====Python Model (Method)====

The **python model** method is a [[style|function method]] that launches a Python script with a calling convention specifically designed to run calculations or machine learning models as an API.

===Additional Info.json Parameters===

This method requires a **path** parameter - this is the path to the Python script. This path can include [[substitutions|substitutions]].

It should also include a **style** parameter. The value of this parameter should be 'query' (run the AI model with user-provided inputs), or **stream** (run the AI model with recorded history).

The 'query' and 'stream' versions of this function take very different parameters.

===Query Parameters===

Your function should have one input for every input of your AI.

<code>
"parameters": [		
		{
			"name": "Temperature",			
			"type": "float",			
			"tag": "temp",
                        "mapping": "Floor Temperature",
                        "source": "Floor.Temperature - Current",
                        "default": 24.5
		}]
</code>

In the example above, we create a parameter called 'temp' of type 'float'. It's human-readable name is 'Temperature'.

When the data arrives in the Python script, the name will be 'Floor Temperature'.

In the user-interface, the default value will be 24.5 degrees, but pressing 'Sync With ARDI' will load the temperature from the live Floor Temperature value. 

===Stream Parameters===

<code>
"parameters": [		
		{
			"name": "Time",			
			"type": "timerange",			
			"tag": "start",
			"startname": "start",
			"endname": "end"
		},		
		{
			"name": "Range",			
			"type": "string",			
			"tag": "range",
			"default": "10 minutes"
		},
		{
			"name": "Samples",			
			"type": "integer",			
			"tag": "grain",
			"default": 200
		}
	]
</code>

Above are the expected parameters for a 'stream' function - the start and end times for the data, and a desired resolution.

===How It Works===

When writing your script, you'll use the **ardifunctions.py** library, which will handle ingesting all of the data for you.

Below is an example file...

<code python>
import ardifunctions
import pandas as pd
import numpy as np

defaults = {}
defaults['Floor Temperature'] = 18

dfunc = ardifunctions.DynamicFunction("http://localhost/s/op")
dfunc.query = "'Floor' ASSET AIPOINTS"
dfunc.Defaults(defaults)
dfunc.ParseParameters()

def Calculate(inputs,results):        
    return {'Output Value': inputs['Floor Temperature'] * 1.2}

dfunc.SingleCalc = Calculate

dfunc.Execute()
</code>

This will call the **Calculate** function - which is given a dictionary of parameter values - to calculate the output value(s). 

This same code works for both the 'query' or 'stream' version of the function - the only difference is that when called on a range of time rather than an instant, the 'Calculate' function will be called many times rather than just once.

===Making Solutions More Efficient===

If using machine learning models or other algorithms that work well with bulk data, you can use [[Bulk Calculation Functions]].