ai.complete

Returns a predicted completion for the given prompt.

Syntax

complete(prompt, options?)

Parameters

• prompt

  The prompt string.

• options optional

  An optional object containing the following properties. Note that not all models support all the properties. If the property is unsupported by the model, it is ignored.

  • model optional

  The name of the model to use. Defaults to the current codebook model. For example, "openchat" or "gpt-3.5-turbo".

  • tokens optional

  The number of tokens to generate. If 0 or not given, the new tokens are generated until the stop text, end-of-text token, or the end of the model’s context size.

  • temperature optional

  The temperature parameter (defaults to 0.5).

  • topK optional

  The top-k parameter (defaults to 40).

  • topP optional

  The top-p parameter (defaults to 0.95).

  • repetitionPenalty optional

  The repetition penalty parameter (defaults to 1.3).

  • seed optional

  The seed for the random number generator, a 32-bit integer. (defaults to a random value).

  Note that the random number generator and the sampling algorithm are currently not stable and will change in the future.

  • stop optional

  The stop text: a string or an array of strings, after generating which the model will stop generating futher tokens.

Return value

A completion string for the text.

Examples

ai.complete("It's always a good idea", { model: "openchat" })

" to be prepared for the unexpected."