refactor(model): 重构模型网关实体和映射逻辑

# Conflicts:
#	go.sum

Dockerfile

@@ -1,43 +1,23 @@
-# 多阶段构建 - 第一阶段：编译（使用已安装的镜像）
-FROM golang:1.26-alpine3.23 AS builder
+# 阶段1: 构建
+FROM golang:alpine AS builder
 
 RUN apk add --no-cache git ca-certificates tzdata
 
+ENV TZ=Asia/Shanghai
+RUN ln -snf /usr/share/zoneinfo/$TZ /etc/localtime && echo $TZ > /etc/timezone
+
 ENV GO111MODULE=on
 ENV GOPROXY=https://goproxy.cn,direct
 ENV CGO_ENABLED=0
 ENV GOTOOLCHAIN=auto
-ENV GOPRIVATE=gitea.com/red-future/common
-
-# 配置git使用私有Gitea仓库（带Token认证）
-RUN git config --global url."http://x-token-auth:619679cd366aefea3a50f0622d842a41f2209e08595767bba49c3836ef57d415@116.204.74.41:3000/red-future/common.git".insteadOf "https://gitea.com/red-future/common.git" && \
-    git config --global credential.helper store
-
 WORKDIR /build
 
-# 复制父目录的 common 模块(因为 go.mod 中使用了本地 replace)
-#COPY ../common /build/common
 COPY . .
 
 RUN go mod download && go mod tidy
 
 RUN go build -ldflags="-s -w" -o main ./main.go
 
-# 第二阶段：运行
-FROM alpine:3.23
-
-ENV TIME_ZONE=Asia/Shanghai
-RUN apk add --no-cache ca-certificates tzdata && \
-    ln -sf /usr/share/zoneinfo/$TIME_ZONE /etc/localtime
-
-WORKDIR /app
-
-# 复制编译好的二进制文件
-COPY --from=builder /build/main .
-COPY --from=builder /build/config.yml ./
-
-# 创建日志目录
-RUN mkdir -p /logs /app/resource/log/run /app/resource/log/server
 
 EXPOSE 3004
 

common/util/convert.go

@@ -0,0 +1,10 @@
+package util
+
+import "github.com/gogf/gf/v2/util/gconv"
+
+// ConvertTo 转换为指定类型
+func ConvertTo[T any](v interface{}) *T {
+	var t T
+	_ = gconv.Struct(v, &t)
+	return &t
+}

common/util/files.go

@@ -1,6 +1,7 @@
 package util
 
 import (
+	"encoding/json"
 	"fmt"
 	"net/http"
 	"os"
@@ -67,3 +68,48 @@ func SaveTmpResult(taskID string, data []byte, ext string) (string, error) {
 	}
 	return path, nil
 }
+
+// SaveTempFileByType
+// 根据传入的数据自动判断：
+// 若是 []byte 且后缀为 .mp3 → 保存二进制音频
+// 若是任意结构体/map → 自动转 JSON 保存
+// 返回：新临时文件路径、错误
+func SaveTempFileByType(taskID string, data any, oldTmpFile string) (string, error) {
+	// 1. 先清理旧临时文件（统一逻辑）
+	if oldTmpFile != "" {
+		_ = os.Remove(oldTmpFile)
+	}
+
+	var tmpPath string
+	var tmpErr error
+
+	// 2. 判断是否是二进制音频（[]byte + .mp3）
+	if audioData, ok := data.([]byte); ok {
+		tmpPath, tmpErr = saveTmpResult(taskID, audioData, ".mp3")
+	} else {
+		// 3. 其他类型 → 序列化为 JSON 保存
+		mappedBytes, err := json.Marshal(data)
+		if err != nil {
+			return "", err
+		}
+		if len(mappedBytes) == 0 {
+			return "", nil
+		}
+		tmpPath, tmpErr = saveTmpResult(taskID, mappedBytes, ".json")
+	}
+
+	if tmpErr != nil || tmpPath == "" {
+		return "", tmpErr
+	}
+
+	return tmpPath, nil
+}
+
+// saveTmpResult 你原有的底层保存文件方法（保留不动）
+func saveTmpResult(taskID string, data []byte, ext string) (string, error) {
+	// 你原来实现，比如：
+	filename := taskID + ext
+	tmpPath := filepath.Join(os.TempDir(), filename)
+	err := os.WriteFile(tmpPath, data, 0644)
+	return tmpPath, err
+}

common/util/headers.go

@@ -3,7 +3,7 @@ package util
 import (
 	"context"
 
-	"gitea.com/red-future/common/utils"
+	"gitea.redpowerfuture.com/red-future/common/utils"
 	"github.com/gogf/gf/v2/frame/g"
 	"github.com/gogf/gf/v2/util/gconv"
 )
@@ -77,24 +77,3 @@ func SetTaskHeadersToCtx(ctx context.Context, headers map[string]string) context
 	}
 	return ctx
 }
-
-// ParseStoredPayload 解析入库的 request_payload，拆出模型调用 payload 与透传 headers
-// 入库格式：{"payload": <any>, "headers": {"Authorization": "...", "X-User-Info":"..."}}
-func ParseStoredPayload(v any) (payload any, headers map[string]string) {
-	if v == nil {
-		return nil, nil
-	}
-	m := gconv.Map(v)
-	if len(m) == 0 {
-		return v, nil
-	}
-	if h, ok := m["headers"]; ok {
-		headers = gconv.MapStrStr(h)
-	}
-	if p, ok := m["payload"]; ok {
-		payload = p
-	} else {
-		payload = v
-	}
-	return
-}

common/util/json.go

@@ -1,28 +0,0 @@
-package util
-
-import (
-	"encoding/json"
-
-	"github.com/gogf/gf/v2/container/gvar"
-)
-
-func ParseJSONField(field any) any {
-	var v *gvar.Var
-	switch val := field.(type) {
-	case *gvar.Var:
-		v = val
-	default:
-		return field
-	}
-
-	if v == nil || v.IsNil() || v.IsEmpty() {
-		return nil
-	}
-
-	str := v.String()
-	var result any
-	if json.Unmarshal([]byte(str), &result) == nil {
-		return result
-	}
-	return str
-}

common/util/mapping.go

@@ -0,0 +1,303 @@
+package util
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"model-gateway/model/entity"
+	"net/http"
+	"net/url"
+	"regexp"
+	"strings"
+	"time"
+
+	"github.com/gogf/gf/v2/encoding/gjson"
+	"github.com/gogf/gf/v2/frame/g"
+	"github.com/gogf/gf/v2/util/gconv"
+	tgjson "github.com/tidwall/gjson"
+)
+
+// ParseAndValidate 解析并校验结果
+func ParseAndValidate(raw map[string]any, model *entity.ModelGatewayModel) (map[string]any, error) {
+	// 1) 解析 content 字符串为 rounds 数组
+	contentVal, ok := raw[entity.ResponseBody]
+	if !ok {
+		return raw, fmt.Errorf("字段 %s 不存在", entity.ResponseBody)
+	}
+	contentStr, ok := contentVal.(string)
+	if !ok || strings.TrimSpace(contentStr) == "" {
+		return raw, fmt.Errorf("字段 %s 为空或不是字符串", entity.ResponseBody)
+	}
+	var arr []any
+	if err := json.Unmarshal([]byte(contentStr), &arr); err != nil {
+		return raw, fmt.Errorf("JSON解析失败: %w", err)
+	}
+	if len(arr) == 0 {
+		return raw, fmt.Errorf("解析后数组为空")
+	}
+
+	// 2) 校验必填字段
+	if len(model.RequiredFields) > 0 {
+		for i, r := range arr {
+			round, ok := r.(map[string]any)
+			if !ok {
+				continue
+			}
+			for _, field := range model.RequiredFields {
+				if gjson.New(round).Get(field).IsNil() {
+					return raw, fmt.Errorf("rounds[%d] 缺少必填字段: %s", i, field)
+				}
+			}
+		}
+	}
+
+	return map[string]any{"total_rounds": len(arr), "rounds": arr}, nil
+}
+
+// ParseStructResult 解析结构结果
+func ParseStructResult(raw map[string]any, responseBody string) map[string]any {
+	contentVal := raw[responseBody]
+	// 是字符串，尝试解析
+	contentStr := gconv.String(contentVal)
+	if contentStr == "" || contentStr == "0" {
+		return map[string]any{
+			"total_rounds": 1,
+			"rounds":       []map[string]any{{responseBody: raw}},
+		}
+	}
+
+	// 尝试解析为数组
+	var arr []any
+	if err := json.Unmarshal([]byte(contentStr), &arr); err == nil && len(arr) > 0 {
+		return map[string]any{
+			"total_rounds": 1,
+			"rounds":       []map[string]any{{responseBody: arr}},
+		}
+	}
+
+	// 尝试解析为单个对象
+	var parsed any
+	if err := json.Unmarshal([]byte(contentStr), &parsed); err == nil {
+		return map[string]any{
+			"total_rounds": 1,
+			"rounds":       []map[string]any{{responseBody: parsed}},
+		}
+	}
+
+	// 兜底：原始字符串作为内容
+	return map[string]any{
+		"total_rounds": 1,
+		"rounds":       []map[string]any{{responseBody: contentStr}},
+	}
+}
+
+// ParseHeadMsgHeaders 从 head_msg JSON 中提取请求头
+// head_msg 格式示例：
+//
+//	{
+//	    "Authorization": "Bearer xxx",
+//	    "Content-Type": "application/json",
+//	    "X-Api-App-Id": "5147401364",
+//	    "X-Api-Access-Key": "VCqRX7..."
+//	}
+func ParseHeadMsgHeaders(headMsg map[string]any) map[string]string {
+	if len(headMsg) == 0 {
+		return nil
+	}
+	out := make(map[string]string, len(headMsg))
+	for k, v := range headMsg {
+		out[k] = gconv.String(v)
+	}
+	return out
+}
+
+// MapResponsePayload 映射模型响应为标准格式
+func MapResponsePayload(mapping map[string]any, result map[string]any) (map[string]any, error) {
+	if len(mapping) == 0 {
+		return result, nil
+	}
+
+	// 把 result 转成 JSON 字符串，tidwall/gjson 需要字符串输入
+	resultBytes, _ := json.Marshal(result)
+	resultStr := string(resultBytes)
+
+	mapped := make(map[string]any)
+
+	for standardField, modelPath := range mapping {
+		path := gconv.String(modelPath)
+		if path == "" {
+			continue
+		}
+
+		value := tgjson.Get(resultStr, path)
+		if !value.Exists() {
+			continue
+		}
+		// 如果是数组路径（含 #），取 Array；否则取单值
+		if strings.Contains(path, "#") {
+			var arr []any
+			for _, v := range value.Array() {
+				arr = append(arr, v.Value())
+			}
+			mapped[standardField] = arr
+		} else {
+			mapped[standardField] = value.Value()
+		}
+	}
+
+	return mapped, nil
+}
+
+//
+//// GetModelBody 获取数据库中保存的模型信息
+//func GetModelBody(v map[string]any) map[string]any {
+//	if v == nil {
+//		return nil
+//	}
+//	if p, ok := v["body"]; ok {
+//		return gconv.Map(p)
+//	}
+//	return v
+//}
+
+// BodyToQuery 将 body 转为 url.Values
+func BodyToQuery(payload map[string]any) (url.Values, error) {
+	q := url.Values{}
+	for k, v := range payload {
+		if v == nil {
+			continue
+		}
+		q.Set(k, gconv.String(v))
+	}
+	return q, nil
+}
+
+// PullTaskResult 轮询查询异步任务结果直到完成
+func PullTaskResult(ctx context.Context, body map[string]any, queryConfig map[string]any, headMsg map[string]any) (map[string]any, error) {
+	// 1) 解析配置
+	// 1.1 提取 taskID
+	taskIDPath := gconv.String(queryConfig["task_id"])
+	taskID := gconv.String(gjson.New(body).Get(taskIDPath).Val())
+	if taskID == "" {
+		return nil, fmt.Errorf("无法从路径 %s 提取 taskID", taskIDPath)
+	}
+	g.Log().Infof(ctx, "[PullTaskResult] taskID=%s", taskID)
+
+	// 1.2 请求地址，替换 {id}
+	queryUrl := gconv.String(queryConfig["url"])
+	queryUrl = replaceURLParams(queryUrl, map[string]any{"id": taskID})
+
+	// 1.3 请求方式
+	method := gconv.String(queryConfig["method"])
+	if method == "" {
+		method = "GET"
+	}
+
+	// 1.4 状态判断配置
+	statusPath := gconv.String(queryConfig["status_path"])
+	statusValues, _ := queryConfig["status_values"].(map[string]any)
+	if statusPath == "" {
+		statusPath = "status"
+	}
+
+	// 1.5 轮询间隔
+	interval := gconv.Int(queryConfig["interval_seconds"])
+	if interval <= 0 {
+		interval = 2
+	}
+
+	// 1.6 请求体
+	reqBodyMap := map[string]any{"task_id": taskID}
+
+	// 2) 轮询请求
+	for {
+		select {
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		default:
+		}
+
+		var reqBody io.Reader
+		if method == "POST" {
+			bs, _ := json.Marshal(reqBodyMap)
+			reqBody = bytes.NewReader(bs)
+		}
+
+		req, err := http.NewRequestWithContext(ctx, method, queryUrl, reqBody)
+		if err != nil {
+			return nil, fmt.Errorf("创建请求失败: %w", err)
+		}
+
+		// 统一用 headMsg 注入请求头
+		for hk, hv := range ParseHeadMsgHeaders(headMsg) {
+			req.Header.Set(hk, hv)
+		}
+
+		client := &http.Client{Timeout: 30 * time.Second}
+		resp, err := client.Do(req)
+		if err != nil {
+			g.Log().Warningf(ctx, "[PullTaskResult] 请求失败 taskID=%s err=%v", taskID, err)
+			time.Sleep(time.Duration(interval) * time.Second)
+			continue
+		}
+
+		raw, _ := io.ReadAll(resp.Body)
+		_ = resp.Body.Close()
+
+		g.Log().Infof(ctx, "[PullTaskResult] taskID=%s statusCode=%d body=%s", taskID, resp.StatusCode, string(raw))
+
+		if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+			time.Sleep(time.Duration(interval) * time.Second)
+			continue
+		}
+
+		var result map[string]any
+		_ = json.Unmarshal(raw, &result)
+
+		statusVal := gjson.New(result).Get(statusPath).Val()
+		statusStr := gconv.String(statusVal)
+		g.Log().Infof(ctx, "[PullTaskResult] 状态 taskID=%s status=%v", taskID, statusVal)
+
+		if matchStatus(statusStr, statusValues["succeeded"]) {
+			g.Log().Infof(ctx, "[PullTaskResult] 任务成功 taskID=%s", taskID)
+			return result, nil
+		}
+
+		if matchStatus(statusStr, statusValues["failed"]) {
+			g.Log().Errorf(ctx, "[PullTaskResult] 任务失败 taskID=%s", taskID)
+			return result, fmt.Errorf("任务失败")
+		}
+
+		time.Sleep(time.Duration(interval) * time.Second)
+	}
+}
+
+func matchStatus(actual string, expected any) bool {
+	expectedStr := gconv.String(expected)
+	if actual == expectedStr {
+		return true
+	}
+	switch v := expected.(type) {
+	case []any:
+		for _, item := range v {
+			if actual == gconv.String(item) {
+				return true
+			}
+		}
+	}
+	return false
+}
+
+// replaceURLParams 替换 URL 中的 {key}
+func replaceURLParams(url string, params map[string]any) string {
+	re := regexp.MustCompile(`\{([^}]+)}`)
+	return re.ReplaceAllStringFunc(url, func(s string) string {
+		key := strings.Trim(s, "{}")
+		if val, ok := params[key]; ok {
+			return gconv.String(val)
+		}
+		return s
+	})
+}

