stableperovskites

A API for stableperovskites, a regression model that predicts the energy above hull of a perovskite oxide. API available here.

Table of Contents

• General Info
• Technologies
• Usage

General Information

This repo shows the code for the API for stableperovskites. The API is containerized and then deployed automatically to fly.io via GitHub Actions.

Technologies

• Web server: Uvicorn
• Logging: Loguru
• Containerization: Docker
• API Serving: FastAPI
• Deployment: fly.io
• CI/CD: GitHub Actions

Usage

You can query the API either via the terminal or in Python code. You only have to supply the following:

• Composition of the perovskite oxide as a string
  • example: Ba1Sr7V8O24
• A site molecules, encoded as A_SITE_1, A_SITE_2, and A_SITE_3
  • example: Ba, Sr, etc.
• B site molecules, encoded as B_SITE_1, B_SITE_2, and B_SITE_3
  • example: V, etc.
• Number of elements
  • example: 4

To query via terminal, you can paste this on your terminal:

curl -X 'POST' \
  'https://stableperovskites-api.fly.dev/api/v1/predict' \
  -H 'accept: application/json' \
  -H 'Content-Type: application/json' \
  -d '{
  "inputs": [
    {
      "COMPOSITION": "Ba1Sr7V8O24",
      "A_SITE_1": "Ba",
      "A_SITE_2": "Sr",
      "B_SITE_1": "V",
      "NUM_ELEMS": 4
    }
  ]
}'

To query via Python code, run the following code:

import requests

headers = {
    'accept': 'application/json',
    'Content-Type': 'application/json',
}

json_data = {
    'inputs': [
        {
            'COMPOSITION': 'Ba1Sr7V8O24',
            'A_SITE_1': 'Ba',
            'A_SITE_2': 'Sr',
            'B_SITE_1': 'V',
            'NUM_ELEMS': 4,
        },
    ],
}

response = requests.post('https://stableperovskites-api.fly.dev/api/v1/predict', headers=headers, json=json_data)

Both returns a JSON object containing the following information:

• Errors (if any)
• Model version
• Predicted energy above hull or perovskite oxide

The API could handle multiple inputs as well, just send over a list of JSONs as input.