coleygroup / aspire-context

context recommendation for ASPIRE project

Askcos Context Service

a standalone context recommendation module from Askcos rewritten as a web service in FastAPI for ASPIRE project

Table of Contents

• link
• Deployment
  • Docker
  • Native
• Testing

First steps

0. (if necessary) install git LFS: git lfs install
1. clone this repo: git clone THIS_REPO
2. pull the files from git LFS: git lfs pull

Deployment

Docker (preferred)

Automatic (via docker compose)

1. docker compose up -d.

   • the -d flag starts the container in the background so you can go about your business.

Manual

1. docker build -t TAG_NAME ..

   • we usually supply -t ASKCOS_CONTEXT
   • By default, all commands are run under the mambauser user. If you feeled compelled, you can change this value by supplying --build-arg MAMBA_USER=USERNAME to docker build.

2. docker run -d --name NAME -p 8000:8000 TAG_NAME

   • TAG_NAME should be same as the above command
   • -p 8000:8000 means (roughly) "expose port 8000 on the host machine to port 8000 on the container." It's standard for web services to be mapped to a port in the range [8000, 9000)

---

Native

Note: all steps should be run from the top-level directory of this repo

0. (if necessary) install (micro)conda
1. build the conda environment: conda env create -f env.yaml -n NAME
2. activate the environment: conda activate NAME
3. install the app package: pip install . --no-deps
4. run the service: uvicorn app.main:app --host localhost --port 8000

Testing

1. to check if the service is running:

   $ curl -X GET localhost:8000/health
   {"message":"Alive!"}