gpt3

README

go-gpt3

An OpenAPI GPT-3 API client enabling Go/Golang programs to interact with the gpt3 APIs.

Supports using the completion APIs with or without streaming.

Usage

Simple usage to call the main gpt-3 API, completion:

client := gpt3.NewClient(apiKey)
resp, err := client.Completion(ctx, gpt3.CompletionRequest{
    Prompt: []string{"2, 3, 5, 7, 11,"},
})

fmt.Print(resp.Choices[0].Text)
// prints " 13, 17, 19, 23, 29, 31", etc

Documentation

Check out the go docs for more detailed documentation on the types and methods provided: https://pkg.go.dev/github.com/PullRequestInc/go-gpt3

Full Examples

Try out any of these examples with putting the contents in a main.go and running go run main.go. I would recommend using go modules in which case you will also need to run go mod init within your test repo. Alternatively you can clone this repo and run the test script with go run cmd/test/main.go.

You will also need to have a .env file that looks like this to use these examples:

API_KEY=<openAI API Key>

package main

import (
	"context"
	"fmt"
	"log"
	"os"

	"github.com/PullRequestInc/go-gpt3"
	"github.com/joho/godotenv"
)

func main() {
	godotenv.Load()

	apiKey := os.Getenv("API_KEY")
	if apiKey == "" {
		log.Fatalln("Missing API KEY")
	}

	ctx := context.Background()
	client := gpt3.NewClient(apiKey)

	resp, err := client.Completion(ctx, gpt3.CompletionRequest{
		Prompt:    []string{"The first thing you should know about javascript is"},
		MaxTokens: gpt3.IntPtr(30),
		Stop:      []string{"."},
		Echo:      true,
	})
	if err != nil {
		log.Fatalln(err)
	}
	fmt.Println(resp.Choices[0].Text)
}

Support

• List Engines API
• Get Engine API
• Completion API (this is the main gpt-3 API)
• Streaming support for the Completion API
• Document Search API
• Overriding default url, user-agent, timeout, and other options

Powered by

Documentation

Index

• Constants
• func IntPtr(i int) *int
• type Client
  • func NewClient(apiKey string, options ...ClientOption) Client
• type ClientOption
  • func WithBaseURL(baseURL string) ClientOption
  • func WithDefaultEngine(engine string) ClientOption
  • func WithTimeout(timeout time.Duration) ClientOption
  • func WithUserAgent(userAgent string) ClientOption
• type CompletionRequest
• type CompletionResponse
• type CompletionResponseChoice
• type EngineObject
• type EnginesResponse
• type SearchData
• type SearchRequest
• type SearchResponse

Constants

const (
	AdaEngine     = "ada"
	BabbageEngine = "babbage"
	CurieEngine   = "curie"
	DavinciEngine = "davinci"
	DefaultEngine = DavinciEngine
)

Engine Types

Functions

func IntPtr

func IntPtr(i int) *int

IntPtr converts an integer to an *int as a convenience

Types

type Client

type Client interface {
	// Engines lists the currently available engines, and provides basic information about each
	// option such as the owner and availability.
	Engines(ctx context.Context) (*EnginesResponse, error)

	// Engine retrieves an engine instance, providing basic information about the engine such
	// as the owner and availability.
	Engine(ctx context.Context, engine string) (*EngineObject, error)

	// Completion creates a completion with the default engine. This is the main endpoint of the API
	// which auto-completes based on the given prompt.
	Completion(ctx context.Context, request CompletionRequest) (*CompletionResponse, error)

	// CompletionStream creates a completion with the default engine and streams the results through
	// multiple calls to onData.
	CompletionStream(ctx context.Context, request CompletionRequest, onData func(*CompletionResponse)) error

	// CompletionWithEngine is the same as Completion except allows overriding the default engine on the client
	CompletionWithEngine(ctx context.Context, engine string, request CompletionRequest) (*CompletionResponse, error)

	// CompletionStreamWithEngine is the same as CompletionStream except allows overriding the default engine on the client
	CompletionStreamWithEngine(ctx context.Context, engine string, request CompletionRequest, onData func(*CompletionResponse)) error

	// Search performs a semantic search over a list of documents with the default engine.
	Search(ctx context.Context, request SearchRequest) (*SearchResponse, error)

