libaiac

Documentation

Index

• Constants
• Variables
• type Client
  • func NewClient(apiKey string) *Client
  • func (client *Client) Ask(ctx context.Context, prompt string, shouldRetry bool, shouldQuit bool, ...) (err error)
  • func (client *Client) GenerateCode(ctx context.Context, prompt string) (code string, err error)
  • func (client *Client) SetFull(full bool) *Client
  • func (client *Client) SetModel(model Model) *Client
• type Model
  • func (m *Model) Decode(ctx *kong.DecodeContext) error

Constants

const (
	// ModelChatGPT represents the gpt-3.5-turbo model used by ChatGPT.
	ModelChatGPT = "gpt-3.5-turbo"

	// ModelTextDaVinci3 represents the text-davinci-003 language generation
	// model.
	ModelTextDaVinci3 = "text-davinci-003"

	// ModelCodeDaVinci2 represents the code-davinci-002 code generation model.
	ModelCodeDaVinci2 = "code-davinci-002"
)

Variables

var (
	// ErrResultTruncated is returned when the OpenAI API returned a truncated
	// result. The reason for the truncation will be appended to the error
	// string.
	ErrResultTruncated = errors.New("result was truncated")

	// ErrNoResults is returned if the OpenAI API returned an empty result. This
	// should not generally happen.
	ErrNoResults = errors.New("no results return from API")

	// ErrUnsupportedModel is returned if the SetModel method is provided with
	// an unsupported model
	ErrUnsupportedModel = errors.New("unsupported model")

	// ErrUnexpectedStatus is returned when the OpenAI API returned a response
	// with an unexpected status code
	ErrUnexpectedStatus = errors.New("OpenAI returned unexpected response")

	// ErrRequestFailed is returned when the OpenAI API returned an error for
	// the request
	ErrRequestFailed = errors.New("request failed")
)

var MaxTokens = 4096

MaxTokens is the maximum amount of tokens supported by the model used. Newer OpenAI models support a maximum of 4096 tokens.

var SupportedModels = []string{ModelChatGPT, ModelTextDaVinci3, ModelCodeDaVinci2}

SupportedModels is a list of all models supported by aiac

Types

type Client

type Client struct {
	*requests.HTTPClient
	// contains filtered or unexported fields
}

Client is a structure used to continuously generate IaC code via OpenAPI/ChatGPT

func NewClient

func NewClient(apiKey string) *Client

NewClient creates a new instance of the Client struct, with the provided input options. Neither the OpenAI API nor ChatGPT are yet contacted at this point.

func (*Client) Ask

func (client *Client) Ask(
	ctx context.Context,
	prompt string,
	shouldRetry bool,
	shouldQuit bool,
	outputPath string,
) (err error)

Ask asks the OpenAI API to generate code based on the provided prompt. It is only meant to be used in command line applications (see GenerateCode for library usage). The generated code will always be printed to standard output, but may optionally be stored in the file whose path is provided by the outputPath argument. To only print to standard output, provide an empty string or a dash ("-") via outputPath. If shouldRetry is true, you will be prompted whether to regenerate the response after it is printed to standard output, in case you are unhappy with the response. If shouldQuit is true, the code is printed to standard output and the function returns, without storing to a file or asking whether to regenerate the response.

func (*Client) GenerateCode

func (client *Client) GenerateCode(ctx context.Context, prompt string) (
	code string,
	err error,
)

GenerateCode sends the provided prompt to the OpenAI API and returns the generated code.

func (*Client) SetFull

func (client *Client) SetFull(full bool) *Client

SetFull sets whether output is returned/stored in full, including explanations (if any), or if only the code is extracted. Defaults to false (meaning only code is extracted).

func (*Client) SetModel

func (client *Client) SetModel(model Model) *Client

SetModel changes the language model to use with the OpenAI API

type Model

type Model string

Model is an enum used to select the language model to use

func (*Model) Decode

func (m *Model) Decode(ctx *kong.DecodeContext) error

Decode is used by the kong library to map CLI-provided values to the Model type