Search code examples
apiflaskallennlp

How do you get a response from a Flask app with curl?


I'm building a Flask test predictor using AllenNLP.

I'm passing 'passage' and 'question' from a .json file to the predictor.

However, when I pass the json file using curl, it doesn't return a response. Is there a special return in Flask to get it?

Code looks like:

 from allennlp.predictors.predictor import Predictor as AllenNLPPredictor


from flask import Flask
from flask import request
app = Flask(__name__)

@app.route("/", methods=['GET','POST'])
def hello():
    return "<h1>Test app!</h1>"


class PythonPredictor:
    def __init__(self, config):
        self.predictor = AllenNLPPredictor.from_path(
            "https://storage.googleapis.com/allennlp-public-models/bidaf-elmo-model-2018.11.30-charpad.tar.gz"
        )

    def predict(self, payload):
        if request.method == "POST":
            prediction = self.predictor.predict(
                passage=payload["passage"], question=payload["question"]
            )
            return prediction["best_span_str"]

Curl command looks like: curl http://127.0.0.1:5000 -X POST -H "Content-Type: application/json" -d @sample.json


Solution

  • Unless I've misunderstood (I'm guessing you're asking how to obtain the JSON submission in your route, and return the result) it sounds like you need to do something like:

    p = PythonPredictor()
    
    @app.route("/", methods=['POST'])
    def hello():
        data = request.get_json()
        result = p.predict(data)
        return result
    

    This effectively runs the data in your sample.json through your PythonPredictor.predict method, and returns that prediction to the client.

    Notice this code creates the instance p outside the route function, so that the NLP model is only loaded when your flask app starts (not on every request). However it looks like this may re-download that file, unless AllenNLPPredictor.from_path does some caching, so it would probably be advisable to manually download that file to your own storage first, and load from there in the PythonPredictor.__init__ function.

    Let me know if any of this needs clarification, or I've missunderstood.