Pycell

-Local Installation-

• Copy pycell.py and table.html into a project folder. By default the code allows for concurrent runs of Pycell by different users, therefore unique files to user runs are stored in a local directory while pycell can be stored in a cloud directory. 

• Create a folder within your documents folder called CodeGen also create a subfolder within CodeGen called documents. Alternatively, within pycell you can change the target root folder to store credentials and output files.

• Install dependent libraries, [pandas, numpy, flask, string, and openai]

• Ensure your openai api keys are stored within keys.txt in CodeGen.

• Run pycell.py.

Commands:

• Pycell first attempts to run what is entered in the cell, as though it is a python command. That means that you can easily generate and process data easily with integrated python support. 

• If it cannot run the entered text as native python it tries to split the text into value and source arguments, which are split by a double colon, "<value>::<source args>". Source args do not have to be given of course. 

• At this point if the value is not prefaced by an '=' sign then either the input cell is fileld with the value, or if source args are given then the input cell is filled with all source args concatenated. 

• When text is prefaced by an '=' sign it is treated as a command. 

  • Python Scripts - 

    • Run python scripts stored in documents folder. E.g entering '=HelloWorld.py' into a cell will run the python script 'HelloWorld', given it's in CodeGen/Documents

    • Run python functions. 

    • To fill cell using a python script the script must print the desired output. 

  • Targets and Sources:

    • Again, pycell  will attempt to execute simple python commands when prefaced with an equals sign here. That is, suppose one has a local variable being used by their local pycell instance, <myVar>. They can print the value of the variable, <myVar>, in the cell using 'print(<myVar>)' or using '=<myVar>'. Note, entering '=print(<myVar>)' will fill the cell with nothing, because print has no output. 

    • Target Arguments - These are the cells which will be filled

      • Structure : [<initial coordinate>, 'to', <final coordinate>, number steps] * Only <initial coordinate> is required when targeting just a single cell.

    • Source Arguments - These are the cells which will be used to fill the target.

      • Structure type A : [<initial coordinate>, 'to', <final coordinate>, number steps] * Only <initial coordinate> is required when sourcing just a single cell. 

      • Structure type B: [<text 1>, <text 2>,...]

    • The length of target args and source args do not have to be identical. Suppose we have [[x0, y0], [x1, y1], [x2,y2]]::[[0, 0], [1, 1]], then [x0, y0] will take the value in cell [0, 0], [x1, y1] will take the value in cell [1, 1], and [x2, y2] will take the value in cell [0, 0]. Note the same operation can be created using the 'to' keyword as [[x0,y0], 'to', [x2, y2], 3]::[[0, 0], [1, 1]]. 

  • Built-In Commands

AI(args) will generate an AI text response given the args. 

promt - type string: Text input for ai text response. 

 form_data -  type array: Leave empty or if referencing the form for input type form_data=form_data.

 conversation - type dictionary : See openai gpt conversation structure

max_tokens, temperature, fine_tuned_model - See openai.  

cform(args) Will convert the dictionary output into a more readable format. 

data - the raw data from an AI output. 

form_data -  type array: Leave empty or if referencing the form for input type form_data=form_data.