Add extra_rate config

.gitignore

@@ -16,3 +16,4 @@
 /botastic
 /test.db
 /config.yaml
+/scripts/deploy.sh

cmd/httpd/httpd.go

@@ -86,7 +86,9 @@ func NewCmdHttpd() *cobra.Command {
 			}
 			indexes := index.New(ctx, milvusClient)
 
-			userz := userServ.New(userServ.Config{}, client, users)
+			userz := userServ.New(userServ.Config{
+				ExtraRate: cfg.Sys.ExtraRate,
+			}, client, users)
 			indexService := indexServ.NewService(ctx, gptHandler, indexes, userz)
 
 			middlewarez := middlewareServ.New(middlewareServ.Config{}, indexService)

cmd/root/root.go

@@ -9,7 +9,6 @@ import (
 	"github.com/pandodao/botastic/cmd/gen"
 	"github.com/pandodao/botastic/cmd/httpd"
 	"github.com/pandodao/botastic/cmd/migrate"
-	"github.com/pandodao/botastic/cmd/worker"
 	"github.com/pandodao/botastic/cmdutil"
 	"github.com/pandodao/botastic/config"
 	"github.com/pandodao/botastic/session"
@@ -70,7 +69,6 @@ func NewCmdRoot(version string) *cobra.Command {
 	cmd.AddCommand(migrate.NewCmdMigrate())
 	cmd.AddCommand(gen.NewCmdGen())
 	cmd.AddCommand(app.NewCmdApp())
-	cmd.AddCommand(worker.NewCmdWorker())
 
 	return cmd
 }

config.example.yaml

@@ -8,6 +8,7 @@ openai:
 
 sys:
   secret_key: ""
+  extra_rate: 0.1
 
 milvus:
   address: "localhost:19530"
@@ -18,6 +19,11 @@ mixpay:
   settlement_asset_id: "31d2ea9c-95eb-3355-b65b-ba096853bc18"
   callback_url: "$API_HOST/api/callback/mixpay"
 
+order_syncer:
+  interval: 1s
+  check_interval: 10s
+  cancel_interval: 2h
+
 auth:
   jwt_secret: "abc123"
   mixin_client_secret: "...."

config/config.go

@@ -47,7 +47,8 @@ type DBConfig struct {
 }
 
 type System struct {
-	SecretKey string `yaml:"secret_key"`
+	ExtraRate float64 `yaml:"extra_rate"`
+	SecretKey string  `yaml:"secret_key"`
 }
 
 type Auth struct {

service/index/index.go

@@ -39,7 +39,7 @@ func (s *serviceImpl) createEmbeddingsWithLimit(ctx context.Context, req gogpt.E
 
 	resp, err := s.gptHandler.CreateEmbeddings(ctx, req)
 	if err == nil {
-		if err := s.userz.ConsumeCreditsByModel(ctx, userID, "text-embedding-ada-002", uint64(resp.Usage.TotalTokens)); err != nil {
+		if err := s.userz.ConsumeCreditsByModel(ctx, userID, gogpt.AdaEmbeddingV2.String(), uint64(resp.Usage.TotalTokens)); err != nil {
 			log.Printf("ConsumeCredits error: %v\n", err)
 		}
 	}

service/user/user.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 
 	"github.com/pandodao/botastic/core"
+	gogpt "github.com/sashabaranov/go-gpt3"
 	"github.com/shopspring/decimal"
 	"gorm.io/gorm"
 
@@ -28,7 +29,7 @@ func New(
 }
 
 type Config struct {
-	MixinClientSecret string
+	ExtraRate float64
 }
 
 type UserService struct {
@@ -149,20 +150,23 @@ func (s *UserService) Topup(ctx context.Context, user *core.User, amount decimal
 func (s *UserService) ConsumeCreditsByModel(ctx context.Context, userID uint64, model string, tokenCount uint64) error {
 	price := decimal.Zero
 	switch model {
-	case "gpt-3.5-turbo":
+	case gogpt.GPT3Dot5Turbo:
 		// $0.002 per 1000 tokens
 		price = decimal.NewFromFloat(0.000002)
-	case "text-davinci-003":
+	case gogpt.GPT3TextDavinci003:
 		// $0.02 per 1000 tokens
 		price = decimal.NewFromFloat(0.00002)
-	case "text-embedding-ada-002":
+	case gogpt.AdaEmbeddingV2.String():
 		// $0.0004 per 1000 tokens
 		price = decimal.NewFromFloat(0.0000004)
 	default:
 		return core.ErrInvalidModel
 	}
 
 	credits := price.Mul(decimal.NewFromInt(int64(tokenCount)))
+	if s.cfg.ExtraRate > 0 {
+		credits = credits.Mul(decimal.NewFromFloat(1 + s.cfg.ExtraRate))
+	}
 	fmt.Printf("model: %v, price: $%s, token: %d, credits: $%s\n", model, price.StringFixed(8), tokenCount, credits.StringFixed(8))
 	return s.ConsumeCredits(ctx, userID, credits)
 }

worker/rotater/rotater.go

@@ -147,7 +147,7 @@ func (w *Worker) run(ctx context.Context) error {
 		}
 
 		switch bot.Model {
-		case "gpt-3.5-turbo", "gpt-3.5-turbo-0301":
+		case gogpt.GPT3Dot5Turbo:
 			// chat completion
 			request := gogpt.ChatCompletionRequest{
 				Model:       bot.Model,
@@ -156,7 +156,7 @@ func (w *Worker) run(ctx context.Context) error {
 			}
 
 			turnReq.ChatRequest = &request
-		default:
+		case gogpt.GPT3TextDavinci003:
 			// text completion
 			prompt := bot.GetPrompt(conv, turn.Request)
 			request := gogpt.CompletionRequest{
@@ -167,6 +167,9 @@ func (w *Worker) run(ctx context.Context) error {
 				User:        conv.GetKey(),
 			}
 			turnReq.Request = &request
+		default:
+			w.UpdateConvTurnAsError(ctx, turn.ID, fmt.Errorf("unsupported model: %s", bot.Model).Error())
+			continue
 		}
 
 		if turn.Status == core.ConvTurnStatusInit {