I am trying to serve a Neural Network using FastAPI. The manual site http://localhost:8000/docs#/default/predict_predict_post works fine and translates into the following curl command: which also works. When I try to query the API using python requests: I only get the "422 Unprocessable Entity" Errors. Where am I going wrong here? Answer You provide a data argument to requests.post, which does

HTTPS POST to query FastAPI using python requests

I am trying to serve a Neural Network using FastAPI.

from fastapi import Depends, FastAPI
from pydantic import BaseModel
from typing import Dict

class iRequest(BaseModel):
    arg1: str
    arg2: str

class iResponse(BaseModel):
    pred: str
    probs: Dict[str, float]

@app.post("/predict", response_model=iResponse)
def predict(request: iRequest, model: Model = Depends(get_model)):
    pred, probs = model.predict(request.arg1, request.arg2)
    return iResponse(pred = pred, probs = probs)

The manual site http://localhost:8000/docs#/default/predict_predict_post works fine and translates into the following curl command:

curl -X POST "http://localhost:8000/predict" -H  "accept: application/json" -H  "Content-Type: application/json" -d "{"arg1":"I am the King","arg2":"You are not my King"}"

which also works. When I try to query the API using python requests:

import requests
data = {"arg1": "I am the King",
        "arg2": "You are not my King"}
r = requests.post("http://localhost:8000/predict", data=data)

I only get the “422 Unprocessable Entity” Errors. Where am I going wrong here?

Answer

You provide a data argument to requests.post, which does a POST with Content-Type: application/x-www-form-urlencoded, which is not JSON.

Consider using requests.post(url, json=data) and you should be fine.

Advertisement

Answer