common/util/streaming.go

@@ -0,0 +1,150 @@
+package util
+
+import (
+	"encoding/base64"
+	"encoding/json"
+	"fmt"
+	"sort"
+	"strings"
+
+	"github.com/gogf/gf/v2/encoding/gjson"
+)
+
+// ================================================================
+
+// ParseStreamResponse 流式响应解析（通用入口）
+func ParseStreamResponse(rawBytes []byte, streamConfig map[string]any) (map[string]any, error) {
+	enabled, _ := streamConfig["enabled"].(bool)
+	if !enabled {
+		return gjson.New(string(rawBytes)).Map(), nil
+	}
+
+	parser, _ := streamConfig["parser"].(string)
+	if parser == "base64_concat" {
+		return parseBase64Stream(rawBytes)
+	}
+
+	return parseSSEStream(rawBytes, streamConfig)
+}
+
+// parseBase64Stream 拼接流式 base64 并解码为二进制（TTS 等音频模型）
+func parseBase64Stream(rawBytes []byte) (map[string]any, error) {
+	lines := strings.Split(string(rawBytes), "\n")
+	var audioBase64 strings.Builder
+
+	for _, line := range lines {
+		line = strings.TrimSpace(line)
+		if line == "" {
+			continue
+		}
+
+		var chunk map[string]any
+		if err := json.Unmarshal([]byte(line), &chunk); err != nil {
+			continue
+		}
+
+		if data, ok := chunk["data"].(string); ok && data != "" {
+			audioBase64.WriteString(data)
+		}
+	}
+
+	cleanBase64 := strings.Map(func(r rune) rune {
+		if r == ' ' || r == '\n' || r == '\r' || r == '\t' {
+			return -1
+		}
+		return r
+	}, audioBase64.String())
+
+	audioBytes, err := base64.StdEncoding.DecodeString(cleanBase64)
+	if err != nil {
+		audioBytes, err = base64.RawStdEncoding.DecodeString(cleanBase64)
+		if err != nil {
+			return nil, fmt.Errorf("base64 解码失败: %w", err)
+		}
+	}
+
+	return map[string]any{"audio": audioBytes}, nil
+}
+
+// parseSSEStream SSE 流式解析（图片模型等）
+func parseSSEStream(rawBytes []byte, streamConfig map[string]any) (map[string]any, error) {
+	events, _ := streamConfig["events"].([]any)
+	if len(events) == 0 {
+		return gjson.New(string(rawBytes)).Map(), nil
+	}
+
+	lines := strings.Split(string(rawBytes), "\n")
+	result := make(map[string]any)
+	var partials []map[string]any
+
+	for _, line := range lines {
+		line = strings.TrimSpace(line)
+		if line == "" || line == "[DONE]" {
+			continue
+		}
+		if strings.HasPrefix(line, "event:") {
+			continue
+		}
+		if strings.HasPrefix(line, "data:") {
+			line = strings.TrimPrefix(line, "data:")
+			line = strings.TrimSpace(line)
+		}
+
+		var chunk map[string]any
+		if err := json.Unmarshal([]byte(line), &chunk); err != nil {
+			continue
+		}
+
+		chunkType, _ := chunk["type"].(string)
+
+		for _, evt := range events {
+			e, _ := evt.(map[string]any)
+			match, _ := e["match"].(string)
+			if !strings.Contains(chunkType, match) {
+				continue
+			}
+
+			fields, _ := e["fields"].(map[string]any)
+			aggregateTo, _ := e["aggregate_to"].(string)
+			evtType, _ := e["type"].(string)
+
+			switch evtType {
+			case "partial":
+				item := make(map[string]any)
+				for localKey, chunkKey := range fields {
+					item[localKey] = chunk[chunkKey.(string)]
+				}
+				partials = append(partials, item)
+
+			case "final":
+				for localKey, chunkKey := range fields {
+					val := gjson.New(chunk).Get(chunkKey.(string))
+					if !val.IsNil() {
+						if _, exists := result[aggregateTo]; !exists {
+							result[aggregateTo] = make(map[string]any)
+						}
+						result[aggregateTo].(map[string]any)[localKey] = val.Val()
+					}
+				}
+			}
+		}
+	}
+
+	if len(partials) > 0 {
+		for _, evt := range events {
+			e, _ := evt.(map[string]any)
+			if e["type"] == "partial" {
+				if orderBy, ok := e["order_by"].(string); ok {
+					sort.Slice(partials, func(i, j int) bool {
+						return fmt.Sprint(partials[i][orderBy]) < fmt.Sprint(partials[j][orderBy])
+					})
+				}
+				result[e["aggregate_to"].(string)] = partials
+				break
+			}
+		}
+	}
+
+	mergedBytes, _ := json.Marshal(result)
+	return gjson.New(mergedBytes).Map(), nil
+}

config.yml

@@ -61,6 +61,10 @@ jaeger:
 
 # 本地调试用：可选自动执行 worker/cleaner（默认关闭）
 asynch:
+  queryPending:
+    enabled: false
+    intervalSeconds: 10  # 每10秒轮询一次
+    limit: 10            # 每次查10条
   worker:
     enabled: false
     intervalSeconds: 5

consts/public/public.go

@@ -1,14 +1,35 @@
 package public
 
+const (
+	CallModeSync   = 0 // 同步调用
+	CallModeAsync  = 1 // 异步调用
+	CallModeStream = 2 // 流式调用
+)
+
+const (
+	TaskStatusPending    = 0 // 排队中
+	TaskStatusRunning    = 1 // 执行中
+	TaskStatusSuccess    = 2 // 成功
+	TaskStatusFailed     = 3 // 失败
+	TaskStatusDownloaded = 4 // 已下载
+)
+
+const (
+	BuildTypePrompt = 1 //提示词构建
+	BuildTypeNode   = 2 //节点构建
+	BuildTypeStruct = 3 //结构构建
+)
+
 // ModelType 模型类型常量
 const (
 	ModelTypeInference = 100 // 推理模型
 
-	ModelTypeImage             = 200 // 图片模型
-	ImageSubTypeTextToImage    = 201 // 图片模型-文生图
-	ImageSubTypeImageToImage   = 202 // 图片模型-图生图
-	ImageSubTypeImageEdit      = 203 // 图片模型-图片编辑
-	ImageSubTypeImageVariation = 204 // 图片模型-图片变体
+	ModelTypeImage               = 200 // 图片模型
+	ImageSubTypeTextToImage      = 201 // 图片模型-文生图
+	ImageSubTypeImageToImage     = 202 // 图片模型-图生图
+	ImageSubTypeImageEdit        = 203 // 图片模型-图片编辑
+	ImageSubTypeImageVariation   = 204 // 图片模型-图片变体
+	ImageSubTypeImageTextToImage = 205 // 图片模型-图文生图
 
 	ModelTypeAudio             = 300 // 音频模型
 	AudioSubTypeTextToSpeech   = 301 // 音频模型-文生音
@@ -28,18 +49,18 @@ const (
 	VideoSubTypeImageToVideo     = 602 // 视频模型-图生视频
 	VideoSubTypeImageTextToVideo = 603 // 视频模型-图文生视频
 	VideoSubTypeVideoToVideo     = 604 // 视频模型-视频生视频
-	VideoSubTypeVideoEdit        = 605 // 视频模型-视频编辑
 )
 
 // ModelTypeName 模型类型名称映射
 var ModelTypeName = map[int]string{
 	ModelTypeInference: "推理模型",
 
-	ModelTypeImage:             "图片模型",
-	ImageSubTypeTextToImage:    "图片模型-文生图",
-	ImageSubTypeImageToImage:   "图片模型-图生图",
-	ImageSubTypeImageEdit:      "图片模型-图片编辑",
-	ImageSubTypeImageVariation: "图片模型-图片变体",
+	ModelTypeImage:               "图片模型",
+	ImageSubTypeTextToImage:      "图片模型-文生图",
+	ImageSubTypeImageToImage:     "图片模型-图生图",
+	ImageSubTypeImageEdit:        "图片模型-图片编辑",
+	ImageSubTypeImageVariation:   "图片模型-图片变体",
+	ImageSubTypeImageTextToImage: "图片模型-图文生图",
 
 	ModelTypeAudio:             "音频模型",
 	AudioSubTypeTextToSpeech:   "音频模型-文生音",
@@ -59,7 +80,6 @@ var ModelTypeName = map[int]string{
 	VideoSubTypeImageToVideo:     "视频模型-图生视频",
 	VideoSubTypeImageTextToVideo: "视频模型-图文生视频",
 	VideoSubTypeVideoToVideo:     "视频模型-视频生视频",
-	VideoSubTypeVideoEdit:        "视频模型-视频编辑",
 }
 
 // 运营商常量

consts/public/table_name.go

@@ -5,8 +5,8 @@ const (
 )
 
 const (
-	TableNameModel = "asynch_models"   // 模型表
-	TableNameTask  = "asynch_task"     // 任务表
-	TableNameOpLog = "logs_model_op"   // 操作日志表
-	TableNameStat  = "logs_model_stat" // 按天统计表（请求次数）
+	TableNameModel = "model_gateway_models"    // 模型表
+	TableNameTask  = "model_gateway_task"      // 任务表
+	TableNameOpLog = "model_gateway_logs_op"   // 操作日志表
+	TableNameStat  = "model_gateway_logs_stat" // 按天统计表
 )

controller/base.go

@@ -1 +0,0 @@
-package controller

controller/model_gateway_logs_stat_controller.go

@@ -2,17 +2,17 @@ package controller
 
 import (
 	"context"
+	statService "model-gateway/service/stat"
 
 	"model-gateway/model/dto"
-	"model-gateway/service"
 )
 
+// ModelGatewayLogsStat 统计控制器
+var ModelGatewayLogsStat = new(stat)
+
 type stat struct{}
 
-// Stat 统计控制器
-var Stat = new(stat)
-
 // ListModelStat 统计列表
 func (c *stat) ListModelStat(ctx context.Context, req *dto.ListModelStatReq) (res *dto.ListModelStatRes, err error) {
-	return service.Stat.List(ctx, req)
+	return statService.ModelGatewayLogsStat.List(ctx, req)
 }

controller/model_gateway_models_controller.go

@@ -2,65 +2,65 @@ package controller
 
 import (
 	"context"
-
 	"model-gateway/model/dto"
-	"model-gateway/service"
+	modelService "model-gateway/service/model"
+	"model-gateway/service/queue"
 )
 
+// ModelGatewayModels 模型配置控制器
+var ModelGatewayModels = new(model)
+
 type model struct{}
 
-// Model 模型配置控制器
-var Model = new(model)
-
 // CreateModel 添加配置
 func (c *model) CreateModel(ctx context.Context, req *dto.CreateModelReq) (res *dto.CreateModelRes, err error) {
-	return service.Model.Create(ctx, req)
+	return modelService.ModelGatewayModels.Create(ctx, req)
 }
 
 // UpdateModel 更改配置
 func (c *model) UpdateModel(ctx context.Context, req *dto.UpdateModelReq) (res *dto.UpdateModelRes, err error) {
-	err = service.Model.Update(ctx, req)
+	err = modelService.ModelGatewayModels.Update(ctx, req)
 	return
 }
 
 // DeleteModel 删除配置
 func (c *model) DeleteModel(ctx context.Context, req *dto.DeleteModelReq) (res *dto.DeleteModelRes, err error) {
-	err = service.Model.Delete(ctx, req)
+	err = modelService.ModelGatewayModels.Delete(ctx, req)
 	return
 }
 
 // GetModel 获取配置详情
 func (c *model) GetModel(ctx context.Context, req *dto.GetModelReq) (res *dto.GetModelRes, err error) {
-	return service.Model.Get(ctx, req)
+	return modelService.ModelGatewayModels.Get(ctx, req)
 }
 
 // ListModel 配置列表
 func (c *model) ListModel(ctx context.Context, req *dto.ListModelReq) (res *dto.ListModelRes, err error) {
-	return service.Model.List(ctx, req)
+	return modelService.ModelGatewayModels.List(ctx, req)
 }
 
 // AutoTune 动态调参（由上层定时任务每小时触发一次）
 func (c *model) AutoTune(ctx context.Context, req *dto.AutoTuneReq) (res *dto.AutoTuneRes, err error) {
-	return service.AutoTune(ctx, req)
+	return queue.AutoTune(ctx, req)
 }
 
 // ListType 模型类型列表
 func (c *model) ListType(ctx context.Context, req *dto.ListTypeReq) (res *dto.TypeItem, err error) {
-	return service.GetModelTypesFromConfig()
+	return modelService.GetModelTypesFromConfig()
 }
 
 // ListOperator 运营商列表
 func (c *model) ListOperator(ctx context.Context, req *dto.ListOperatorReq) (res *dto.ListOperatorRes, err error) {
-	return service.GetOperatorList()
+	return modelService.GetOperatorList()
 }
 
 // UpdateChatModel 更新是否为聊天模型
 func (c *model) UpdateChatModel(ctx context.Context, req *dto.UpdateChatModelReq) (res *dto.UpdateChatModelRes, err error) {
-	err = service.Model.UpdateChatModel(ctx, req)
+	err = modelService.ModelGatewayModels.UpdateChatModel(ctx, req)
 	return
 }
 
