Parameterize maxTokens

This commit is contained in:
Matt Low 2023-11-05 18:29:57 +00:00
parent 0c2149663f
commit 7cb2a4e017
2 changed files with 15 additions and 19 deletions

View File

@ -7,6 +7,9 @@ import (
"github.com/spf13/cobra" "github.com/spf13/cobra"
) )
// TODO: allow setting with flag
const MAX_TOKENS = 256
var rootCmd = &cobra.Command{ var rootCmd = &cobra.Command{
Use: "lmcli", Use: "lmcli",
Short: "Interact with Large Language Models", Short: "Interact with Large Language Models",
@ -120,7 +123,7 @@ var newCmd = &cobra.Command{
response <- HandleDelayedResponse(receiver) response <- HandleDelayedResponse(receiver)
}() }()
err = CreateChatCompletionStream(messages, receiver) err = CreateChatCompletionStream(messages, MAX_TOKENS, receiver)
if err != nil { if err != nil {
Fatal("%v\n", err) Fatal("%v\n", err)
} }
@ -160,7 +163,7 @@ var promptCmd = &cobra.Command{
receiver := make(chan string) receiver := make(chan string)
go HandleDelayedResponse(receiver) go HandleDelayedResponse(receiver)
err := CreateChatCompletionStream(messages, receiver) err := CreateChatCompletionStream(messages, MAX_TOKENS, receiver)
if err != nil { if err != nil {
Fatal("%v\n", err) Fatal("%v\n", err)
} }

View File

@ -8,7 +8,7 @@ import (
openai "github.com/sashabaranov/go-openai" openai "github.com/sashabaranov/go-openai"
) )
func CreateChatCompletionRequest(messages []Message) *openai.ChatCompletionRequest { func CreateChatCompletionRequest(messages []Message, maxTokens int) openai.ChatCompletionRequest {
chatCompletionMessages := []openai.ChatCompletionMessage{} chatCompletionMessages := []openai.ChatCompletionMessage{}
for _, m := range messages { for _, m := range messages {
chatCompletionMessages = append(chatCompletionMessages, openai.ChatCompletionMessage{ chatCompletionMessages = append(chatCompletionMessages, openai.ChatCompletionMessage{
@ -17,23 +17,19 @@ func CreateChatCompletionRequest(messages []Message) *openai.ChatCompletionReque
}) })
} }
return &openai.ChatCompletionRequest{ return openai.ChatCompletionRequest{
Model: openai.GPT4, Model: openai.GPT3Dot5Turbo,
MaxTokens: 256,
Messages: chatCompletionMessages, Messages: chatCompletionMessages,
Stream: true, MaxTokens: maxTokens,
} }
} }
// CreateChatCompletion submits a Chat Completion API request and returns the // CreateChatCompletion submits a Chat Completion API request and returns the
// response. // response.
func CreateChatCompletion(messages []Message) (string, error) { func CreateChatCompletion(messages []Message, maxTokens int) (string, error) {
client := openai.NewClient(config.OpenAI.APIKey) client := openai.NewClient(config.OpenAI.APIKey)
resp, err := client.CreateChatCompletion( req := CreateChatCompletionRequest(messages, maxTokens)
context.Background(), resp, err := client.CreateChatCompletion(context.Background(), req)
*CreateChatCompletionRequest(messages),
)
if err != nil { if err != nil {
return "", err return "", err
} }
@ -43,16 +39,13 @@ func CreateChatCompletion(messages []Message) (string, error) {
// CreateChatCompletionStream submits a streaming Chat Completion API request // CreateChatCompletionStream submits a streaming Chat Completion API request
// and streams the received response to the provided output channel. // and streams the received response to the provided output channel.
func CreateChatCompletionStream(messages []Message, output chan string) error { func CreateChatCompletionStream(messages []Message, maxTokens int, output chan string) error {
client := openai.NewClient(config.OpenAI.APIKey) client := openai.NewClient(config.OpenAI.APIKey)
ctx := context.Background() req := CreateChatCompletionRequest(messages, maxTokens)
req := CreateChatCompletionRequest(messages)
req.Stream = true
defer close(output) defer close(output)
stream, err := client.CreateChatCompletionStream(ctx, *req) stream, err := client.CreateChatCompletionStream(context.Background(), req)
if err != nil { if err != nil {
return err return err
} }