Updating HTTP service example to consoletest

python.rst

Python

This is an example of how you can use the web API from Python.

To test this

$ python -m dffml.util.testing.consoletest docs/python.rst

First start the server

Warning

By no means should you use the following command in a production environment! You should instead see the security docs!

$ dffml service http server -insecure -cors '*' -addr 0.0.0.0 -port 8080

Configure a model

configure.py

import sys
import json
import urllib.request

# The URL that the HTTP service is running on
# http://127.0.0.1:8080/
LABEL = sys.argv[-1]
MODEL = sys.argv[-2]
ENDPOINT = sys.argv[-3]

print(urllib.request.urlopen(
  f"{ENDPOINT}/configure/model/{MODEL}/{LABEL}",
  data=json.dumps(
    {
      "model": {
        "plugin": None,
        "config": {
          "directory": {
            "plugin": [
              "mymodel"
            ],
            "config": {}
          },
          "predict": {
            "plugin": "Y:int:1",
            "config": {}
          },
          "features": {
            "plugin": [
              "X:int:1",
            ],
            "config": {}
          }
        }
      }
    }
  ).encode()
).read())

$ python configure.py http://localhost:8080 slr mymodel

Create a context for a model

context.py

import sys
import json
import urllib.request

# The URL that the HTTP service is running on
# http://127.0.0.1:8080/
CONTEXT = sys.argv[-1]
MODEL = sys.argv[-2]
ENDPOINT = sys.argv[-3]

print(urllib.request.urlopen(
  f"{ENDPOINT}/context/model/{MODEL}/{CONTEXT}",
).read())

$ python context.py http://localhost:8080 mymodel mymodelcontext

print("Hello World")

*Note*: On successful creation and configuration the server will return  {"error": null}

## Context Creation

URL = "https://127.0.0.1:5000/context/mdoel/{label}/{ctx_label}".format(label = "mymodel", ctx_label = "ctx_mymodel")
result = requests.get(url = URL, params = {})

*Note*: On successful creation of a context the server will return {"error": null}

## Train the Model

URL = "https://127.0.0.1:5000/model/{ctx_label}/train".format(ctx_label = "ctx_mymodel")
params = {
  [
    "my_training_dataset"
  ]
}
result = requests.post(url = URL, params = PARAMS)

*Note*: On successful execution the server will return {"error": null}

## Assess Accuracy

URL = "https://127.0.0.1:5000/model/{ctx_label}/accuracy".format(ctx_label = "ctx_mymodel")
params = {
  [
    "my_test_dataset"
  ]
}
result = requests.post(url = URL, params = PARAMS)

*Note*: On successful execution the response will be a JSON object containing the accuracy as a float value : {"accuracy": 0.42}

## Make Prediction

URL = "https://127.0.0.1:5000/model/{ctx_label}/predict/0".format(ctx_label = "ctx_mymodel")
PARAMS = {
  {
    "42": {
      "features": {
        "by_ten": 420
      }
    }
  }
}

*Note*: The JSON passed as param maps key of the record to the JSON representation of dffml.record.Record as received by the source record endpoint

result = requests.post(url = URL, params = PARAMS)

*Note*: On successful execution the response will be a JSON object similar to this:

response = {
  "iterkey": null,
  "records": {
    "42": {
      "key": "42",
      "features": {
        "by_ten": 420
      },
      "prediction": {
        "confidence": 42,
        "value": 4200
      },
      "last_updated": "2019-10-15T08:19:41Z",
      "extra": {}
    }
  }
}