	// SearchWithEngine performs a semantic search over a list of documents with the specified engine.
	SearchWithEngine(ctx context.Context, engine string, request SearchRequest) (*SearchResponse, error)
}

A Client is an API client to communicate with the OpenAI gpt-3 APIs

func NewClient

func NewClient(apiKey string, options ...ClientOption) Client

NewClient returns a new OpenAI GPT-3 API client. An apiKey is required to use the client

type ClientOption

type ClientOption func(*client) error

ClientOption are options that can be passed when creating a new client

func WithBaseURL

func WithBaseURL(baseURL string) ClientOption

WithBaseURL is a client option that allows you to override the default base url of the client. The default base url is "https://api.openai.com/v1"

func WithDefaultEngine

func WithDefaultEngine(engine string) ClientOption

WithDefaultEngine is a client option that allows you to override the default engine of the client

func WithTimeout

func WithTimeout(timeout time.Duration) ClientOption

WithTimeout is a client option that allows you to override the default timeout duration of requests for the client. The default is 30 seconds.

func WithUserAgent

func WithUserAgent(userAgent string) ClientOption

WithUserAgent is a client option that allows you to override the default user agent of the client

type CompletionRequest

type CompletionRequest struct {
	// The engine ID
	EngineID string `json:"engine_id,omitempty"`
	// A list of string prompts to use.
	// TODO there are other prompt types here for using token integers that we could add support for.
	Prompt []string `json:"prompt"`
	// How many tokens to complete up to. Max of 512
	MaxTokens *int `json:"max_tokens,omitempty"`
	// Sampling temperature to use
	Temperature *float32 `json:"temperature,omitempty"`
	// Alternative to temperature for nucleus sampling
	TopP *float32 `json:"top_p,omitempty"`
	// How many choice to create for each prompt
	N *float32 `json:"n"`
	// Include the probabilities of most likely tokens
	LogProbs *int `json:"logprobs"`
	// Echo back the prompt in addition to the completion
	Echo bool `json:"echo"`
	// Up to 4 sequences where the API will stop generating tokens. Response will not contain the stop sequence.
	Stop []string `json:"stop,omitempty"`
	// PresencePenalty number between 0 and 1 that penalizes tokens that have already appeared in the text so far.
	PresencePenalty float32 `json:"presence_penalty"`
	// FrequencyPenalty number between 0 and 1 that penalizes tokens on existing frequency in the text so far.
	FrequencyPenalty float32 `json:"presence_penalty"`

	// Whether to stream back results or not. Don't set this value in the request yourself
	// as it will be overriden depending on if you use CompletionStream or Completion methods.
	Stream bool `json:"stream,omitempty"`
}

CompletionRequest is a request for the completions API

type CompletionResponse

type CompletionResponse struct {
	ID      string                     `json:"id"`
	Object  string                     `json:"object"`
	Created int                        `json:"created"`
	Model   string                     `json:"model"`
	Choices []CompletionResponseChoice `json:"choices"`
}

CompletionResponse is the full response from a request to the completions API

type CompletionResponseChoice

type CompletionResponseChoice struct {
	Text         string `json:"text"`
	Index        int    `json:"index"`
	LogProbs     *int   `json:"logprobs"`
	FinishReason string `json:"finish_reason"`
}

CompletionResponseChoice is one of the choices returned in the response to the Completions API

type EngineObject

type EngineObject struct {
	ID     string `json:"id"`
	Object string `json:"object"`
	Owner  string `json:"owner"`
	Ready  bool   `json:"ready"`
}

EngineObject contained in an engine reponse

type EnginesResponse

type EnginesResponse struct {
	Data   []EngineObject `json:"data"`
	Object string         `json:"object"`
}

EnginesResponse is returned from the Engines API

type SearchData

type SearchData struct {
	Document int     `json:"document"`
	Object   string  `json:"search_resulet"`
	Score    float64 `json:"score"`
}

SearchData is a single search result from the document search API

type SearchRequest

type SearchRequest struct {
	EngineID  string   `json:"engine_id"`
	Documents []string `json:"documents"`
	Query     string   `json:"query"`
}

SearchRequest is a request for the document search API

type SearchResponse

type SearchResponse struct {
	Data   []SearchData `json:"data"`
	Object string       `json:"object"`
}

SearchResponse is the full response from a request to the document search API