Flaskerk

Provide OpenAPI document and validation for flask service.

Mainly built for Machine Learning Model services.

If you're using Falcon, check my another library Falibrary.

Attention

I've created a new library named SpecTree. It supports all the features of Flaskerk, but with a better interface. Give it a try! This library may be archived in the future.

Features

• Generate API document with Redoc UI or Swagger UI :yum:
• Less boilerplate code, annotations are really easy-to-use :sparkles:
• Validate query, JSON data, response data with pydantic :wink:
• Better HTTP exceptions for API services (default & customized) (JSON instead of HTML) :grimacing:

Quick Start

install with pip install flaskerk (Python 3.6+)

Simple demo

from flask import Flask, request, jsonify
from flaskerk import Flaskerk
from pydantic import BaseModel

class Query(BaseModel):
    text: str

app = Flask(__name__)
api = Flaskerk()

@app.route('/api/classify')
@api.validate(query=Query)
def classify():
    print(request.query)
    return jsonify(label=0)

if __name__ == "__main__":
    api.register(app)
    app.run()

Changes you need to make:

• create model with pydantic
• decorate the route function with Flaskerk.validate()
• specify which part you need in validate
  • query (args in url)
    • builtin converters (string, path, any, int, float, uuid)
  • data (JSON data from request)
  • resp (response) this will be transformed to JSON data after validation
  • x (HTTP Exceptions list)
  • tags (tags for this API route)
• register to Flask application

After that, this library will help you validate the incoming request and provide API document in /docs.

For more details, check the document.

More feature

from flask import Flask, request
from pydantic import BaseModel, Schema
from random import random
from flaskerk import Flaskerk, HTTPException

app = Flask(__name__)
api = Flaskerk(
    title='Demo Service',
    version='1.0',
    ui='swagger',
)

class Query(BaseModel):
    text: str

class Response(BaseModel):
    label: int
    score: float = Schema(
        ...,
        gt=0,
        lt=1,
    )

class Data(BaseModel):
    uid: str
    limit: int
    vip: bool

e233 = HTTPException(code=233, msg='lucky for you')

@app.route('/api/predict/<string(length=2):source>/<string(length=2):target>', methods=['POST'])
@api.validate(query=Query, data=Data, resp=Response, x=[e233], tags=['model'])
def predict(source, target):
    """
    predict demo

    demo for `query`, `data`, `resp`, `x`
    """
    print(f'=> from {source} to {target}')  # path
    print(f'Data: {request.json_data}')  # Data
    print(f'Query: {request.query}')  # Query
    if random() < 0.5:
        e233.abort('bad luck')
    return Response(label=int(10 * random()), score=random())

if __name__ == '__main__':
    api.register(app)
    app.run()

try it with http POST ':5000/api/predict/zh/en?text=hello' uid=0b01001001 limit=5 vip=true

Open the docs in http://127.0.0.1:5000/docs .

For more examples, check examples.

FAQ

Can I just do the validation without generating API document?

Sure. If you don't register it to Flask application, there won't be document routes.