 // GetIsChatModel 获取当前会话模型
 func (c *model) GetIsChatModel(ctx context.Context, req *dto.GetIsChatModelReq) (res *dto.GetIsChatModelRes, err error) {
-	return service.Model.GetIsChatModel(ctx)
+	return modelService.ModelGatewayModels.GetIsChatModel(ctx)
 }

controller/model_gateway_task_controller.go

@@ -0,0 +1,43 @@
+package controller
+
+import (
+	"context"
+	taskService "model-gateway/service/task"
+
+	"model-gateway/model/dto"
+)
+
+// ModelGatewayTask 任务控制器
+var ModelGatewayTask = new(task)
+
+type task struct{}
+
+// CreateTask 根据 modelName 创建异步任务，返回 taskId
+func (c *task) CreateTask(ctx context.Context, req *dto.CreateTaskReq) (res *dto.CreateTaskRes, err error) {
+	return taskService.ModelGatewayTask.Create(ctx, req)
+}
+
+// GetTaskResult 获取单条任务结果（返回 *dto.GetTaskResultRes）
+func (c *task) GetTaskResult(ctx context.Context, req *dto.GetTaskResultReq) (res *dto.GetTaskResultRes, err error) {
+	return taskService.ModelGatewayTask.GetResult(ctx, req.TaskID)
+}
+
+// GetTaskBatch 批量查询任务（返回 *[]dto.GetTaskBatchItem）
+func (c *task) GetTaskBatch(ctx context.Context, req *dto.GetTaskBatchReq) (res *dto.GetTaskBatchRes, err error) {
+	return taskService.ModelGatewayTask.GetBatch(ctx, req)
+}
+
+// ListTask 任务列表分页查询
+func (c *task) ListTask(ctx context.Context, req *dto.ListTaskReq) (res *dto.ListTaskRes, err error) {
+	return taskService.ModelGatewayTask.List(ctx, req)
+}
+
+// ModelTaskCallback 接收模型异步任务的回调通知 —— 待调整
+func (c *task) ModelTaskCallback(ctx context.Context, req *dto.ModelTaskCallbackReq) (res *dto.ModelTaskCallbackRes, err error) {
+	return taskService.ModelGatewayTask.ModelTaskCallback(ctx, req)
+}
+
+// QueryPendingTasks 批量轮询进行中的异步任务 —— 待调整
+func (c *task) QueryPendingTasks(ctx context.Context, req *dto.QueryPendingTasksReq) (res *dto.QueryPendingTasksRes, err error) {
+	return taskService.ModelGatewayTask.QueryPendingTasks(ctx, req)
+}

controller/task_controller.go

@@ -1,43 +0,0 @@
-package controller
-
-import (
-	"context"
-
-	"model-gateway/model/dto"
-	"model-gateway/service"
-)
-
-type task struct{}
-
-// Task 任务控制器
-var Task = new(task)
-
-// CreateTask 根据 modelName 创建异步任务，返回 taskId
-func (c *task) CreateTask(ctx context.Context, req *dto.CreateTaskReq) (res *dto.CreateTaskRes, err error) {
-	return service.Task.Create(ctx, req)
-}
-
-// GetTaskResult 获取任务结果（只返回 oss 地址 + state）
-func (c *task) GetTaskResult(ctx context.Context, req *dto.GetTaskResultReq) (res *dto.GetTaskResultRes, err error) {
-	return service.Task.GetResult(ctx, req.TaskID)
-}
-
-// GetTaskBatch 批量查询任务（成功任务标记为已下载）
-func (c *task) GetTaskBatch(ctx context.Context, req *dto.GetTaskBatchReq) (res *dto.GetTaskBatchRes, err error) {
-	return service.Task.GetBatch(ctx, req)
-}
-
-// ListTask 任务列表分页查询
-func (c *task) ListTask(ctx context.Context, req *dto.ListTaskReq) (res *dto.ListTaskRes, err error) {
-	return service.Task.List(ctx, req)
-}
-
-// RunWork 手动触发一次 worker（由上层定时任务调用）
-func (c *task) RunWork(ctx context.Context, req *dto.RunWorkReq) (res *dto.RunWorkRes, err error) {
-	return service.AsyncWorker.RunOnce(ctx, req)
-}
-
-// CleanWork 手动触发一次 cleaner（由上层定时任务调用）
-func (c *task) CleanWork(ctx context.Context, req *dto.CleanWorkReq) (res *dto.CleanWorkRes, err error) {
-	return service.Cleaner.RunOnce(ctx)
-}

dao/model_dao.go

@@ -1,169 +0,0 @@
-package dao
-
-import (
-	"context"
-	"model-gateway/consts/public"
-	"model-gateway/model/dto"
-	"model-gateway/model/entity"
-
-	"gitea.com/red-future/common/db/gfdb"
-	"github.com/gogf/gf/v2/frame/g"
-	"github.com/gogf/gf/v2/util/gconv"
-)
-
-var Model = &modelDao{}
-
-type modelDao struct{}
-
-// Insert 插入
-func (d *modelDao) Insert(ctx context.Context, req *entity.AsynchModel) (id int64, err error) {
-	m := new(entity.AsynchModel)
-	err = gconv.Struct(req, &m)
-	if err != nil {
-		return
-	}
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
-		Insert(m)
-	if err != nil {
-		return
-	}
-	return r.LastInsertId()
-}
-
-// Update 更新
-func (d *modelDao) Update(ctx context.Context, req *entity.AsynchModel) (rows int64, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
-		OmitEmpty().
-		Data(&req).
-		Where(entity.AsynchModelCol.Id, req.Id).
-		Update()
-	if err != nil {
-		return
-	}
-	return r.RowsAffected()
-}
-
-// Delete 删除
-func (d *modelDao) Delete(ctx context.Context, req *entity.AsynchModel) (rows int64, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
-		OmitEmpty().
-		Where(entity.AsynchModelCol.Id, req.Id).
-		Delete()
-	if err != nil {
-		return
-	}
-	return r.RowsAffected()
-}
-
-// Get 按ID获取（带租户隔离，只查当前租户）
-func (d *modelDao) Get(ctx context.Context, req *entity.AsynchModel, fields ...string) (m *entity.AsynchModel, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
-		OmitEmpty().
-		Where(entity.AsynchModelCol.Id, req.Id).
-		Where(entity.AsynchModelCol.Creator, req.Creator).
-		Where(entity.AsynchModelCol.IsChatModel, req.IsChatModel).
-		Where(entity.AsynchModelCol.ModelName, req.ModelName).
-		Fields(fields).One()
-	if err != nil {
-		return
-	}
-	err = r.Struct(&m)
-	return
-}
-
-// GetByAcrossTenant 按ID获取（跨租户，查所有租户）
-func (d *modelDao) GetByAcrossTenant(ctx context.Context, req *entity.AsynchModel, fields ...string) (m *entity.AsynchModel, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
-		NoTenantId(ctx).
-		OmitEmpty().
-		Where(entity.AsynchModelCol.Id, req.Id).
-		Where(entity.AsynchModelCol.Creator, req.Creator).
-		Where(entity.AsynchModelCol.IsChatModel, req.IsChatModel).
-		Where(entity.AsynchModelCol.ModelName, req.ModelName).
-		Fields(fields).One()
-	if err != nil {
-		return
-	}
-	err = r.Struct(&m)
-	return
-}
-
-// GetByCreatorAndPlatform 按创建者、平台获取
-func (d *modelDao) GetByCreatorAndPlatform(ctx context.Context, req *dto.ListModelReq) (list []*entity.AsynchModel, total int, err error) {
-	// 基础 SQL
-	sql := `
-SELECT DISTINCT ON (model_name) *
-FROM asynch_models
-WHERE deleted_at IS NULL 
-  AND (? = '' OR model_name LIKE ?)
-  AND (? = 0 OR model_type = ?)
-`
-	args := []any{
-		req.ModelName, "%" + req.ModelName + "%",
-		req.ModelType, req.ModelType,
-	}
-	if !g.IsEmpty(req.IsPrivate) {
-		sql += ` AND is_private = ? `
-		args = append(args, req.IsPrivate)
-	}
-	if req.IsOwner != nil && *req.IsOwner == 0 {
-		if req.Enabled != nil && *req.Enabled == 1 {
-			sql += ` AND creator = ? AND is_owner = ? AND enabled=1 `
-		} else if req.Enabled != nil && *req.Enabled == 0 {
-			sql += ` AND creator = ? AND is_owner = ? AND enabled=0 `
-		} else {
-			sql += ` AND creator = ? AND is_owner = ? `
-		}
-
-		args = append(args, req.Creator)
-		args = append(args, req.IsOwner)
-	} else if req.IsOwner != nil && *req.IsOwner == 1 {
-		if req.Enabled != nil && *req.Enabled == 1 {
-			sql += ` AND ((creator = ? AND is_owner = ? AND enabled=1) OR (is_owner = 0 AND enabled=1)) `
-		} else if req.Enabled != nil && *req.Enabled == 0 {
-			sql += ` AND ((creator = ? AND is_owner = ? AND enabled=0) OR (is_owner = 0 AND enabled=1)) `
-		} else {
-			sql += ` AND ((creator = ? AND is_owner = ?) OR (is_owner = 0 AND enabled=1)) `
-		}
-		args = append(args, req.Creator)
-		args = append(args, req.IsOwner)
-	}
-
-	// 最后拼接排序
-	sql += ` ORDER BY model_name, is_owner DESC, created_at DESC`
-
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).GetAll(ctx, sql, args...)
-	if err != nil {
-		return nil, 0, err
-	}
-
-	err = r.Structs(&list)
-	if err != nil {
-		return nil, 0, err
-	}
-
-	total = len(list)
-	return
-}
-
-// GetByModelNameForTenant 后台任务使用：按 tenant_id + model_name 查询，不依赖 gfdb Hook/Trace/用户上下文
-func (d *modelDao) GetByModelNameForTenant(ctx context.Context, tenantId uint64, modelName string) (m *entity.AsynchModel, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).GetAll(ctx,
-		"SELECT * FROM "+public.TableNameModel+" WHERE tenant_id=? AND model_name=? AND deleted_at IS NULL LIMIT 1",
-		tenantId, modelName,
-	)
-	if err != nil {
-		return nil, err
-	}
-	if r.IsEmpty() {
-		return nil, nil
-	}
-	var list []*entity.AsynchModel
-	if err := r.Structs(&list); err != nil {
-		return nil, err
-	}
-	if len(list) == 0 {
-		return nil, nil
-	}
-	return list[0], nil
-}

dao/model_gateway_logs_op.go

@@ -0,0 +1,23 @@
+package dao
+
+import (
+	"context"
+
+	"model-gateway/consts/public"
+	"model-gateway/model/entity"
+
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
+)
+
+var ModelGatewayLogsOp = &modelGatewayLogsOpDao{}
+
+type modelGatewayLogsOpDao struct{}
+
+// Insert 插入操作日志
+func (d *modelGatewayLogsOpDao) Insert(ctx context.Context, req *entity.ModelGatewayLogsOp) (int64, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameOpLog).Insert(req)
+	if err != nil {
+		return 0, err
+	}
+	return r.LastInsertId()
+}

dao/model_gateway_logs_stat_dao.go

@@ -0,0 +1,52 @@
+package dao
+
+import (
+	"context"
+	"time"
+
+	"model-gateway/consts/public"
+	"model-gateway/model/entity"
+
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
+	"github.com/gogf/gf/v2/os/gtime"
+	"github.com/gogf/gf/v2/util/gconv"
+)
+
+var ModelGatewayLogsStat = &modelGatewayLogsStatDao{}
+
+type modelGatewayLogsStatDao struct{}
+
+// IncRequestCount 原子累加：按天+租户+创建人+模型 +1
+func (d *modelGatewayLogsStatDao) IncRequestCount(ctx context.Context, day time.Time, tenantId uint64, creator, modelName string) error {
+	_, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameStat).
+		Data(&entity.ModelGatewayLogsStat{
+			Day:          gtime.New(day),
+			TenantId:     tenantId,
+			Creator:      creator,
+			ModelName:    modelName,
+			RequestCount: 1,
+		}).
+		OnDuplicate("request_count", "request_count+1").
+		Insert()
+	return err
+}
+
+// List 分页查询统计
+func (d *modelGatewayLogsStatDao) List(ctx context.Context, pageNum, pageSize int, req *entity.ModelGatewayLogsStat) (list []*entity.ModelGatewayLogsStat, total int64, err error) {
+	model := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameStat).
+		OmitEmpty().
+		Where(entity.ModelGatewayLogsStatCols.Creator, req.Creator).
+		WhereLike(entity.ModelGatewayLogsStatCols.ModelName, "%"+req.ModelName+"%").
+		OrderDesc(entity.ModelGatewayLogsStatCols.Day).
+		OrderDesc(entity.ModelGatewayLogsStatCols.RequestCount)
+	if pageNum > 0 && pageSize > 0 {
+		model = model.Page(pageNum, pageSize)
+	}
+	r, totalInt, err := model.AllAndCount(false)
+	if err != nil {
+		return nil, 0, err
+	}
+	total = gconv.Int64(totalInt)
+	err = r.Structs(&list)
+	return
+}

dao/model_gateway_models_dao.go

@@ -0,0 +1,201 @@
+package dao
+
+import (
+	"context"
+	"model-gateway/consts/public"
+	"model-gateway/model/dto"
+	"model-gateway/model/entity"
+	"strconv"
+
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
+	"github.com/gogf/gf/v2/frame/g"
+)
+
+var ModelGatewayModels = &modelGatewayModelsDao{}
+
+type modelGatewayModelsDao struct{}
+
+// Insert 插入
+func (d *modelGatewayModelsDao) Insert(ctx context.Context, req *entity.ModelGatewayModel) (int64, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).Insert(req)
+	if err != nil {
+		return 0, err
+	}
+	return r.LastInsertId()
+}
+
+// Update 更新
+func (d *modelGatewayModelsDao) Update(ctx context.Context, req *entity.ModelGatewayModel) (int64, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
+		OmitEmpty().
+		Data(req).
+		Where(entity.ModelGatewayModelCol.Id, req.Id).
+		Update()
+	if err != nil {
+		return 0, err
+	}
+	return r.RowsAffected()
+}
+
+// Delete 删除
+func (d *modelGatewayModelsDao) Delete(ctx context.Context, req *entity.ModelGatewayModel) (int64, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
+		OmitEmpty().
+		Where(entity.ModelGatewayModelCol.Id, req.Id).
+		Delete()
+	if err != nil {
+		return 0, err
+	}
+	return r.RowsAffected()
+}
+
+// Get 获取模型
+func (d *modelGatewayModelsDao) Get(ctx context.Context, req *entity.ModelGatewayModel, fields ...string) (*entity.ModelGatewayModel, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
+		OmitEmpty().
+		Where(entity.ModelGatewayModelCol.Id, req.Id).
+		Where(entity.ModelGatewayModelCol.Creator, req.Creator).
+		Where(entity.ModelGatewayModelCol.ModelName, req.ModelName).
+		Fields(fields).One()
+	if err != nil {
+		return nil, err
+	}
+	var m entity.ModelGatewayModel
+	err = r.Struct(&m)
+	return &m, err
+}
+
+//// Get 按ID获取（带租户隔离，只查当前租户）
+//func (d *modelGatewayModelsDao) Get(ctx context.Context, req *entity.AsynchModel, fields ...string) (m *entity.AsynchModel, err error) {
+//	var whereCondition strings.Builder
+//	var queryParams []interface{}
+//	if !g.IsEmpty(req.Id) {
+//		whereCondition.WriteString(fmt.Sprintf(" AND %s = (?) ", entity.AsynchModelCol.Id))
+//		queryParams = append(queryParams, req.Id)
+//	}
+//	if !g.IsEmpty(req.Creator) {
+//		whereCondition.WriteString(fmt.Sprintf(" AND %s = (?) ", entity.AsynchModelCol.Creator))
+//		queryParams = append(queryParams, req.Creator)
+//	}
+//	if !g.IsEmpty(req.IsChatModel) {
+//		whereCondition.WriteString(fmt.Sprintf(" AND %s = (?) ", entity.AsynchModelCol.IsChatModel))
+//		queryParams = append(queryParams, req.IsChatModel)
+//	}
+//	if !g.IsEmpty(req.ModelName) {
+//		whereCondition.WriteString(fmt.Sprintf(" AND %s = (?) ", entity.AsynchModelCol.ModelName))
+//		queryParams = append(queryParams, req.ModelName)
+//	}
+//	// 完整 SQL
+//	sql := `SELECT * FROM "asynch_models" WHERE "deleted_at" IS NULL` + whereCondition.String()
+//	r, err := gfdb.DB(ctx, public.DbNameModelGateway).GetAll(ctx, sql, queryParams...)
+//	if err != nil {
+//		return
+//	}
+//	var i []*entity.AsynchModel
+//	if err = r.Structs(&i); err != nil {
+//		return nil, err
+//	}
+//	for _, item := range i {
+//		m = item
+//	}
+//	return
+//}
+
+// GetByAcrossTenant 跨租户查询
+func (d *modelGatewayModelsDao) GetByAcrossTenant(ctx context.Context, req *entity.ModelGatewayModel, fields ...string) (*entity.ModelGatewayModel, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameModel).
+		NoTenantId(ctx).
+		OmitEmpty().
+		Where(entity.ModelGatewayModelCol.Id, req.Id).
+		Where(entity.ModelGatewayModelCol.Creator, req.Creator).
+		Where(entity.ModelGatewayModelCol.ModelName, req.ModelName).
+		Fields(fields).One()
+	if err != nil {
+		return nil, err
+	}
+	var m entity.ModelGatewayModel
+	err = r.Struct(&m)
+	return &m, err
+}
+
+// GetByCreatorAndPlatform 按创建者、平台获取
+func (d *modelGatewayModelsDao) GetByCreatorAndPlatform(ctx context.Context, req *dto.ListModelReq) (list []*entity.ModelGatewayModel, total int, err error) {
+	sql := `
+SELECT DISTINCT ON (model_name) *
+FROM asynch_models
+WHERE deleted_at IS NULL 
+  AND (? = '' OR model_name LIKE ?)
+`
+	args := []any{
+		req.ModelName, "%" + req.ModelName + "%",
+	}
+
+	// modelType: 传 6 模糊匹配 6%
+	if req.ModelType > 0 {
+		prefix := strconv.Itoa(req.ModelType)[:1] // 截取第一位
+		sql += ` AND model_type::text LIKE ? `
+		args = append(args, prefix+"%")
+	}
+
+	if !g.IsEmpty(req.IsPrivate) {
+		sql += ` AND is_private = ? `
+		args = append(args, req.IsPrivate)
+	}
+
+	if req.IsOwner != nil && *req.IsOwner == 0 {
+		if req.Enabled != nil && *req.Enabled == 1 {
+			sql += ` AND creator = ? AND is_owner = ? AND enabled=1 `
+		} else if req.Enabled != nil && *req.Enabled == 0 {
+			sql += ` AND creator = ? AND is_owner = ? AND enabled=0 `
+		} else {
+			sql += ` AND creator = ? AND is_owner = ? `
+		}
+		args = append(args, req.Creator, req.IsOwner)
+	} else if req.IsOwner != nil && *req.IsOwner == 1 {
+		if req.Enabled != nil && *req.Enabled == 1 {
+			sql += ` AND ((creator = ? AND is_owner = ? AND enabled=1) OR (is_owner = 0 AND enabled=1)) `
+		} else if req.Enabled != nil && *req.Enabled == 0 {
+			sql += ` AND ((creator = ? AND is_owner = ? AND enabled=0) OR (is_owner = 0 AND enabled=1)) `
+		} else {
+			sql += ` AND ((creator = ? AND is_owner = ?) OR (is_owner = 0 AND enabled=1)) `
+		}
+		args = append(args, req.Creator, req.IsOwner)
+	}
+
+	sql += ` ORDER BY model_name, is_owner DESC, created_at DESC`
+
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).GetAll(ctx, sql, args...)
+	if err != nil {
+		return nil, 0, err
+	}
+
+	err = r.Structs(&list)
+	if err != nil {
+		return nil, 0, err
+	}
+
+	total = len(list)
+	return
+}
+
+// GetByModelNameForTenant 后台任务使用：按 tenant_id + model_name 查询，不依赖 gfdb Hook/Trace/用户上下文
+func (d *modelGatewayModelsDao) GetByModelNameForTenant(ctx context.Context, tenantId uint64, modelName string) (*entity.ModelGatewayModel, error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).GetAll(ctx,
+		"SELECT * FROM "+public.TableNameModel+" WHERE tenant_id=? AND model_name=? AND deleted_at IS NULL LIMIT 1",
+		tenantId, modelName,
+	)
+	if err != nil {
+		return nil, err
+	}
+	if r.IsEmpty() {
+		return nil, nil
+	}
+	var list []*entity.ModelGatewayModel
+	if err := r.Structs(&list); err != nil {
+		return nil, err
+	}
+	if len(list) == 0 {
+		return nil, nil
+	}
+	return list[0], nil
+}

