Completion-cli NPM

completion-cli

npx completion-cli --help

Answers questions. This tool was done for personal use, but feel free to use it.

Requirements

NodeJS v19+

Api

Define OPENAI_API_KEY environment variable with your OpenAI key.

Simplest one. Request creative reply of 2000 tokens long.

npx completion-cli -t 2000 -T 1 -p "Explain me the world:"

Send the text from blogpost.txt using supported option in the config file.

I tend to use this option almost exclusively.

npx completion-cli -j ./path/to/config.json -f ./path/to/blogpost.txt

Example of a json config here

Options

Pass either cli flags or a JSON-config path. The CLI and JSON options are combined, CLI overwrites JSON. But only one of --prompt or --filePath or json.prompt is allowed.

--help
--filePath or -f, the relative path to the text file. Example: completion-cli -f ./path/to/blogpost.txt
--jsonConfig or -j, path to a json file with any of the Open AI Supported Options(#supported-onot both, and any of these overwrites ptions). Example: completion-cli -j ./path/to/config.json

The CLI only supports a small subset of OpenAI options

--prompt or -p: one way to pass text. Example: completion-cli -p Explain me the world:. The length limit is 15000 bytes (ascii characters.)
--model or -m, the default model used is text-davinci-003. Other flavours are: text-curie-001, text-babbage-001. Default text-davinci-003
--tokens or -t the expected response length in tokens (max is 4096, about 1700 english words.)`
--temperature or -T (capital T). How creative the model is. Float between 0 and 1. Default 0.
--echo or -e. Explicit boolean. Indicates whether to concatenate the response to the prompt or not. You need to specify --echo [true, false]

All the values in the object are optional. For explanations and default values check the schema file.

{
  "model": string;
  "prompt": string;
  "suffix": string;
  "max_tokens": number;
  "temperature": number;
  "top_p": number;
  "echo": boolean;
  "stop": string | string[];
  "presence_penalty": number;
  "frequency_penalty": number;
  "prePromptString": string;
  "postPromptString": string;
}