hrubanj/model_serving_assignment

Setup environment

All code should be compatible with Python >=3.9.

Use either virtual environment:

make initialize-venv

Or docker (if you do not have compatible python version):

docker run -it --rm  -p "5000:5000" -v "$(pwd):/tmp" python:3.10-slim-buster /bin/sh
apt-get update && apt-get install make && cd tmp

Install requirements:

make install-requirements

Run tests

Run all tests via:

make run-tests

Run specific test file via:

python3 -m pytest tests/<directory_name>

Run Path Analyzer

Run analyzer function that does not account for gaps between months. (Note that gaps can still be present but this function just returns a minimum and maximum)

Example:

from path_analyzer.analyzer import report_min_max_no_gaps


report_min_max_no_gaps(
    "s3://my-bucket", # bucket
    "xxx/yyy/zzz/def", # full path
    "/Users/jiri/PycharmProjects/model_serving_assignment/data/min_max_no_gaps.json", # output file
)

Run Model Serving API

Set environmental variable for directory where data gets saved. For example:

 export DATA_DIRECTORY="data/"

make run-api

The API is exposed at http://localhost:5000.

Endpoints

All POST requests must be in JSON format.

GET "/"

Status code: 200

Returns application name and version.

GET "/status"

Status code: 200

Returns metrics that can be scraped by Prometheus.

GET "/metrics"

Status code: 200

Returns metrics that can be scraped by Prometheus.

GET "/status"

Status code: 200

Returns "OK" if the application is running.

POST "/label_sentiment"

Request format:

{
  "text": "string",
  "languageCode": "string"
}

Example:

{
  "text": "This task is awesome. Nope, just kidding.",
  "languageCode": "en"
}

Successful response:

Status code: 200

{
  "sentiment": {
    "schema": {
      "type": "string",
      "enum": [
        "neutral",
        "positive",
        "negative"
      ]
    }
  }
}

Example:

{
  "sentiments": "positive"
}

Errors:

Invalid request format:

Status code: 400

If request is not a valid JSON, status code 400 is returned.

Invalid request content:

Status code: 400

If required fields are not present or cannot be coerced to string, returns error description with status code 400.

Example:

{
  "errors": [
    {
      "loc": [
        "languageCode"
      ],
      "msg": "field required",
      "type": "value_error.missing"
    }
  ]
}

POST "/save_sentiment"

Request format:

{
  "text": "string",
  "languageCode": "string",
  "isGoodTranslation": "boolean",
  "sentiment": {
    "schema": {
      "type": "string",
      "enum": [
        "neutral",
        "positive",
        "negative"
      ]
    }
  }
}

Example:

{
  "text": "Hello",
  "languageCode": "EN",
  "isGoodTranslation": true,
  "sentiment": "positive"
}

Successful response:

Status code: 201

{}

Errors:

Invalid request format:

Status code: 400

If request is not a valid JSON, status code 400 is returned.

Invalid request content:

Status code: 400

If required fields are not present or cannot be coerced to correct data types, returns error description with status code 400.

Example:

{
  "errors": [
    {
      "loc": [
        "languageCode"
      ],
      "msg": "field required",
      "type": "value_error.missing"
    }
  ]
}

Duplicate item:

Status code: 409

Item with the same content is already saved.

Example:

{
  "errors": [
    [
      "item already exists"
    ]
  ]
}