dao/model_gateway_task_dao.go

@@ -0,0 +1,159 @@
+package dao
+
+import (
+	"context"
+	"fmt"
+	"model-gateway/consts/public"
+	"model-gateway/model/entity"
+
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
+	"github.com/gogf/gf/v2/database/gdb"
+	"github.com/gogf/gf/v2/util/gconv"
+)
+
+var ModelGatewayTask = &modelGatewayTaskDao{}
+
+type modelGatewayTaskDao struct{}
+
+// Insert 插入
+func (d *modelGatewayTaskDao) Insert(ctx context.Context, req *entity.ModelGatewayTask) (id int64, err error) {
+	m := new(entity.ModelGatewayTask)
+	err = gconv.Struct(req, &m)
+	if err != nil {
+		return
+	}
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).Insert(m)
+	if err != nil {
+		return
+	}
+	return r.LastInsertId()
+}
+
+// Update 更新（按ID）
+func (d *modelGatewayTaskDao) Update(ctx context.Context, req *entity.ModelGatewayTask) (rows int64, err error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		OmitEmpty().
+		Data(req).
+		Where(entity.ModelGatewayTaskCol.Id, req.Id).
+		Update()
+	if err != nil {
+		return
+	}
+	return r.RowsAffected()
+}
+
+// Get 获取（按TaskID 或 ID）
+func (d *modelGatewayTaskDao) Get(ctx context.Context, req *entity.ModelGatewayTask) (m *entity.ModelGatewayTask, err error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		OmitEmpty().
+		Where(entity.ModelGatewayTaskCol.TaskID, req.TaskID).
+		Where(entity.ModelGatewayTaskCol.Id, req.Id).
+		One()
+	if err != nil {
+		return
+	}
+	err = r.Struct(&m)
+	return
+}
+
+// List 分页查询
+func (d *modelGatewayTaskDao) List(ctx context.Context, pageNum, pageSize int, req *entity.ModelGatewayTask) (list []*entity.ModelGatewayTask, total int64, err error) {
+	model := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		OmitEmpty().
+		Where(entity.ModelGatewayTaskCol.Creator, req.Creator).
+		Where(entity.ModelGatewayTaskCol.ModelName, "%"+req.ModelName+"%").
+		Where(entity.ModelGatewayTaskCol.BizName, req.BizName).
+		Where(entity.ModelGatewayTaskCol.State, req.State).
+		Where(entity.ModelGatewayTaskCol.TaskID, req.TaskID).
+		OrderDesc(entity.ModelGatewayTaskCol.CreatedAt)
+	if pageNum > 0 && pageSize > 0 {
+		model = model.Page(pageNum, pageSize)
+	}
+	r, totalInt, err := model.AllAndCount(false)
+	if err != nil {
+		return nil, 0, err
+	}
+	total = gconv.Int64(totalInt)
+	err = r.Structs(&list)
+	return
+}
+
+// Delete 删除（软删，按ID）
+func (d *modelGatewayTaskDao) Delete(ctx context.Context, req *entity.ModelGatewayTask) (rows int64, err error) {
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		Where(entity.ModelGatewayTaskCol.Id, req.Id).
+		Delete()
+	if err != nil {
+		return
+	}
+	return r.RowsAffected()
+}
+
+// ListByTaskIDs 批量查询
+func (d *modelGatewayTaskDao) ListByTaskIDs(ctx context.Context, taskIDs []string) (list []*entity.ModelGatewayTask, err error) {
+	if len(taskIDs) == 0 {
+		return nil, nil
+	}
+	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		WhereIn(entity.ModelGatewayTaskCol.TaskID, taskIDs).
+		All()
+	if err != nil {
+		return nil, err
+	}
+	err = r.Structs(&list)
+	return
+}
+
+// MarkDownloadedByID 标记已下载
+func (d *modelGatewayTaskDao) MarkDownloadedByID(ctx context.Context, id int64) error {
+	_, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		Where(entity.ModelGatewayTaskCol.Id, id).
+		Where(entity.ModelGatewayTaskCol.State, 2).
+		Data(map[string]any{entity.ModelGatewayTaskCol.State: 4}).
+		Update()
+	return err
+}
+
+// GetPendingAsyncTasks 获取进行中的异步任务
+func (d *modelGatewayTaskDao) GetPendingAsyncTasks(ctx context.Context, limit int) ([]*entity.ModelGatewayTask, error) {
+	var tasks []*entity.ModelGatewayTask
+	err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
+		Where(entity.ModelGatewayTaskCol.State, 1).
+		Limit(limit).
+		Scan(&tasks)
+	return tasks, err
+}
+
+// ======================== 事务抢占 ========================
+
+// ClaimByID 按主键抢占，返回抢占后的任务
+func (d *modelGatewayTaskDao) ClaimByID(ctx context.Context, id int64) (*entity.ModelGatewayTask, error) {
+	var task entity.ModelGatewayTask
+	err := gfdb.DB(ctx, public.DbNameModelGateway).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
+		r, err := tx.Model(public.TableNameTask).
+			Where(entity.ModelGatewayTaskCol.Id, id).
+			Where(entity.ModelGatewayTaskCol.State, public.TaskStatusPending).
+			Limit(1).
+			LockUpdate().
+			One()
+		if err != nil {
+			return err
+		}
+		if r.IsEmpty() {
+			return fmt.Errorf("任务已被抢占或不存在: id=%d", id)
+		}
+		if err := r.Struct(&task); err != nil {
+			return err
+		}
+		_, err = tx.Model(public.TableNameTask).
+			Data(&entity.ModelGatewayTask{State: public.TaskStatusRunning}).
+			Where(entity.ModelGatewayTaskCol.Id, id).
+			OmitEmpty().
+			Update()
+		return err
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &task, nil
+}

dao/op_log_dao.go

@@ -1,30 +0,0 @@
-package dao
-
-import (
-	"context"
-
-	"model-gateway/consts/public"
-	"model-gateway/model/entity"
-
-	"gitea.com/red-future/common/db/gfdb"
-	"github.com/gogf/gf/v2/util/gconv"
-)
-
-type opLogDao struct{}
-
-var OpLog = &opLogDao{}
-
-// Insert 插入
-func (d *opLogDao) Insert(ctx context.Context, req *entity.LogsModelOp) (id int64, err error) {
-	m := new(entity.LogsModelOp)
-	err = gconv.Struct(req, &m)
-	if err != nil {
-		return
-	}
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameOpLog).
-		Insert(m)
-	if err != nil {
-		return 0, err
-	}
-	return r.LastInsertId()
-}

dao/stat_dao.go

@@ -1,60 +0,0 @@
-package dao
-
-import (
-	"context"
-	"fmt"
-	"time"
-
-	"model-gateway/consts/public"
-	"model-gateway/model/entity"
-
-	"gitea.com/red-future/common/db/gfdb"
-	"github.com/gogf/gf/v2/os/gtime"
-)
-
-type statDao struct{}
-
-var Stat = &statDao{}
-
-// IncRequestCount 原子累加（支持分布式/多协程）：按天+租户+创建人+模型 +1
-func (d *statDao) IncRequestCount(ctx context.Context, day time.Time, tenantId int64, creator, modelName string) error {
-	sql := fmt.Sprintf(`
-INSERT INTO %s(day, tenant_id, creator, model_name, request_count, created_at, updated_at)
-VALUES(?, ?, ?, ?, 1, NOW(), NOW())
-ON CONFLICT (day, tenant_id, creator, model_name)
-DO UPDATE SET request_count = %s.request_count + 1, updated_at = NOW()`,
-		public.TableNameStat, public.TableNameStat,
-	)
-	_, err := gfdb.DB(ctx, public.DbNameModelGateway).Exec(ctx, sql, gtime.New(day).Format("Y-m-d"), tenantId, creator, modelName)
-	return err
-}
-
-func (d *statDao) List(ctx context.Context, pageNum, pageSize int, startDay, endDay string, tenantId *int64, creator, modelName string) (list []*entity.LogsModelStat, total int64, err error) {
-	m := gfdb.DB(ctx).Model(ctx, public.TableNameStat).Where("1=1")
-	if startDay != "" {
-		m = m.Where("day >= ?", startDay)
-	}
-	if endDay != "" {
-		m = m.Where("day <= ?", endDay)
-	}
-	if tenantId != nil {
-		m = m.Where("tenant_id = ?", *tenantId)
-	}
-	if creator != "" {
-		m = m.WhereLike("creator", "%"+creator+"%")
-	}
-	if modelName != "" {
-		m = m.WhereLike("model_name", "%"+modelName+"%")
-	}
-	m = m.OrderDesc("day").OrderDesc("request_count")
-	if pageNum > 0 && pageSize > 0 {
-		m = m.Page(pageNum, pageSize)
-	}
-	r, totalInt, err := m.AllAndCount(false)
-	if err != nil {
-		return nil, 0, err
-	}
-	total = int64(totalInt)
-	err = r.Structs(&list)
-	return
-}

dao/task_dao.go

@@ -1,100 +0,0 @@
-package dao
-
-import (
-	"context"
-	"model-gateway/consts/public"
-	"model-gateway/model/entity"
-
-	"gitea.com/red-future/common/db/gfdb"
-	"github.com/gogf/gf/v2/database/gdb"
-	"github.com/gogf/gf/v2/os/gtime"
-	"github.com/gogf/gf/v2/util/gconv"
-)
-
-var Task = &taskDao{}
-
-type taskDao struct{}
-
-// Insert 插入
-func (d *taskDao) Insert(ctx context.Context, req *entity.AsynchTask) (id int64, err error) {
-	m := new(entity.AsynchTask)
-	err = gconv.Struct(req, &m)
-	if err != nil {
-		return
-	}
-	r, err := gfdb.DB(ctx).Model(ctx, public.TableNameTask).
-		Insert(m)
-	if err != nil {
-		return
-	}
-	return r.LastInsertId()
-}
-
-// Get 获取
-func (d *taskDao) Get(ctx context.Context, req *entity.AsynchTask, fields ...string) (m *entity.AsynchTask, err error) {
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
-		OmitEmpty().
-		Where(entity.AsynchTaskCol.TaskID, req.TaskID).
-		Fields(fields).One()
-	if err != nil {
-		return
-	}
-	err = r.Struct(&m)
-	return
-}
-
-// ListByTaskIDs 批量查询任务（会受 gfdb 的租户 Hook 影响，只返回当前租户数据）
-func (d *taskDao) ListByTaskIDs(ctx context.Context, taskIDs []string) (m []*entity.AsynchTask, err error) {
-	if len(taskIDs) == 0 {
-		return nil, nil
-	}
-	r, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
-		OmitEmpty().
-		WhereIn(entity.AsynchTaskCol.TaskID, taskIDs).
-		All()
-	if err != nil {
-		return nil, err
-	}
-	err = r.Structs(&m)
-	return
-}
-
-// MarkDownloadedByID 将成功任务标记为已下载(state=4)，并写入过期时间
-func (d *taskDao) MarkDownloadedByID(ctx context.Context, id int64, expireAt *gtime.Time) error {
-	data := gdb.Map{
-		entity.AsynchTaskCol.State:    4,
-		entity.AsynchTaskCol.ExpireAt: expireAt,
-		entity.AsynchTaskCol.Updater:  "",
-	}
-	_, err := gfdb.DB(ctx, public.DbNameModelGateway).Model(ctx, public.TableNameTask).
-		Where(entity.AsynchTaskCol.Id, id).
-		Where(entity.AsynchTaskCol.State, 2).
-		Data(data).
-		Update()
-	return err
-}
-
-// List 任务分页查询（受 gfdb 租户 Hook 影响）
-func (d *taskDao) List(ctx context.Context, pageNum, pageSize int, modelNameLike, taskIDLike string, state *int) (list []*entity.AsynchTask, total int64, err error) {
-	m := gfdb.DB(ctx).Model(ctx, public.TableNameTask).Where("deleted_at IS NULL")
-	if modelNameLike != "" {
-		m = m.WhereLike(entity.AsynchTaskCol.ModelName, "%"+modelNameLike+"%")
-	}
-	if taskIDLike != "" {
-		m = m.WhereLike(entity.AsynchTaskCol.TaskID, "%"+taskIDLike+"%")
-	}
-	if state != nil {
-		m = m.Where(entity.AsynchTaskCol.State, *state)
-	}
-	m = m.OrderDesc(entity.AsynchTaskCol.CreatedAt)
-	if pageNum > 0 && pageSize > 0 {
-		m = m.Page(pageNum, pageSize)
-	}
-	r, totalInt, err := m.AllAndCount(false)
-	if err != nil {
-		return nil, 0, err
-	}
-	total = gconv.Int64(totalInt)
-	err = r.Structs(&list)
-	return
-}

