#First we have to import libraries
#Think of libraries as "pre-written programs" that help us accelerate what we do in Python

#Gradio is a web interface library for deploying machine learning models
import gradio as gr

#Pickle is a library that lets us work with machine learning models, which in Python are typically in a "pickle" file format
import pickle

#Orange is the Python library used by... well, Orange!
from Orange.data import * 

#This is called a function. This function can be "called" by our website (when we click submit). Every time it's called, the function runs.
#Within our function, there are inputs (bedrooms1, bathrooms1, etc.). These are passed from our website front end, which we will create further below.
def make_prediction(gender1,married1,dependents1,education1,self_employed1,applicantincome1,coapplicantincome1,loanamount1,loanamountterm1,credit_history1,property_area1):
    
	#Because we already trained a model on these variables, any inputs we feed to our model has to match the inputs it was trained on.
	#Even if you're not familiar with programming, you can probably decipher the below code.
    gender=DiscreteVariable("Gender",values=["Male","Female"])
    married=DiscreteVariable("Married",values=["Yes","No"])
    dependents=DiscreteVariable("Dependents",values=["0","1","2","3+"])
    education=DiscreteVariable("Education",values=["Graduate","Not Graduate"])
    self_employed=DiscreteVariable("Self_Employed",values=["Yes","No"])
    applicantincome=ContinuousVariable("ApplicantIncome")
    coapplicantincome=ContinuousVariable("CoapplicantIncome")
    loanamount=ContinuousVariable("LoanAmount")
    loanamountterm=ContinuousVariable("Loan_Amount_Term")
    credit_history=DiscreteVariable("Credit_History",values=["0","1"])
    property_area=DiscreteVariable("Property_Area",values=["Rural","Semiurban","Urban"])

	#This code is a bit of housekeeping. 
	#Since our model is expecting discrete inputs (just like in Orange), we need to convert our numeric values to strings
    dependents1=str(dependents1)
    credit_history1=str(credit_history1)
    applicantincome1=float((applicantincome1-5403.46)/1.13)
    print(applicantincome1)
    coapplicantincome1=float((coapplicantincome1-1621.25)/1.80347)
    print(coapplicantincome1)
    loanamount1=float((loanamount1-146.41)/0.58)
    print(loanamount1)
    loanamountterm1=float((loanamountterm1-342)/0.19)
    print(loanamountterm1)
	#A domain is essentially an Orange file definition. Just like the one you set with the "file node" in the tool.
    domain=Domain([gender,married,dependents,education,self_employed,applicantincome,coapplicantincome,loanamount,loanamountterm,credit_history,property_area])
	
	#Our data is the data being passed by the website inputs. This gets mapped to our domain, which we defined above.
    data=Table(domain,[[gender1,married1,dependents1,education1,self_employed1,applicantincome1,coapplicantincome1,loanamount1,loanamountterm1,credit_history1,property_area1]])
	
	#Next, we can work on our predictions!

	#This tiny piece of code loads our model (pickle load).
    with open("model.pkcls", "rb") as f:
		#Then feeds our data into the model, then sets the "preds" variable to the prediction output for our class variable, which is price.	
        clf  = pickle.load(f)
        ar=clf(data)
        preds=clf.domain.class_var.str_val(ar)
		#Finally, we send the prediction to the website.
        return preds

#Now that we have defined our prediction function, we need to create our web interface.
#This code creates the input components for our website. Gradio has this well documented and it's pretty easy to modify.
TheGender=gr.Dropdown(["Male","Female"],label="Whats your gender?")
IsMarried=gr.Dropdown(["Yes","No"],label="Are you married?")
HasDependents=gr.Dropdown(["0","1","2","3+"], label="How many dependents do you have?")
#HasDependents=gr.Slider(minimum=0,maximum=3,step=1,label="How many dependents do you have?")
IsEducated=gr.Dropdown(["Graduate","Not Graduate"],label="Whats your education status?")
IsSelfEmployed=gr.Dropdown(["Yes","No"],label="Are you self-employed?")
ApplicantIncom=gr.Number(label="Whats the applicant income?")
CoApplicantIncome=gr.Number(label="Whats the co-applicant income? If any!")
LoanAmount=gr.Number(label="Whats the Loan Amount?")
LoanAmountTerm=gr.Number(label="Whats the Loan Amount Term?")
HasCreditHistory=gr.Dropdown(["0","1"],label="Do you have credit history?")
PropertyArea=gr.Dropdown(['Rural','Semiurban','Urban'],label='What is the area where your property is located?')

# Next, we have to tell Gradio what our model is going to output. In this case, it's going to be a text result (house prices).
output = gr.Textbox(label="Loan Approval Status: ")

#Then, we just feed all of this into Gradio and launch the web server. 
#Our fn (function) is our make_prediction function above, which returns our prediction based on the inputs.
app = gr.Interface(fn = make_prediction, inputs=[TheGender, IsMarried, HasDependents,IsEducated,IsSelfEmployed,ApplicantIncom,CoApplicantIncome,LoanAmount,LoanAmountTerm,HasCreditHistory,PropertyArea], outputs=output)
app.launch()