_DB_THREAD_QUEUE_TIMEOUT_COUNT

application= FastAPI()

client = AsyncIOMotorClient(str(MONGODB_URL),

                                   maxPoolSize=MAX_CONNECTIONS_COUNT,

                                   minPoolSize=MIN_CONNECTIONS_COUNT,

                                   waitQueueMultiple = MAX_DB_THREADS_WAIT_COUNT,

                                   waitQueueTimeoutMS = MAX_DB_THREAD_QUEUE_TIMEOUT_COUNT )

async def get_database() -> AsyncIOMotorClient:

   

    return client

@application.post("/createStudent")

async def create_student(student = Body(...), db: AsyncIOMotorClient = Depends(get_database)):

    new_student = await db["college"]["students"].insert_one(student)

    created_student = await db["college"]["students"].find_one({"_id": new_student.inserted_id})

    return JSONResponse(status_code=status.HTTP_201_CREATED, content=created_student)

@application.post("/createTeacher")

async def create_teacher(teacher = Body(...), db: AsyncIOMotorClient = Depends(get_database)):

    new_teacher = await db["college"]["students"].insert_one(teacher)

    created_teacher = await db["college"]["students"].find_one({"_id": new_teacher.inserted_id})

    return JSONResponse(status_code=status.HTTP_201_CREATED, content=created_teacher)

handler = Mangum(application)

For every API request, new connection is created. How to cache db so that new request uses old connection. Every time new request is created so that lambda compute time is increased dramatically after db hits max connection limit.

There are examples for node js but for python I could not find anywhere