dao/task_dao_bg.go

@@ -1,278 +0,0 @@
-package dao
-
-import (
-	"context"
-	"fmt"
-	"time"
-
-	"model-gateway/consts/public"
-	"model-gateway/model/entity"
-
-	"gitea.com/red-future/common/db/gfdb"
-	"github.com/gogf/gf/v2/database/gdb"
-	"github.com/gogf/gf/v2/os/gtime"
-)
-
-// ClaimPendingGlobal 后台任务使用：全局抢占 pending 任务（不加 tenant 过滤）
-func (d *taskDao) ClaimPendingGlobal(ctx context.Context, batchSize int) (tasks []*entity.AsynchTask, err error) {
-	if batchSize <= 0 {
-		batchSize = 1
-	}
-	err = gfdb.DB(ctx).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
-		sql := fmt.Sprintf(
-			`SELECT id, tenant_id, creator, model_name, task_id, biz_name, callback_url, model_key, input_ref, request_payload, phase, tmp_file
-			   FROM %s
-			  WHERE deleted_at IS NULL AND state = 0
-			  ORDER BY enqueue_at ASC
-			  LIMIT %d
-			  FOR UPDATE SKIP LOCKED`,
-			public.TableNameTask,
-			batchSize,
-		)
-		r, err := tx.GetAll(sql)
-		if err != nil {
-			return err
-		}
-		if r.IsEmpty() {
-			tasks = nil
-			return nil
-		}
-		if err := r.Structs(&tasks); err != nil {
-			return err
-		}
-		now := time.Now()
-		for _, t := range tasks {
-			_, err = tx.Exec(
-				fmt.Sprintf(`UPDATE %s SET state=1, started_at=?, updated_at=? WHERE id=?`, public.TableNameTask),
-				now, now, t.Id,
-			)
-			if err != nil {
-				return err
-			}
-		}
-		return nil
-	})
-	return
-}
-
-// ClaimPendingByTaskIDGlobal 按 task_id 定向抢占单个 pending 任务（不加 tenant 过滤）
-// 用于 createTask 创建成功后立即异步尝试执行当前任务，避免只依赖后续 runWork 扫描队列。
-func (d *taskDao) ClaimPendingByTaskIDGlobal(ctx context.Context, taskID string) (task *entity.AsynchTask, err error) {
-	if taskID == "" {
-		return nil, nil
-	}
-	err = gfdb.DB(ctx).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
-		sql := fmt.Sprintf(
-			`SELECT id, tenant_id, creator, model_name, task_id, biz_name, callback_url, model_key, input_ref, request_payload, phase, tmp_file
-			   FROM %s
-			  WHERE deleted_at IS NULL AND state = 0 AND task_id = ?
-			  LIMIT 1
-			  FOR UPDATE SKIP LOCKED`,
-			public.TableNameTask,
-		)
-		r, err := tx.GetOne(sql, taskID)
-		if err != nil {
-			return err
-		}
-		if r.IsEmpty() {
-			task = nil
-			return nil
-		}
-		if err := r.Struct(&task); err != nil {
-			return err
-		}
-		now := time.Now()
-		_, err = tx.Exec(
-			fmt.Sprintf(`UPDATE %s SET state=1, started_at=?, updated_at=? WHERE id=?`, public.TableNameTask),
-			now, now, task.Id,
-		)
-		return err
-	})
-	return
-}
-
-func (d *taskDao) UpdateSuccessGlobal(ctx context.Context, id int64, ossFile, fileType, textResult string, fileSize int64, expireAt *gtime.Time, expendTokens int) error {
-	now := gtime.Now()
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s
-SET state=2,
-    oss_file=?,
-    file_type=?,
-    text_result=?,
-	expend_tokens=?,
-    file_size=?,
-    error_msg='',
-    finished_at=?,
-    duration_seconds=EXTRACT(EPOCH FROM (? - created_at))::BIGINT,
-    expire_at=NULL,
-    phase=0,
-    tmp_file='',
-    updated_at=?
-WHERE id=?`, public.TableNameTask),
-		ossFile, fileType, textResult, expendTokens, fileSize, now, now, now, id,
-	)
-	return err
-}
-
-func (d *taskDao) UpdateFailedGlobal(ctx context.Context, id int64, errorMsg string) error {
-	now := gtime.Now()
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s
-SET state=3,
-    error_msg=?,
-    finished_at=?,
-    duration_seconds=EXTRACT(EPOCH FROM (? - created_at))::BIGINT,
-    phase=0,
-    tmp_file='',
-    updated_at=?
-WHERE id=?`, public.TableNameTask),
-		errorMsg, now, now, now, id,
-	)
-	return err
-}
-
-// UpdateFailedKeepTmpGlobal OSS 上传失败：保留 phase/tmp_file，下一轮仅重试 OSS 上传
-func (d *taskDao) UpdateFailedKeepTmpGlobal(ctx context.Context, id int64, errorMsg string) error {
-	now := gtime.Now()
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s SET state=3, error_msg=?, finished_at=?, phase=1, updated_at=? WHERE id=?`, public.TableNameTask),
-		errorMsg, now, now, id,
-	)
-	return err
-}
-
-// UpdateTmpAfterModelGlobal 模型调用成功后，写入临时文件路径并标记 phase=1
-func (d *taskDao) UpdateTmpAfterModelGlobal(ctx context.Context, id int64, tmpFile string) error {
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s SET phase=1, tmp_file=?, updated_at=NOW() WHERE id=?`, public.TableNameTask),
-		tmpFile, id,
-	)
-	return err
-}
-
-func (d *taskDao) RollbackToPendingGlobal(ctx context.Context, id int64) error {
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s SET state=0, enqueue_at=NOW(), updated_at=NOW() WHERE id=? AND state=1`, public.TableNameTask),
-		id,
-	)
-	return err
-}
-
-// ListExpiredDownloadedGlobal 获取已下载(state=4)且过期的任务，用于清理
-func (d *taskDao) ListExpiredDownloadedGlobal(ctx context.Context, limit int) (list []*entity.AsynchTask, err error) {
-	if limit <= 0 {
-		limit = 200
-	}
-	r, err := gfdb.DB(ctx).GetAll(ctx,
-		fmt.Sprintf(`SELECT * FROM %s WHERE deleted_at IS NULL AND state=4 AND expire_at IS NOT NULL AND expire_at < ? LIMIT ?`, public.TableNameTask),
-		gtime.Now(), limit,
-	)
-	if err != nil {
-		return nil, err
-	}
-	err = r.Structs(&list)
-	return
-}
-
-// ListFailedRetryableGlobal 获取失败(state=3)且仍可重试的任务
-// retry_count 不含首次执行；retry_times 表示失败后最多再重试 N 次
-func (d *taskDao) ListFailedRetryableGlobal(ctx context.Context, limit int) (list []*entity.AsynchTask, err error) {
-	if limit <= 0 {
-		limit = 200
-	}
-	r, err := gfdb.DB(ctx).GetAll(ctx,
-		fmt.Sprintf(`
-SELECT t.*,
-       m.retry_queue_max_seconds AS retry_queue_max_seconds
-  FROM %s t
-  JOIN %s m
-    ON t.tenant_id = m.tenant_id
-   AND t.model_name = m.model_name
- WHERE t.deleted_at IS NULL
-   AND t.state = 3
-   AND t.retry_count < m.retry_times
- ORDER BY t.updated_at ASC
- LIMIT ?`, public.TableNameTask, public.TableNameModel),
-		limit,
-	)
-	if err != nil {
-		return nil, err
-	}
-	err = r.Structs(&list)
-	return
-}
-
-// RequeueForRetryGlobal 将任务重新入队（state=0），并将 retry_count +1
-// enqueueAt 用于控制重试任务在队列中的位置：
-// - enqueueAt 越早，越靠前（ClaimPendingGlobal 按 enqueue_at ASC 抢占）
-func (d *taskDao) RequeueForRetryGlobal(ctx context.Context, id int64, enqueueAt time.Time) error {
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s SET state=0, retry_count=retry_count+1, enqueue_at=?, updated_at=NOW() WHERE id=? AND state=3 AND deleted_at IS NULL`, public.TableNameTask),
-		enqueueAt, id,
-	)
-	return err
-}
-
-// ListFailedExhaustedGlobal 获取失败(state=3)且超过重试次数的任务，用于硬删除
-func (d *taskDao) ListFailedExhaustedGlobal(ctx context.Context, limit int) (list []*entity.AsynchTask, err error) {
-	if limit <= 0 {
-		limit = 200
-	}
-	r, err := gfdb.DB(ctx).GetAll(ctx,
-		fmt.Sprintf(`
-SELECT t.*
-  FROM %s t
-  JOIN %s m
-    ON t.tenant_id = m.tenant_id
-   AND t.model_name = m.model_name
- WHERE t.deleted_at IS NULL
-   AND t.state = 3
-   AND t.retry_count >= m.retry_times
- ORDER BY t.updated_at ASC
- LIMIT ?`, public.TableNameTask, public.TableNameModel),
-		limit,
-	)
-	if err != nil {
-		return nil, err
-	}
-	err = r.Structs(&list)
-	return
-}
-
-// HardDeleteByIDGlobal 硬删除任务记录
-func (d *taskDao) HardDeleteByIDGlobal(ctx context.Context, id int64) error {
-	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`DELETE FROM %s WHERE id=?`, public.TableNameTask),
-		id,
-	)
-	return err
-}
-
-// ListTimeoutTasksGlobal 根据模型配置 expected_seconds 判定超时任务：
-// - state in (0,1)
-// - 模型 expected_seconds > 0
-// - now - created_at >= expected_seconds
-func (d *taskDao) ListTimeoutTasksGlobal(ctx context.Context, limit int) (list []*entity.AsynchTask, err error) {
-	if limit <= 0 {
-		limit = 200
-	}
-	r, err := gfdb.DB(ctx).GetAll(ctx,
-		fmt.Sprintf(`
-SELECT t.*
-  FROM %s t
-  JOIN %s m
-    ON t.tenant_id = m.tenant_id
-   AND t.model_name = m.model_name
- WHERE t.deleted_at IS NULL
-   AND t.state IN (0,1)
-   AND m.expected_seconds > 0
-   AND t.created_at < (NOW() - (m.expected_seconds || ' seconds')::interval)
- LIMIT ?`, public.TableNameTask, public.TableNameModel),
-		limit,
-	)
-	if err != nil {
-		return nil, err
-	}
-	err = r.Structs(&list)
-	return
-}

go.mod

@@ -1,22 +1,14 @@
 module model-gateway
 
-go 1.26.0
+go 1.26.1
 
 require (
-	gitea.com/red-future/common v0.0.19
-	github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.0
-	github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.0
-	github.com/gogf/gf/v2 v2.10.0
+	gitea.redpowerfuture.com/red-future/common v0.0.23
+	github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.2
+	github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.2
+	github.com/gogf/gf/v2 v2.10.2
 	github.com/google/uuid v1.6.0
-	github.com/tidwall/gjson v1.14.2
-)
-
-require (
-	github.com/r3labs/diff/v2 v2.15.1 // indirect
-	github.com/tidwall/match v1.1.1 // indirect
-	github.com/tidwall/pretty v1.2.0 // indirect
-	github.com/vmihailenco/msgpack v4.0.4+incompatible // indirect
-	google.golang.org/appengine v1.6.7 // indirect
+	github.com/tidwall/gjson v1.19.0
 )
 
 require (
@@ -57,7 +49,7 @@ require (
 	github.com/hashicorp/go-rootcerts v1.0.2 // indirect
 	github.com/hashicorp/golang-lru v1.0.2 // indirect
 	github.com/hashicorp/serf v0.10.1 // indirect
-	github.com/klauspost/compress v1.18.2 // indirect
+	github.com/klauspost/compress v1.18.0 // indirect
 	github.com/lib/pq v1.10.9 // indirect
 	github.com/magiconair/properties v1.8.10 // indirect
 	github.com/mattn/go-colorable v0.1.14 // indirect
@@ -69,11 +61,14 @@ require (
 	github.com/olekukonko/ll v0.0.9 // indirect
 	github.com/olekukonko/tablewriter v1.1.0 // indirect
 	github.com/pkg/errors v0.9.1 // indirect
+	github.com/r3labs/diff/v2 v2.15.1 // indirect
 	github.com/redis/go-redis/v9 v9.12.1 // indirect
 	github.com/rivo/uniseg v0.4.7 // indirect
-	github.com/tidwall/sjson v1.2.5
+	github.com/tidwall/match v1.1.1 // indirect
+	github.com/tidwall/pretty v1.2.0 // indirect
 	github.com/tiger1103/gfast-token v1.0.10 // indirect
 	github.com/vcaesar/cedar v0.30.0 // indirect
+	github.com/vmihailenco/msgpack v4.0.4+incompatible // indirect
 	go.mongodb.org/mongo-driver/v2 v2.4.0 // indirect
 	go.opencensus.io v0.23.0 // indirect
 	go.opentelemetry.io/auto/sdk v1.1.0 // indirect
@@ -85,9 +80,10 @@ require (
 	go.opentelemetry.io/otel/trace v1.38.0 // indirect
 	go.opentelemetry.io/proto/otlp v1.7.1 // indirect
 	golang.org/x/exp v0.0.0-20250305212735-054e65f0b394 // indirect
-	golang.org/x/net v0.48.0 // indirect
-	golang.org/x/sys v0.39.0 // indirect
-	golang.org/x/text v0.32.0 // indirect
+	golang.org/x/net v0.47.0 // indirect
+	golang.org/x/sys v0.38.0 // indirect
+	golang.org/x/text v0.31.0 // indirect
+	google.golang.org/appengine v1.6.7 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20250825161204-c5933d9347a5 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20250825161204-c5933d9347a5 // indirect
 	google.golang.org/grpc v1.75.0 // indirect

