from flask import Flask, request,jsonify ,Response 
import msgpack
import numpy as np
app = Flask(__name__)

import gzip




import onnxruntime

ort_session = onnxruntime.InferenceSession("super_resolution.onnx", providers=["CPUExecutionProvider"])

@app.route('/')
def hello_world():
    return 'hello world'

@app.route('/post', methods=['POST'])
def register():
    # print(request.headers)
    # print(request.data)
    # print(len(request.data))
    data=msgpack.unpackb(gzip.decompress(request.data) )

    # print(len(data["c"]),data["c"])
    

    data=np.frombuffer(data["spectral_data_bin"],dtype=np.uint16).reshape(int(len(data["spectral_data_bin"])/224/512/2),1,224,512)
    data=np.mean(data,axis=0, keepdims=True)
    # print(data.shape)

    # input_data=torch.tensor(data,dtype=torch.float32)/4095
    # output=model(input_data)
    # print(output.shape)

    ort_inputs = {ort_session.get_inputs()[0].name: data.astype(np.float32)/4095}
    output = ort_session.run(None, ort_inputs)[0]
    print(output)
    
    response={}
    response["Temperature"]= float(output[0,0]*(1663-1496)+1496)
    response["C"]= float(output[0,1]*(0.829-0.079)+0.079)
    response["P"]=float(output[0,1]*(0.797-0.001)+0.001)
    response["S"]=0.04402615384615385
    response["Mn"]=0.138787 
    response["Ni"]=0.035104
    response["Mo"]=0.093789
    response["Cr"]=0.002983
    d=gzip.compress(msgpack.packb(response))
    print(d)

    # d=jsonify(response)
    # print(d,type(d))

    response = Response()
    response.data=d

    print(msgpack.unpackb(gzip.decompress(d)))


    return response

if __name__ == '__main__':
    app.run(debug=True,host="0.0.0.0",port=22111)