Merge pull request #47 from mylxsw/develop

Develop

config.yaml

@@ -92,9 +92,9 @@ fallback-openai-autoproxy: false
 # OpenAI 服务配置
 fallback-openai-organization: ""
 # 备用的 OpenAI 服务器地址，配置多个时会自动在多个服务之间平衡负载，不要忘记在在 URL 后面添加 /v1
-fallback-openai-servers: ["https://api.openai.com/v1"]
+fallback-openai-servers: [ "https://api.openai.com/v1" ]
 # 备用的 OpenAI Keys，如果指定多个，会在多个服务之间平衡负载
-fallback-openai-keys: [""]
+fallback-openai-keys: [ "" ]
 
 ######## OpenAI Dalle 配置 ########
 # OpenAI Dalle 服务配置，用于文生图
@@ -105,8 +105,8 @@ openai-dalle-autoproxy: false
 openai-dalle-azure: false
 openai-dalle-apiversion: "2023-05-15"
 openai-dalle-organization: ""
-openai-dalle-servers: ["https://api.openai.com/v1"]
-openai-dalle-keys: [""]
+openai-dalle-servers: [ "https://api.openai.com/v1" ]
+openai-dalle-keys: [ "" ]
 
 ######## Anthropic #######
 
@@ -198,7 +198,7 @@ gpt360-apikey: ""
 # https://github.com/songquanpeng/one-api
 
 # 支持的模型列表，当前支持：chatglm_turbo, chatglm_pro, chatglm_std, chatglm_lite, PaLM-2
-oneapi-support-models: []
+oneapi-support-models: [ ]
 enable-oneapi: false
 oneapi-server: ""
 oneapi-key: ""
@@ -208,7 +208,7 @@ oneapi-key: ""
 # https://openrouter.ai/
 
 # 支持的模型列表，
-openrouter-support-models: ["01-ai/yi-34b-chat"]
+openrouter-support-models: [ "01-ai/yi-34b-chat" ]
 openrouter-autoproxy: false
 enable-openrouter: false
 openrouter-server: "https://openrouter.ai/api/v1"
@@ -273,8 +273,8 @@ getimgai-key: ""
 # API 申请：https://www.lepton.ai/
 enable-leptonai: false
 leptonai-autoproxy: false
-leptonai-qr-servers: ["https://aiqr.lepton.run"]
-leptonai-keys: [""]
+leptonai-qr-servers: [ "https://aiqr.lepton.run" ]
+leptonai-keys: [ "" ]
 
 # 艺术字字体文件，只支持 ttf 格式
 font-path: "path/to/your/font/MaoKenShiJinHei-2.ttf"
@@ -420,6 +420,9 @@ cnlocal-model: "generalv2"
 default-img2img-model: lb-realistic-versionv4.0
 default-txt2img-model: sb-stable-diffusion-xl-1024-v1-0
 
+# 图生图图像识别处理模型，用于识别图像内容，生成图生图的提示语，目前支持 xfyun，留空则表示不启用
+img2img-recognition-provider: ""
+
 ######## 虚拟模型配置 ########
 enable-virtual-model: false
 virtual-model-implementation: "openai"

config/config.go

@@ -253,6 +253,9 @@ type Config struct {
 	DefaultImageToImageModel string `json:"default_image_to_image_model" yaml:"default_image_to_image_model"`
 	DefaultTextToImageModel  string `json:"default_text_to_image_model" yaml:"default_text_to_image_model"`
 
+	// 图生图图像识别处理模型，用于识别图像内容，生成图生图的提示语
+	ImageToImageRecognitionProvider string `json:"img2img-recognition-provider" yaml:"img2img-recognition-provider"`
+
 	// 虚拟模型
 	EnableVirtualModel bool         `json:"enable_virtual_model" yaml:"enable_virtual_model"`
 	VirtualModel       VirtualModel `json:"virtual_model" yaml:"virtual_model"`
@@ -529,6 +532,8 @@ func Register(ins *app.App) {
 			DefaultImageToImageModel: ctx.String("default-img2img-model"),
 			DefaultTextToImageModel:  ctx.String("default-txt2img-model"),
 