go.sum

@@ -1,6 +1,6 @@
 cloud.google.com/go v0.26.0/go.mod h1:aQUYkXzVsufM+DwF1aE+0xfcU+56JwCaLick0ClmMTw=
-gitea.com/red-future/common v0.0.19 h1:9/WrfCFUCeFUYwuhBYF+JOQi5F5xuOy+gVnf2ZvHZu4=
-gitea.com/red-future/common v0.0.19/go.mod h1:6/nqIucVzmjOyqDTIq71feYBXXFNBy0rFwzaQ0/Ueoo=
+gitea.redpowerfuture.com/red-future/common v0.0.23 h1:xieoA00iKOCDm5SO9iXn+cSyMKBAlZwI0fuEVPWrHLg=
+gitea.redpowerfuture.com/red-future/common v0.0.23/go.mod h1:50U1Xi+Ie56z09S5LQbZvaken0Mxv3OeS9LgR7U/ZRY=
 github.com/BurntSushi/toml v0.3.1/go.mod h1:xHWCNGjB5oqiDr8zfno3MHue2Ht5sIBksp03qcyfWMU=
 github.com/BurntSushi/toml v1.5.0 h1:W5quZX/G/csjUnuI8SUYlsHs9M38FC7znL0lIO+DvMg=
 github.com/BurntSushi/toml v1.5.0/go.mod h1:ukJfTF/6rtPPRCnwkur4qwRxa8vTRFBF0uk2lLoLwho=
@@ -77,16 +77,16 @@ github.com/go-logr/logr v1.4.3/go.mod h1:9T104GzyrTigFIr8wt5mBrctHMim0Nb2HLGrmQ4
 github.com/go-logr/stdr v1.2.2 h1:hSWxHoqTgW2S2qGc0LTAI563KZ5YKYRhT3MFKZMbjag=
 github.com/go-logr/stdr v1.2.2/go.mod h1:mMo/vtBO5dYbehREoey6XUKy/eSumjCCveDpRre4VKE=
 github.com/go-stack/stack v1.8.0/go.mod h1:v0f6uXyyMGvRgIKkXu+yp6POWl0qKG85gN/melR3HDY=
-github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.0 h1:39+jbTenm7KBj4hO2C8ANAxVHpX/7OuRDs1VcGC9ylA=
-github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.0/go.mod h1:B0s0fVzn0W220E8UTpSGzrrGKsop5KcB90twBeLCiz0=
-github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.0 h1:N/F9CuDdUZLoM1nVRqrDE/33pDZuhVxpNY4wYdeIaBs=
-github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.0/go.mod h1:x6uoJGfZOtirIRQls8xUlYzC6f7T/eULPUa9er368X0=
+github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.2 h1:u8EpP24GkprogROnJ7htMov9Fc66pTP1eVYrWxiCYOs=
+github.com/gogf/gf/contrib/drivers/pgsql/v2 v2.10.2/go.mod h1:GmvM3r8GVByVMi4RD2+MCs5+CfxVXPMeT8mVDkAaAXE=
+github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.2 h1:iTQegT+lEg/wDKvj2mi3W1wrdrwFarjokf88EXVVgu4=
+github.com/gogf/gf/contrib/nosql/redis/v2 v2.10.2/go.mod h1:ZRw3GNz5cq4uYrW4TPSVyrYWaoqzujKdWro/AOcGBaE=
 github.com/gogf/gf/contrib/registry/consul/v2 v2.9.5 h1:eUqwJ/qNH8lJ6yssiqskazgp1ACQuNU6zXlLOZVuXTQ=
 github.com/gogf/gf/contrib/registry/consul/v2 v2.9.5/go.mod h1:sjQyMry9+0POYZCA6lHXBxO77WoNKkruJpRB4xKqk5k=
 github.com/gogf/gf/contrib/trace/otlphttp/v2 v2.9.5 h1:tHUEZYB5GTqEYYVDYnlGobf1xISARKDE4KHVlgjwTec=
 github.com/gogf/gf/contrib/trace/otlphttp/v2 v2.9.5/go.mod h1:cfzTn2HS9RDX8f5pUVkbGxUWcSosouqfNQ1G6cY0V88=
-github.com/gogf/gf/v2 v2.10.0 h1:rzDROlyqGMe/eM6dCalSR8dZOuMIdLhmxKSH1DGhbFs=
-github.com/gogf/gf/v2 v2.10.0/go.mod h1:Svl1N+E8G/QshU2DUbh/3J/AJauqCgUnxHurXWR4Qx0=
+github.com/gogf/gf/v2 v2.10.2 h1:46IO0Uc8e85/FqdftJFskfDejJLBL0JBnGS5qOftUu8=
+github.com/gogf/gf/v2 v2.10.2/go.mod h1:Svl1N+E8G/QshU2DUbh/3J/AJauqCgUnxHurXWR4Qx0=
 github.com/gogo/protobuf v1.1.1/go.mod h1:r8qH/GZQm5c6nD/R0oafs1akxWv10x8SbQlK7atdtwQ=
 github.com/gogo/protobuf v1.3.2 h1:Ov1cvc58UF3b5XjBnZv7+opcTcQFZebYjWzi34vdm4Q=
 github.com/gogo/protobuf v1.3.2/go.mod h1:P1XiOD3dCwIKUDQYPy72D8LYyHL2YPYrpS2s69NZV8Q=
@@ -185,8 +185,8 @@ github.com/json-iterator/go v1.1.9/go.mod h1:KdQUCv79m/52Kvf8AW2vK1V8akMuk1QjK/u
 github.com/julienschmidt/httprouter v1.2.0/go.mod h1:SYymIcj16QtmaHHD7aYtjjsJG7VTCxuUUipMqKk8s4w=
 github.com/kisielk/errcheck v1.5.0/go.mod h1:pFxgyoBC7bSaBwPgfKdkLd5X25qrDl4LWUI2bnpBCr8=
 github.com/kisielk/gotool v1.0.0/go.mod h1:XhKaO+MFFWcvkIS/tQcRk01m1F5IRFswLeQ+oQHNcck=
-github.com/klauspost/compress v1.18.2 h1:iiPHWW0YrcFgpBYhsA6D1+fqHssJscY/Tm/y2Uqnapk=
-github.com/klauspost/compress v1.18.2/go.mod h1:R0h/fSBs8DE4ENlcrlib3PsXS61voFxhIs2DeRhCvJ4=
+github.com/klauspost/compress v1.18.0 h1:c/Cqfb0r+Yi+JtIEq73FWXVkRonBlf0CRNYc8Zttxdo=
+github.com/klauspost/compress v1.18.0/go.mod h1:2Pp+KzxcywXVXMr50+X0Q/Lsb43OQHYWRCY2AiWywWQ=
 github.com/konsorten/go-windows-terminal-sequences v1.0.1/go.mod h1:T0+1ngSBFLxvqU3pZ+m/2kptfBszLMUkC4ZK/EgS/cQ=
 github.com/kr/logfmt v0.0.0-20140226030751-b84e30acd515/go.mod h1:+0opPa2QZZtGFBFZlji/RkVcI2GknAs/DXo4wKdlNEc=
 github.com/kr/pretty v0.1.0/go.mod h1:dAy3ld7l9f0ibDNOQOHHMYYIIbhfbHSm3C4ZsoJORNo=
@@ -288,14 +288,12 @@ github.com/stretchr/testify v1.6.1/go.mod h1:6Fq8oRcR53rry900zMqJjRRixrwX3KX962/
 github.com/stretchr/testify v1.7.2/go.mod h1:R6va5+xMeoiuVRoj+gSkQ7d3FALtqAAGI1FQKckRals=
 github.com/stretchr/testify v1.11.1 h1:7s2iGBzp5EwR7/aIZr8ao5+dra3wiQyKjjFuvgVKu7U=
 github.com/stretchr/testify v1.11.1/go.mod h1:wZwfW3scLgRK+23gO65QZefKpKQRnfz6sD981Nm4B6U=
-github.com/tidwall/gjson v1.14.2 h1:6BBkirS0rAHjumnjHF6qgy5d2YAJ1TLIaFE2lzfOLqo=
-github.com/tidwall/gjson v1.14.2/go.mod h1:/wbyibRr2FHMks5tjHJ5F8dMZh3AcwJEMf5vlfC0lxk=
+github.com/tidwall/gjson v1.19.0 h1:xwxm7n691Uf3u5OFjzngavjGTh55KX5q/9w9xHW88JU=
+github.com/tidwall/gjson v1.19.0/go.mod h1:V37/opeE/JbLUOfH0QTXiNez2l0RUjYUhpT4szFQAfc=
 github.com/tidwall/match v1.1.1 h1:+Ho715JplO36QYgwN9PGYNhgZvoUSc9X2c80KVTi+GA=
 github.com/tidwall/match v1.1.1/go.mod h1:eRSPERbgtNPcGhD8UCthc6PmLEQXEWd3PRB5JTxsfmM=
 github.com/tidwall/pretty v1.2.0 h1:RWIZEg2iJ8/g6fDDYzMpobmaoGh5OLl4AXtGUGPcqCs=
 github.com/tidwall/pretty v1.2.0/go.mod h1:ITEVvHYasfjBbM0u2Pg8T2nJnzm8xPwvNhhsoaGGjNU=
-github.com/tidwall/sjson v1.2.5 h1:kLy8mja+1c9jlljvWTlSazM7cKDRfJuR/bOJhcY5NcY=
-github.com/tidwall/sjson v1.2.5/go.mod h1:Fvgq9kS/6ociJEDnK0Fk1cpYF4FIW6ZF7LAe+6jwd28=
 github.com/tiger1103/gfast-token v1.0.10 h1:fNiBE/Dq5iTHvTGlCx3DmXa2o4hr0NtumFpffZ39k6s=
 github.com/tiger1103/gfast-token v1.0.10/go.mod h1:a/21mxmj7zFeNvjhZSC0XpEAFHfb1aT2k6DXnufFU1s=
 github.com/tv42/httpunix v0.0.0-20150427012821-b75d8614f926/go.mod h1:9ESjWnEqriFuLhtthL60Sar/7RFoluCcXsuvEwTV5KM=
@@ -344,8 +342,8 @@ golang.org/x/lint v0.0.0-20190227174305-5b3e6a55c961/go.mod h1:wehouNa3lNwaWXcvx
 golang.org/x/lint v0.0.0-20190313153728-d0100b6bd8b3/go.mod h1:6SW0HCj/g11FgYtHlgUYUwCkIfeOF89ocIRzGO/8vkc=
 golang.org/x/mod v0.2.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
 golang.org/x/mod v0.3.0/go.mod h1:s0Qsj1ACt9ePp/hMypM3fl4fZqREWJwdYDEqhRiZZUA=
-golang.org/x/mod v0.30.0 h1:fDEXFVZ/fmCKProc/yAXXUijritrDzahmwwefnjoPFk=
-golang.org/x/mod v0.30.0/go.mod h1:lAsf5O2EvJeSFMiBxXDki7sCgAxEUcZHXoXMKT4GJKc=
+golang.org/x/mod v0.29.0 h1:HV8lRxZC4l2cr3Zq1LvtOsi/ThTgWnUk/y64QSs8GwA=
+golang.org/x/mod v0.29.0/go.mod h1:NyhrlYXJ2H4eJiRy/WDBO6HMqZQ6q9nk4JzS3NuCK+w=
 golang.org/x/net v0.0.0-20180724234803-3673e40ba225/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20180826012351-8a410e7b638d/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
 golang.org/x/net v0.0.0-20181114220301-adae6a3d119a/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
@@ -361,8 +359,8 @@ golang.org/x/net v0.0.0-20201021035429-f5854403a974/go.mod h1:sp8m0HH+o8qH0wwXwY
 golang.org/x/net v0.0.0-20201110031124-69a78807bb2b/go.mod h1:sp8m0HH+o8qH0wwXwYZr8TS3Oi6o0r6Gce1SSxlDquU=
 golang.org/x/net v0.0.0-20210226172049-e18ecbb05110/go.mod h1:m0MpNAwzfU5UDzcl9v0D8zg8gWTRqZa9RBIspLL5mdg=
 golang.org/x/net v0.0.0-20210410081132-afb366fc7cd1/go.mod h1:9tjilg8BloeKEkVJvy7fQ90B1CfIiPueXVOjqfkSzI8=
-golang.org/x/net v0.48.0 h1:zyQRTTrjc33Lhh0fBgT/H3oZq9WuvRR5gPC70xpDiQU=
-golang.org/x/net v0.48.0/go.mod h1:+ndRgGjkh8FGtu1w1FGbEC31if4VrNVMuKTgcAAnQRY=
+golang.org/x/net v0.47.0 h1:Mx+4dIFzqraBXUugkia1OOvlD6LemFo1ALMHjrXDOhY=
+golang.org/x/net v0.47.0/go.mod h1:/jNxtkgq5yWUGYkaZGqo27cfGZ1c5Nen03aYrrKpVRU=
 golang.org/x/oauth2 v0.0.0-20180821212333-d2e6202438be/go.mod h1:N/0e6XlmueqKjAGxoOufVs8QHGRruUQn6yWY3a++T0U=
 golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20181108010431-42b317875d0f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
@@ -371,8 +369,8 @@ golang.org/x/sync v0.0.0-20190423024810-112230192c58/go.mod h1:RxMgew5VJxzue5/jJ
 golang.org/x/sync v0.0.0-20190911185100-cd5d95a43a6e/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20201020160332-67f06af15bc9/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
 golang.org/x/sync v0.0.0-20210220032951-036812b2e83c/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
-golang.org/x/sync v0.19.0 h1:vV+1eWNmZ5geRlYjzm2adRgW2/mcpevXNg50YZtPCE4=
-golang.org/x/sync v0.19.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
+golang.org/x/sync v0.18.0 h1:kr88TuHDroi+UVf+0hZnirlk8o8T+4MrK6mr60WkH/I=
+golang.org/x/sync v0.18.0/go.mod h1:9KTHXmSnoGruLpwFjVSX0lNNA75CykiMECbovNTZqGI=
 golang.org/x/sys v0.0.0-20180823144017-11551d06cbcc/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180830151530-49385e6e1522/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
 golang.org/x/sys v0.0.0-20180905080454-ebe1bf3edb33/go.mod h1:STP8DvDyc/dI5b8T5hshtkjS+E42TnysNCUPdjciGhY=
@@ -397,15 +395,15 @@ golang.org/x/sys v0.0.0-20220503163025-988cb79eb6c6/go.mod h1:oPkhp1MJrh7nUepCBc
 golang.org/x/sys v0.0.0-20220728004956-3c1f35247d10/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.0.0-20221010170243-090e33056c14/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
 golang.org/x/sys v0.6.0/go.mod h1:oPkhp1MJrh7nUepCBck5+mAzfO9JrbApNNgaTdGDITg=
-golang.org/x/sys v0.39.0 h1:CvCKL8MeisomCi6qNZ+wbb0DN9E5AATixKsvNtMoMFk=
-golang.org/x/sys v0.39.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
+golang.org/x/sys v0.38.0 h1:3yZWxaJjBmCWXqhN1qh02AkOnCQ1poK6oF+a7xWL6Gc=
+golang.org/x/sys v0.38.0/go.mod h1:OgkHotnGiDImocRcuBABYBEXf8A9a87e/uXjp9XT3ks=
 golang.org/x/term v0.0.0-20201126162022-7de9c90e9dd1/go.mod h1:bj7SfCRtBDWHUb9snDiAeCFNEtKQo2Wmx5Cou7ajbmo=
 golang.org/x/text v0.3.0/go.mod h1:NqM8EUOU14njkJ3fqMW+pc6Ldnwhi/IjpwHt7yyuwOQ=
 golang.org/x/text v0.3.2/go.mod h1:bEr9sfX3Q8Zfm5fL9x+3itogRgK3+ptLWKqgva+5dAk=
 golang.org/x/text v0.3.3/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
 golang.org/x/text v0.3.6/go.mod h1:5Zoc/QRtKVWzQhOtBMvqHzDpF6irO9z98xDceosuGiQ=
-golang.org/x/text v0.32.0 h1:ZD01bjUt1FQ9WJ0ClOL5vxgxOI/sVCNgX1YtKwcY0mU=
-golang.org/x/text v0.32.0/go.mod h1:o/rUWzghvpD5TXrTIBuJU77MTaN0ljMWE47kxGJQ7jY=
+golang.org/x/text v0.31.0 h1:aC8ghyu4JhP8VojJ2lEHBnochRno1sgL6nEi9WGFGMM=
+golang.org/x/text v0.31.0/go.mod h1:tKRAlv61yKIjGGHX/4tP1LTbc13YSec1pxVEWXzfoeM=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190114222345-bf090417da8b/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/tools v0.0.0-20190226205152-f727befe758c/go.mod h1:9Yl7xja0Znq3iFh3HoIrodX9oNMXvdceNzlUR8zjMvY=
@@ -415,8 +413,8 @@ golang.org/x/tools v0.0.0-20190907020128-2ca718005c18/go.mod h1:b+2E5dAYhXwXZwtn
 golang.org/x/tools v0.0.0-20191119224855-298f0cb1881e/go.mod h1:b+2E5dAYhXwXZwtnZ6UAqBI28+e2cm9otk0dWdXHAEo=
 golang.org/x/tools v0.0.0-20200619180055-7c47624df98f/go.mod h1:EkVYQZoAsY45+roYkvgYkIh4xh/qjgUK9TdY2XT94GE=
 golang.org/x/tools v0.0.0-20210106214847-113979e3529a/go.mod h1:emZCQorbCU4vsT4fOWvOPXz4eW1wZW4PmDk9uLelYpA=
-golang.org/x/tools v0.39.0 h1:ik4ho21kwuQln40uelmciQPp9SipgNDdrafrYA4TmQQ=
-golang.org/x/tools v0.39.0/go.mod h1:JnefbkDPyD8UU2kI5fuf8ZX4/yUeh9W877ZeBONxUqQ=
+golang.org/x/tools v0.38.0 h1:Hx2Xv8hISq8Lm16jvBZ2VQf+RLmbd7wVUsALibYI/IQ=
+golang.org/x/tools v0.38.0/go.mod h1:yEsQ/d/YK8cjh0L6rZlY8tgtlKiBNTL14pGDJPJpYQs=
 golang.org/x/xerrors v0.0.0-20190717185122-a985d3407aa7/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191011141410-1b5146add898/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
@@ -426,6 +424,7 @@ gonum.org/v1/gonum v0.16.0/go.mod h1:fef3am4MQ93R2HHpKnLk4/Tbh/s0+wqD5nfa6Pnwy4E
 google.golang.org/appengine v1.1.0/go.mod h1:EbEs0AVv82hx2wNQdGPgUI5lhzA/G0D9YwlJXL52JkM=
 google.golang.org/appengine v1.4.0/go.mod h1:xpcJRLb0r/rnEns0DIKYYv+WjYCduHsrkT7/EB5XEv4=
 google.golang.org/appengine v1.6.6/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
+google.golang.org/appengine v1.6.7 h1:FZR1q0exgwxzPzp/aF+VccGrSfxfPpkBqjIIEq3ru6c=
 google.golang.org/appengine v1.6.7/go.mod h1:8WjMMxjGQR8xUklV/ARdw2HLXBOI7O7uCIDZVag1xfc=
 google.golang.org/genproto v0.0.0-20180817151627-c66870c02cf8/go.mod h1:JiN7NxoALGmiZfu7CAH4rXhgtRTLTxftemlI0sWmxmc=
 google.golang.org/genproto v0.0.0-20190819201941-24fa4b261c55/go.mod h1:DMBHOl98Agz4BDEuKkezgsaosCRResVns1a3J2ZsMNc=

