asgi-ml

Distibuted ML-based service prototype.

Components

Web: FastAPI
ML: scikit-learn
Task Management: Celery
Tasks Queue: RabbitMQ
Results backend: Redis

How to launch

Make sure a have a Docker CLI and docker-compose executable on your machine:

docker --version
docker-compose --version

Create a folder named env on project root and put values-dev.env file on that folder:

mkdir env
touch env/values-dev.env

Fill the values-dev.env file with values:

APP_REDIS_HOST="redis"
APP_REDIS_PORT=6379
APP_REDIS_DB=0

APP_RABBITMQ_HOST='rabbitmq'
APP_RABBITMQ_PORT=5672
APP_RABBITMQ_USERNAME="user"
APP_RABBITMQ_PASSWORD="pwd"

APP_ML_N_ESTIMATORS=100
APP_ML_MAX_DEPTH=10
APP_ML_MIN_SAMPLES_ON_LEAF=2
APP_ML_CRITERION="entropy"
APP_ML_TEST_SIZE=0.05

Finally build / pull all needed images and run the deployment

docker-compose build
docker-compose up -d

You`ll see a message that Uvicorn instance is running on localhost:8080

You can see API docs at http://0.0.0.0:8080/docs endpoint.