+			ImageToImageRecognitionProvider: ctx.String("img2img-recognition-provider"),
+
 			EnableVirtualModel: ctx.Bool("enable-virtual-model"),
 			VirtualModel: VirtualModel{
 				Implementation: ctx.String("virtual-model-implementation"),

config/flag.go

@@ -206,6 +206,8 @@ func initCmdFlags(ins *app.App) {
 	ins.AddStringFlag("default-img2img-model", "lb-realistic-versionv4.0", "默认的图生图模型，值取自数据表 image_model.model_id")
 	ins.AddStringFlag("default-txt2img-model", "sb-stable-diffusion-xl-1024-v1-0", "默认的文生图模型，值取自数据表 image_model.model_id")
 
+	ins.AddStringFlag("img2img-recognition-provider", "", "图生图图像识别处理模型，用于识别图像内容，生成图生图的提示语，目前支持 xfyun，留空则表示不启用")
+
 	ins.AddBoolFlag("enable-virtual-model", "是否启用虚拟模型")
 	ins.AddStringFlag("virtual-model-implementation", "openai", "虚拟模型实现厂商")
 	ins.AddStringFlag("virtual-model-nanxian-rel", "gpt-3.5-turbo", "南贤大模型实现")

docker-build.sh

@@ -1,6 +1,6 @@
 #!/usr/bin/env bash
 
-VERSION=1.0.10
+VERSION=1.0.12
 
 docker buildx build --platform=linux/amd64,linux/arm64 -t mylxsw/aidea-server:$VERSION . --push
 

internal/queue/lepton.go → internal/queue/artistic.go

@@ -10,7 +10,7 @@ import (
 	"github.com/mylxsw/aidea-server/pkg/ai/lepton"
 	"github.com/mylxsw/aidea-server/pkg/ai/openai"
 	"github.com/mylxsw/aidea-server/pkg/image"
-	repo2 "github.com/mylxsw/aidea-server/pkg/repo"
+	"github.com/mylxsw/aidea-server/pkg/repo"
 	"github.com/mylxsw/aidea-server/pkg/uploader"
 	"github.com/mylxsw/aidea-server/pkg/youdao"
 	"github.com/mylxsw/asteria/log"
@@ -68,20 +68,20 @@ func NewArtisticTextCompletionTask(payload any) *asynq.Task {
 	return asynq.NewTask(TypeArtisticTextCompletion, data)
 }
 
-func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao.Translater, up *uploader.Uploader, rep *repo2.Repository, oai openai.Client) TaskHandler {
+func BuildArtisticTextCompletionHandler(leptonClient *lepton.Lepton, translator youdao.Translater, up *uploader.Uploader, rep *repo.Repository, oai openai.Client) TaskHandler {
 	return func(ctx context.Context, task *asynq.Task) (err error) {
 		var payload ArtisticTextCompletionPayload
 		if err := json.Unmarshal(task.Payload(), &payload); err != nil {
 			return err
 		}
 
-		if err := rep.Queue.Update(context.TODO(), payload.GetID(), repo2.QueueTaskStatusRunning, nil); err != nil {
+		if err := rep.Queue.Update(context.TODO(), payload.GetID(), repo.QueueTaskStatusRunning, nil); err != nil {
 			log.WithFields(log.Fields{"payload": payload}).Errorf("set task status to running failed: %s", err)
 			return err
 		}
 
 		if payload.CreatedAt.Add(5 * time.Minute).Before(time.Now()) {
-			rep.Queue.Update(context.TODO(), payload.GetID(), repo2.QueueTaskStatusFailed, ErrorResult{Errors: []string{"任务处理超时"}})
+			rep.Queue.Update(context.TODO(), payload.GetID(), repo.QueueTaskStatusFailed, ErrorResult{Errors: []string{"任务处理超时"}})
 			log.WithFields(log.Fields{"payload": payload}).Errorf("task expired")
 			return nil
 		}
@@ -92,8 +92,8 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 				err = err2.(error)
 
 				// 更新创作岛历史记录
-				if err := rep.Creative.UpdateRecordByTaskID(ctx, payload.GetUID(), payload.GetID(), repo2.CreativeRecordUpdateRequest{
-					Status: repo2.CreativeStatusFailed,
+				if err := rep.Creative.UpdateRecordByTaskID(ctx, payload.GetUID(), payload.GetID(), repo.CreativeRecordUpdateRequest{
+					Status: repo.CreativeStatusFailed,
 					Answer: err.Error(),
 				}); err != nil {
 					log.WithFields(log.Fields{"payload": payload}).Errorf("update creative failed: %s", err)
@@ -104,7 +104,7 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 				if err := rep.Queue.Update(
 					context.TODO(),
 					payload.GetID(),
-					repo2.QueueTaskStatusFailed,
+					repo.QueueTaskStatusFailed,
 					ErrorResult{
 						Errors: []string{err.Error()},
 					},
@@ -146,7 +146,7 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 			panic(fmt.Errorf("不支持该艺术字类型: %v", payload.Type))
 		}
 