main.go

@@ -3,17 +3,17 @@ package main
 import (
 	"context"
 	"model-gateway/model/dto"
+	"model-gateway/service/task"
 	"os"
 	"os/signal"
 	"syscall"
 	"time"
 
 	"model-gateway/controller"
-	"model-gateway/service"
 
-	"gitea.com/red-future/common/http"
-	"gitea.com/red-future/common/jaeger"
-	_ "gitea.com/red-future/common/swagger"
+	"gitea.redpowerfuture.com/red-future/common/http"
+	"gitea.redpowerfuture.com/red-future/common/jaeger"
+	_ "gitea.redpowerfuture.com/red-future/common/swagger"
 	_ "github.com/gogf/gf/contrib/drivers/pgsql/v2"
 	_ "github.com/gogf/gf/contrib/nosql/redis/v2"
 	"github.com/gogf/gf/v2/frame/g"
@@ -26,9 +26,9 @@ func main() {
 
 	// 注册路由
 	http.RouteRegister([]interface{}{
-		controller.Model,
-		controller.Task,
-		controller.Stat,
+		controller.ModelGatewayModels,
+		controller.ModelGatewayTask,
+		controller.ModelGatewayLogsStat,
 	})
 
 	// 本地调试：可选自动触发 worker/cleaner（由配置文件控制）
@@ -46,14 +46,10 @@ func main() {
 }
 
 func startAutoRunner(ctx context.Context) {
-	// worker
-	if g.Cfg().MustGet(ctx, "asynch.worker.enabled").Bool() {
-		interval := g.Cfg().MustGet(ctx, "asynch.worker.intervalSeconds").Int()
-		if interval <= 0 {
-			interval = 5
-		}
-		batchSize := g.Cfg().MustGet(ctx, "asynch.worker.batchSize").Int()
-		goroutines := g.Cfg().MustGet(ctx, "asynch.worker.goroutines").Int()
+	// queryPending
+	if g.Cfg().MustGet(ctx, "asynch.queryPending.enabled").Bool() {
+		interval := g.Cfg().MustGet(ctx, "asynch.queryPending.intervalSeconds", 10).Int()
+		limit := g.Cfg().MustGet(ctx, "asynch.queryPending.limit", 10).Int()
 		ticker := time.NewTicker(time.Duration(interval) * time.Second)
 		go func() {
 			defer ticker.Stop()
@@ -62,34 +58,11 @@ func startAutoRunner(ctx context.Context) {
 				case <-ctx.Done():
 					return
 				case <-ticker.C:
-					if _, err := service.AsyncWorker.RunOnce(ctx, &dto.RunWorkReq{
-						BatchSize:  batchSize,
-						Goroutines: goroutines,
-					}); err != nil {
-						g.Log().Warningf(ctx, "[auto-worker] run once failed: %v", err)
+					if _, err := task.ModelGatewayTask.QueryPendingTasks(ctx, &dto.QueryPendingTasksReq{Limit: limit}); err != nil {
+						g.Log().Warningf(ctx, "[auto-queryPending] run once failed: %v", err)
 					}
 				}
 			}
 		}()
 	}
-
-	// cleaner
-	if g.Cfg().MustGet(ctx, "asynch.cleaner.enabled").Bool() {
-		interval := g.Cfg().MustGet(ctx, "asynch.cleaner.intervalSeconds").Int()
-		if interval <= 0 {
-			interval = 30
-		}
-		ticker := time.NewTicker(time.Duration(interval) * time.Second)
-		go func() {
-			defer ticker.Stop()
-			for {
-				select {
-				case <-ctx.Done():
-					return
-				case <-ticker.C:
-					_, _ = service.Cleaner.RunOnce(ctx)
-				}
-			}
-		}()
-	}
 }

model/dto/model_dto.go

@@ -1,161 +0,0 @@
-package dto
-
-import (
-	"gitea.com/red-future/common/beans"
-	"github.com/gogf/gf/v2/frame/g"
-)
-
-// CreateModelReq 添加模型配置
-type CreateModelReq struct {
-	g.Meta               `path:"/createModel" method:"post" tags:"模型管理" summary:"创建模型配置" dc:"添加新的模型配置"`
-	ModelName            string `p:"modelName" json:"modelName" v:"required#modelName不能为空" dc:"模型名称（唯一标识）"`
-	ModelType            int    `p:"modelType" json:"modelType" v:"required#modelType不能为空" dc:"模型类型：1-文本生成 2-图像生成 3-语音 4-视频 5-多模态"`
-	BaseURL              string `p:"baseUrl" json:"baseUrl" v:"required#baseUrl不能为空" dc:"模型服务基础地址（如 gateway(s)://host:port）"`
-	HttpMethod           string `p:"httpMethod" json:"httpMethod" dc:"请求方式：GET/POST（默认POST）"`
-	HeadMsg              string `p:"headMsg" json:"headMsg" dc:"请求头绑定（支持多个，逗号分隔），示例：Authorization:Bearer xxx,Content-Type:application/json"`
-	IsPrivate            *int   `p:"isPrivate" json:"isPrivate" v:"in:0,1#私有化参数只能为0或1" dc:"是否私有化：0-私有（默认） 1-公共"`
-	Enabled              *int   `p:"enabled" json:"enabled" v:"in:0,1#启用参数只能为0或1" dc:"是否启用：0-禁用，1-启用（默认1）"`
-	IsChatModel          *int   `p:"isChatModel" json:"isChatModel" v:"in:0,1#对话模型参数只能为0或1" dc:"是否为对话模型：0-否，1-是（默认0）"`
-	IsOwner              *int   `p:"isOwner" json:"isOwner" v:"in:0,1#是否为所有者参数只能为0或1" dc:"是否为所有者：0-否，1-是（默认0）"`
-	OperatorName         string `p:"operatorName" json:"operatorName" v:"required#operatorName不能为空" dc:"运营商名称"`
-	TokenConfig          any    `p:"tokenConfig" json:"tokenConfig" dc:"token计算配置"`
-	ApiKey               string `p:"apiKey" json:"apiKey" dc:"调用凭证/密钥，用于模型认证"`
-	Form                 any    `p:"form" json:"form" dc:"动态表单配置（JSON），用于前端渲染配置项"`
-	RequestMapping       any    `p:"requestMapping" json:"requestMapping" dc:"请求映射"`
-	ResponseMapping      any    `p:"responseMapping" json:"responseMapping" dc:"返回映射"`
-	ResponseBody         any    `p:"responseBody" json:"responseBody" dc:"返回主体"`
-	ResponseTokenField   string `p:"responseTokenField" json:"responseTokenField" dc:"响应中消耗token的字段映射"`
-	MaxConcurrency       int    `p:"maxConcurrency" json:"maxConcurrency" dc:"最大并发数（默认10）"`
-	QueueLimit           int    `p:"queueLimit" json:"queueLimit" dc:"排队队列上限（默认1000）"`
-	TimeoutSeconds       int    `p:"timeoutSeconds" json:"timeoutSeconds" dc:"请求超时时间（秒，默认600）"`
-	ExpectedSeconds      int    `p:"expectedSeconds" json:"expectedSeconds" dc:"模型预计执行时间（秒，默认600）"`
-	RetryTimes           int    `p:"retryTimes" json:"retryTimes" dc:"失败重试次数（默认3）"`
-	RetryQueueMaxSeconds int    `p:"retryQueueMaxSeconds" json:"retryQueueMaxSeconds" dc:"失败重试最大排队时间（秒，默认600）"`
-	AutoCleanSeconds     int    `p:"autoCleanSeconds" json:"autoCleanSeconds" dc:"任务完成后自动清理时间（秒，默认86400）"`
-	Remark               string `p:"remark" json:"remark" dc:"备注说明"`
-}
-
-type CreateModelRes struct {
-	ID int64 `json:"id,string" dc:"配置ID"`
-}
-
-type UpdateModelReq struct {
-	g.Meta               `path:"/updateModel" method:"put" tags:"模型管理" summary:"更新模型配置" dc:"更新指定ID的模型配置"`
-	ID                   int64  `p:"id" json:"id" v:"required#id不能为空" dc:"配置ID"`
-	ModelName            string `p:"modelName" json:"modelName" dc:"模型名称（唯一标识）"`
-	ModelType            int    `p:"modelType" json:"modelType" dc:"模型类型ID列表（逗号分隔）（可选更新）"`
-	BaseURL              string `p:"baseUrl" json:"baseUrl" dc:"模型服务基础地址"`
-	HttpMethod           string `p:"httpMethod" json:"httpMethod" dc:"请求方式：GET/POST（可选更新）"`
-	HeadMsg              string `p:"headMsg" json:"headMsg" dc:"请求头绑定（可选更新）"`
-	ApiKey               string `p:"apiKey" json:"apiKey" dc:"调用凭证/密钥，用于模型认证（可选更新）"`
-	Form                 any    `p:"form" json:"form" dc:"动态表单配置（JSON）（可选更新）"`
-	RequestMapping       any    `p:"requestMapping" json:"requestMapping" dc:"请求参数映射（可选更新）"`
-	ResponseMapping      any    `p:"responseMapping" json:"responseMapping" dc:"返回参数映射（可选更新）"`
-	ResponseBody         any    `p:"responseBody" json:"responseBody" dc:"返回主体（可选更新）"`
-	ResponseTokenField   string `p:"responseTokenField" json:"responseTokenField" dc:"响应中消耗token的字段映射"`
-	Enabled              *int   `p:"enabled" json:"enabled" dc:"是否启用：0-禁用，1-启用（可选更新）"`
-	IsPrivate            *int   `p:"isPrivate" json:"isPrivate" v:"in:0,1#私有化参数只能为0或1" dc:"是否私有化：0-私有（默认） 1-公共"`
-	IsChatModel          *int   `p:"isChatModel" json:"isChatModel" v:"in:0,1#对话模型参数只能为0或1" dc:"是否为对话模型：0-否，1-是（默认0）"`
-	IsOwner              *int   `p:"isOwner" json:"isOwner" v:"in:0,1#是否为所有者参数只能为0或1" dc:"是否为所有者：0-否，1-是（默认0）"`
-	OperatorName         string `p:"operatorName" json:"operatorName" v:"required#operatorName不能为空" dc:"运营商名称"`
-	TokenConfig          any    `p:"tokenConfig" json:"tokenConfig" dc:"token计算配置"`
-	MaxConcurrency       int    `p:"maxConcurrency" json:"maxConcurrency" dc:"最大并发数（可选更新）"`
-	QueueLimit           int    `p:"queueLimit" json:"queueLimit" dc:"排队队列上限（可选更新）"`
-	TimeoutSeconds       int    `p:"timeoutSeconds" json:"timeoutSeconds" dc:"请求超时时间（秒）（可选更新）"`
-	ExpectedSeconds      int    `p:"expectedSeconds" json:"expectedSeconds" dc:"模型预计执行时间（秒）（可选更新）"`
-	RetryTimes           int    `p:"retryTimes" json:"retryTimes" dc:"失败重试次数（可选更新）"`
-	RetryQueueMaxSeconds int    `p:"retryQueueMaxSeconds" json:"retryQueueMaxSeconds" dc:"失败重试最大排队时间（秒）（可选更新）"`
-	AutoCleanSeconds     int    `p:"autoCleanSeconds" json:"autoCleanSeconds" dc:"自动清理间隔（秒）（可选更新）"`
-	Remark               string `p:"remark" json:"remark" dc:"备注说明（可选更新）"`
-}
-
-type UpdateModelRes struct {
-	ID int64 `json:"id,string" dc:"配置ID"`
-}
-
-// DeleteModelReq 删除模型配置
-type DeleteModelReq struct {
-	g.Meta `path:"/deleteModel" method:"delete" tags:"模型管理" summary:"删除模型配置" dc:"删除指定ID的模型配置"`
-	ID     int64 `p:"id" json:"id,string" v:"required#id不能为空" dc:"配置ID"`
-}
-
-type DeleteModelRes struct {
-	ID int64 `json:"id,string" dc:"配置ID"`
-}
-
-// GetModelReq 获取模型配置详情
-type GetModelReq struct {
-	g.Meta  `path:"/getModel" method:"get" tags:"模型管理" summary:"获取模型配置" dc:"根据模型ID获取配置详情"`
-	ID      int64  `p:"id" json:"id,string" v:"required#id不能为空" dc:"配置ID"`
-	Creator string `p:"creator" json:"creator" dc:"创建人"`
-}
-
-type GetModelRes struct {
-	Model any `json:"model" dc:"模型配置详情"`
-}
-
-// ListModelReq 配置列表
-type ListModelReq struct {
-	g.Meta    `path:"/listModel" method:"get" tags:"模型管理" summary:"模型配置列表" dc:"分页获取模型配置列表"`
-	Page      *beans.Page `json:"page"`
-	ModelName string      `p:"modelName" json:"modelName" dc:"模型名称（模糊查询，可选）"`
-	ModelType int         `p:"modelType" json:"modelType" dc:"模型类型"`
-	Enabled   *int        `p:"enabled" json:"enabled" dc:"是否启用：0-禁用，1-启用"`
-	IsPrivate *int        `p:"isPrivate" json:"isPrivate" dc:"是否私有化 0-私有 1-公共"`
-	IsOwner   *int        `p:"isOwner" json:"isOwner" dc:"是否为所有者 0-否 1-是"`
-	Creator   string      `p:"creator" json:"creator" dc:"创建人"`
-}
-
-type ListModelRes struct {
-	List  any `json:"list" dc:"列表数据"`
-	Total int `json:"total" dc:"总数"`
-}
-
-// AutoTuneReq 动态调参（由上层定时任务每小时触发一次）
-type AutoTuneReq struct {
-	g.Meta        `path:"/autoTune" method:"post" tags:"模型管理" summary:"动态调参" dc:"按 model_name 维度统计指定时间窗口内执行耗时(P90)，动态生成运行时 max_concurrency/queue_limit（不超过配置上限），写入 Redis 供 Worker/CreateTask 使用；windowSeconds 不传默认 3600"`
-	WindowSeconds int `p:"windowSeconds" json:"windowSeconds" dc:"统计窗口秒数；不传/<=0 默认 3600（1小时）"`
-}
-
-type AutoTuneRes struct {
-	List any `json:"list" dc:"调参结果列表"`
-}
-
-type ModelTypeModelItem struct {
-	ID   int64  `json:"id" dc:"模型主键ID"`
-	Name string `json:"name" dc:"模型名称"`
-	Form any    `json:"form" dc:"动态表单配置（JSON数组），用于前端渲染"`
-}
-
-// ListModelTypeReq 模型类型列表（分页）
-type ListTypeReq struct {
-	g.Meta `path:"/listType" method:"get" tags:"模型类型列表" summary:"模型类型列表" dc:"分页获取模型类型列表"`
-}
-
-type TypeItem struct {
-	Type map[int]string `json:"type" dc:"模型类型ID到名称的映射"`
-}
-
-type ListOperatorReq struct {
-	g.Meta `path:"/listOperator" method:"get" tags:"模型管理" summary:"获取运营商列表" dc:"获取运营商列表"`
-}
-
-type ListOperatorRes struct {
-	List []string `json:"list" dc:"运营商名称到ID的映射"`
-}
-
-type UpdateChatModelReq struct {
-	g.Meta `path:"/updateChatModel" method:"post" tags:"模型管理" summary:"更新聊天模型" dc:"更新指定模型的聊天模型"`
-	Id     int64 `p:"id" json:"id" v:"required#model不能为空" dc:"模型id"`
-}
-type UpdateChatModelRes struct {
-	ID int64 `json:"id,string" dc:"模型ID"`
-}
-
-type GetIsChatModelReq struct {
-	g.Meta `path:"/getIsChatModel" method:"get" tags:"模型管理" summary:"获取模型是否为聊天模型" dc:"根据模型ID获取是否为聊天模型"`
-}
-
-type GetIsChatModelRes struct {
-	Model any `json:"model" dc:"模型详情"`
-}

