Skip to content

Latest commit

 

History

History
87 lines (67 loc) · 4.23 KB

CONFIG.md

File metadata and controls

87 lines (67 loc) · 4.23 KB
Raw

Configuration documentation

Full Documentation Here

API Key

Add a .env file in your root directory

OPENAI_API_KEY="sk-XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX"

or set the OPENAI_API_KEY as a system environement variable.
Find your api key here

Config file

You can specify a configuration file:

node node_modules/gptdoc -c .myconfig

If none provided, the script will look for .gptdoc

Here is a sample configuration file

{
    "DEBUG": false,
    "framework": "JSDOC",
    "language": "TS",
    "tab_size": 4,
    "disableHeader": false,
    "files": {
        "src": "./src",
        "dest": "./gpt",
        "recursive": true
    },
    "openai": {
        "temperature": 0.7,
        "top_p": 1,
        "max_tokens": 256,
        "model": "text-davinci-003"
    },
    "prompt": "Add property tags with @prop",
    "minify": true
}

Config Properties

See config types

Property type description
DEBUG boolean Debug mode allows you to read & write files without prompting the OpenAI API. Placeholder doc comments are written instead.
framework string A value fed to the model to enforce a documentation framework. ex: JSDOC, typedoc, yui-doc
language string A value fed to the model to enforce a certain language
tab_size number source code's tab size in spaces
prompt string Additional prompt instructions sent to OpenAI
minify boolean whether or not to minify the code sent to OpenAI
disableHeader boolean whether or not to disable the header at the top of generated files
log boolean verbose output in the console
files IFileConfig File configuration
openai IModelConfig OpenAI configuration

File Config

IFileConfig

Property type description
src string The source directory or file
dest string The destination directory or file
recursive boolean Search files recursively through the source directory

OpenAI Config

IModelConfig

Property type description
temperature number Attention to unexpected vocabulary, ranges from 0 to 2
top_p number Common token distribution, ranges from 0 to 1
max_tokens number The maximum number of tokens the model can respond
model string model name, refer to OpenAI API Models Documentations. Use text-davinci-003 if you don't know which model to use. gpt-4 is the best model for this case, but it costs a lot more. Keep a Human in the loop, results might not be accurate