-		res, err := client.ImageGenerate(ctx, lepton.QRImageRequest{
+		res, err := leptonClient.ImageGenerate(ctx, lepton.QRImageRequest{
 			Prompt:            prompt,
 			NegativePrompt:    negativePrompt,
 			Model:             payload.ArtisticType,
@@ -171,7 +171,7 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 
 		imageUrls, err := res.UploadResources(ctx, up, payload.GetUID())
 		if err != nil {
-			log.With(payload).Errorf("上传图片到七牛云存储失败: %", err)
+			log.With(payload).Errorf("上传图片到七牛云存储失败: %v", err)
 			panic(fmt.Errorf("上传图片到七牛云存储失败: %w", err))
 		}
 
@@ -190,14 +190,14 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 			panic(err)
 		}
 
-		req := repo2.CreativeRecordUpdateRequest{
-			Status:    repo2.CreativeStatusSuccess,
+		req := repo.CreativeRecordUpdateRequest{
+			Status:    repo.CreativeStatusSuccess,
 			Answer:    string(retJson),
 			QuotaUsed: payload.GetQuota(),
 		}
 
 		if prompt != payload.Prompt || negativePrompt != payload.NegativePrompt {
-			ext := repo2.CreativeRecordUpdateExtArgs{}
+			ext := repo.CreativeRecordUpdateExtArgs{}
 			if prompt != payload.Prompt {
 				ext.RealPrompt = prompt
 			}
@@ -219,15 +219,15 @@ func BuildArtisticTextCompletionHandler(client *lepton.Lepton, translator youdao
 			ctx,
 			payload.GetUID(),
 			payload.GetQuota(),
-			repo2.NewQuotaUsedMeta("leptonai", modelUsed...),
+			repo.NewQuotaUsedMeta("leptonai", modelUsed...),
 		); err != nil {
 			log.Errorf("used quota add failed: %s", err)
 		}
 
 		return rep.Queue.Update(
 			context.TODO(),
 			payload.GetID(),
-			repo2.QueueTaskStatusSuccess,
+			repo.QueueTaskStatusSuccess,
 			CompletionResult{
 				Resources:   imageUrls,
 				ValidBefore: time.Now().Add(7 * 24 * time.Hour),

internal/queue/consumer/provider.go

@@ -95,6 +95,7 @@ func (p Provider) Boot(resolver infra.Resolver) {
 		userSvc *service.UserService,
 		dalleClient *openai.DalleImageClient,
 		leptonClient *lepton.Lepton,
+		aiProvider *chat.AIProvider,
 	) {
 		log.Debugf("register all queue handlers")
 		mux.HandleFunc(queue.TypeOpenAICompletion, queue.BuildOpenAICompletionHandler(openaiClient, rep))
@@ -106,7 +107,7 @@ func (p Provider) Boot(resolver infra.Resolver) {
 		mux.HandleFunc(queue.TypeSignup, queue.BuildSignupHandler(rep, mailer, ding))
 		mux.HandleFunc(queue.TypePayment, queue.BuildPaymentHandler(rep, mailer, que, ding))
 		mux.HandleFunc(queue.TypeBindPhone, queue.BuildBindPhoneHandler(rep, mailer))
-		mux.HandleFunc(queue.TypeImageGenCompletion, queue.BuildImageCompletionHandler(leapClient, stabaiClient, deepaiClient, fromstonClient, dashscopeClient, getimgaiClient, translater, uploader, rep, openaiClient, dalleClient))
+		mux.HandleFunc(queue.TypeImageGenCompletion, queue.BuildImageCompletionHandler(conf, aiProvider, leapClient, stabaiClient, deepaiClient, fromstonClient, dashscopeClient, getimgaiClient, translater, uploader, rep, openaiClient, dalleClient))
 		mux.HandleFunc(queue.TypeFromStonCompletion, queue.BuildFromStonCompletionHandler(fromstonClient, uploader, rep))
 		mux.HandleFunc(queue.TypeDashscopeImageCompletion, queue.BuildDashscopeImageCompletionHandler(dashscopeClient, uploader, rep, translater, openaiClient))
 		mux.HandleFunc(queue.TypeGetimgAICompletion, queue.BuildGetimgAICompletionHandler(getimgaiClient, translater, uploader, rep, openaiClient))