model/dto/model_gateway_logs_stat_dto.go

@@ -0,0 +1,20 @@
+package dto
+
+import "github.com/gogf/gf/v2/frame/g"
+
+// ListModelStatReq 统计列表
+type ListModelStatReq struct {
+	g.Meta    `path:"/listModelStat" method:"get" tags:"统计" summary:"模型请求统计列表" dc:"按天统计模型请求次数，支持分页与条件筛选"`
+	PageNum   int    `p:"pageNum" json:"pageNum" dc:"页码（默认1）"`
+	PageSize  int    `p:"pageSize" json:"pageSize" dc:"每页条数（默认10）"`
+	StartDay  string `p:"startDay" json:"startDay" dc:"开始日期（YYYY-MM-DD，可选）"`
+	EndDay    string `p:"endDay" json:"endDay" dc:"结束日期（YYYY-MM-DD，可选）"`
+	TenantID  *int64 `p:"tenantId" json:"tenantId" dc:"租户ID（可选）"`
+	Creator   string `p:"creator" json:"creator" dc:"创建人（可选，模糊匹配）"`
+	ModelName string `p:"modelName" json:"modelName" dc:"模型名称（可选，模糊匹配）"`
+}
+
+type ListModelStatRes struct {
+	List  any   `json:"list" dc:"列表数据"`
+	Total int64 `json:"total" dc:"总数"`
+}

model/dto/model_gateway_models_dto.go

@@ -0,0 +1,186 @@
+package dto
+
+import (
+	"model-gateway/model/entity"
+
+	"gitea.redpowerfuture.com/red-future/common/beans"
+	"github.com/gogf/gf/v2/frame/g"
+)
+
+// CreateModelReq 添加模型配置
+type CreateModelReq struct {
+	g.Meta           `path:"/createModel" method:"post" tags:"模型管理" summary:"创建模型配置" dc:"添加新的模型配置"`
+	ModelName        string           `p:"modelName" json:"modelName" v:"required#模型名称不能为空" dc:"模型名称（唯一标识）"`
+	ModelType        int              `p:"modelType" json:"modelType" v:"required#模型类型不能为空" dc:"模型类型"`
+	BaseURL          string           `p:"baseUrl" json:"baseUrl" v:"required#模型地址不能为空" dc:"模型服务地址"`
+	HttpMethod       string           `p:"httpMethod" json:"httpMethod" dc:"请求方式：GET/POST（默认POST）"`
+	HeadMsg          map[string]any   `p:"headMsg" json:"headMsg" dc:"请求头JSON结构"`
+	IsPrivate        *int             `p:"isPrivate" json:"isPrivate" dc:"是否私有化：0-私有 1-公共"`
+	Enabled          *int             `p:"enabled" json:"enabled" dc:"是否启用：0-停用 1-启用"`
+	IsChatModel      *int             `p:"isChatModel" json:"isChatModel" dc:"是否为对话模型：0-否 1-是"`
+	CallModel        *int             `p:"callModel" json:"callModel" dc:"调用模式：0-同步 1-异步 2-流式"`
+	RequiredFields   []string         `p:"requiredFields" json:"requiredFields" dc:"必填字段"`
+	IsOwner          *int             `p:"isOwner" json:"isOwner" dc:"是否为所有者：0-否 1-是"`
+	ApiKey           string           `p:"apiKey" json:"apiKey" dc:"调用凭证/密钥"`
+	Form             []map[string]any `p:"form" json:"form" dc:"动态表单配置"`
+	RequestMapping   map[string]any   `p:"requestMapping" json:"requestMapping" dc:"请求映射"`
+	ResponseMapping  map[string]any   `p:"responseMapping" json:"responseMapping" dc:"返回映射"`
+	OperatorName     string           `p:"operatorName" json:"operatorName" dc:"运营商名称"`
+	TokenConfig      map[string]any   `p:"tokenConfig" json:"tokenConfig" dc:"token计算配置"`
+	ExtendMapping    map[string]any   `p:"extendMapping" json:"extendMapping" dc:"附加映射"`
+	QueryConfig      map[string]any   `p:"queryConfig" json:"queryConfig" dc:"查询/回调配置"`
+	StreamConfig     map[string]any   `p:"streamConfig" json:"streamConfig" dc:"流式输出配置"`
+	FirstFrame       string           `p:"firstFrame" json:"firstFrame" dc:"首帧图片参数"`
+	LastFrame        string           `p:"lastFrame" json:"lastFrame" dc:"尾帧图片参数"`
+	MaxConcurrency   int              `p:"maxConcurrency" json:"maxConcurrency" dc:"最大并发数（默认10）"`
+	TimeoutSeconds   int              `p:"timeoutSeconds" json:"timeoutSeconds" dc:"请求超时时间（秒，默认600）"`
+	RetryTimes       int              `p:"retryTimes" json:"retryTimes" dc:"失败重试次数（默认3）"`
+	AutoCleanSeconds int              `p:"autoCleanSeconds" json:"autoCleanSeconds" dc:"任务完成后自动清理时间（秒，默认86400）"`
+	CallbackUrl      string           `p:"callbackUrl" json:"callbackUrl" dc:"回调地址"`
+}
+
+type CreateModelRes struct {
+	ID int64 `json:"id,string" dc:"配置ID"`
+}
+
+type UpdateModelReq struct {
+	g.Meta           `path:"/updateModel" method:"put" tags:"模型管理" summary:"更新模型配置" dc:"更新指定ID的模型配置"`
+	ID               int64            `p:"id" json:"id" v:"required#id不能为空" dc:"配置ID"`
+	ModelName        string           `p:"modelName" json:"modelName" dc:"模型名称"`
+	ModelType        int              `p:"modelType" json:"modelType" dc:"模型类型"`
+	BaseURL          string           `p:"baseUrl" json:"baseUrl" dc:"模型服务地址"`
+	HttpMethod       string           `p:"httpMethod" json:"httpMethod" dc:"请求方式：GET/POST"`
+	HeadMsg          map[string]any   `p:"headMsg" json:"headMsg" dc:"请求头JSON结构"`
+	IsPrivate        *int             `p:"isPrivate" json:"isPrivate" dc:"是否私有化：0-私有 1-公共"`
+	Enabled          *int             `p:"enabled" json:"enabled" dc:"是否启用：0-停用 1-启用"`
+	IsChatModel      *int             `p:"isChatModel" json:"isChatModel" dc:"是否为对话模型：0-否 1-是"`
+	CallModel        *int             `p:"callModel" json:"callModel" dc:"调用模式：0-同步 1-异步 2-流式"`
+	RequiredFields   []string         `p:"requiredFields" json:"requiredFields" dc:"必填字段"`
+	IsOwner          *int             `p:"isOwner" json:"isOwner" dc:"是否为所有者：0-否 1-是"`
+	ApiKey           string           `p:"apiKey" json:"apiKey" dc:"调用凭证/密钥"`
+	Form             []map[string]any `p:"form" json:"form" dc:"动态表单配置"`
+	RequestMapping   map[string]any   `p:"requestMapping" json:"requestMapping" dc:"请求映射"`
+	ResponseMapping  map[string]any   `p:"responseMapping" json:"responseMapping" dc:"返回映射"`
+	OperatorName     string           `p:"operatorName" json:"operatorName" dc:"运营商名称"`
+	TokenConfig      map[string]any   `p:"tokenConfig" json:"tokenConfig" dc:"token计算配置"`
+	ExtendMapping    map[string]any   `p:"extendMapping" json:"extendMapping" dc:"附加映射"`
+	QueryConfig      map[string]any   `p:"queryConfig" json:"queryConfig" dc:"查询/回调配置"`
+	StreamConfig     map[string]any   `p:"streamConfig" json:"streamConfig" dc:"流式输出配置"`
+	FirstFrame       string           `p:"firstFrame" json:"firstFrame" dc:"首帧图片参数"`
+	LastFrame        string           `p:"lastFrame" json:"lastFrame" dc:"尾帧图片参数"`
+	MaxConcurrency   int              `p:"maxConcurrency" json:"maxConcurrency" dc:"最大并发数"`
+	TimeoutSeconds   int              `p:"timeoutSeconds" json:"timeoutSeconds" dc:"请求超时时间（秒）"`
+	RetryTimes       int              `p:"retryTimes" json:"retryTimes" dc:"失败重试次数"`
+	AutoCleanSeconds int              `p:"autoCleanSeconds" json:"autoCleanSeconds" dc:"任务完成后自动清理时间（秒）"`
+	CallbackUrl      string           `p:"callbackUrl" json:"callbackUrl" dc:"回调地址"`
+}
+
+type UpdateModelRes struct {
+	ID int64 `json:"id,string" dc:"配置ID"`
+}
+
+// DeleteModelReq 删除模型配置
+type DeleteModelReq struct {
+	g.Meta `path:"/deleteModel" method:"delete" tags:"模型管理" summary:"删除模型配置" dc:"删除指定ID的模型配置"`
+	ID     int64 `p:"id" json:"id,string" v:"required#id不能为空" dc:"配置ID"`
+}
+
+type DeleteModelRes struct {
+	ID int64 `json:"id,string" dc:"配置ID"`
+}
+
+// GetModelReq 获取模型配置详情
+type GetModelReq struct {
+	g.Meta      `path:"/getModel" method:"get" tags:"模型管理" summary:"获取模型配置" dc:"根据模型ID获取配置详情"`
+	ID          int64  `p:"id" json:"id,string" dc:"配置ID"`
+	Creator     string `p:"creator" json:"creator" dc:"创建人"`
+	IsChatModel *int   `p:"isChatModel" json:"isChatModel" dc:"是否为聊天模型"`
+	ModelName   string `p:"modelName" json:"modelName" dc:"模型名称（唯一标识）"`
+}
+
+type GetModelRes struct {
+	Model *entity.ModelGatewayModel `json:"model" dc:"模型配置详情"`
+}
+
+// ListModelReq 配置列表
+type ListModelReq struct {
+	g.Meta    `path:"/listModel" method:"get" tags:"模型管理" summary:"模型配置列表" dc:"分页获取模型配置列表"`
+	Page      *beans.Page `json:"page"`
+	ModelName string      `p:"modelName" json:"modelName" dc:"模型名称（模糊查询，可选）"`
+	ModelType int         `p:"modelType" json:"modelType" dc:"模型类型"`
+	Enabled   *int        `p:"enabled" json:"enabled" dc:"是否启用：0-禁用，1-启用"`
+	IsPrivate *int        `p:"isPrivate" json:"isPrivate" dc:"是否私有化 0-私有 1-公共"`
+	IsOwner   *int        `p:"isOwner" json:"isOwner" dc:"是否为所有者 0-否 1-是"`
+	Creator   string      `p:"creator" json:"creator" dc:"创建人"`
+}
+
+type ListModelRes struct {
+	List  any `json:"list" dc:"列表数据"`
+	Total int `json:"total" dc:"总数"`
+}
+
+// AutoTuneReq 动态调参（由上层定时任务每小时触发一次）
+type AutoTuneReq struct {
+	g.Meta        `path:"/autoTune" method:"post" tags:"模型管理" summary:"动态调参" dc:"按 model_name 维度统计指定时间窗口内执行耗时(P90)，动态生成运行时 max_concurrency/queue_limit（不超过配置上限），写入 Redis 供 Worker/CreateTask 使用；windowSeconds 不传默认 3600"`
+	WindowSeconds int `p:"windowSeconds" json:"windowSeconds" dc:"统计窗口秒数；不传/<=0 默认 3600（1小时）"`
+}
+
+type AutoTuneRes struct {
+	List any `json:"list" dc:"调参结果列表"`
+}
+
+type ModelTypeModelItem struct {
+	ID   int64  `json:"id" dc:"模型主键ID"`
+	Name string `json:"name" dc:"模型名称"`
+	Form any    `json:"form" dc:"动态表单配置（JSON数组），用于前端渲染"`
+}
+
+// ListModelTypeReq 模型类型列表（分页）
+type ListTypeReq struct {
+	g.Meta `path:"/listType" method:"get" tags:"模型类型列表" summary:"模型类型列表" dc:"分页获取模型类型列表"`
+}
+
+type TypeItem struct {
+	Type map[int]string `json:"type" dc:"模型类型ID到名称的映射"`
+}
+
+type ListOperatorReq struct {
+	g.Meta `path:"/listOperator" method:"get" tags:"模型管理" summary:"获取运营商列表" dc:"获取运营商列表"`
+}
+
+type ListOperatorRes struct {
+	List []string `json:"list" dc:"运营商名称到ID的映射"`
+}
+
+type UpdateChatModelReq struct {
+	g.Meta `path:"/updateChatModel" method:"post" tags:"模型管理" summary:"更新聊天模型" dc:"更新指定模型的聊天模型"`
+	Id     int64 `p:"id" json:"id" v:"required#model不能为空" dc:"模型id"`
+}
+type UpdateChatModelRes struct {
+	ID int64 `json:"id,string" dc:"模型ID"`
+}
+
+type GetIsChatModelReq struct {
+	g.Meta `path:"/getIsChatModel" method:"get" tags:"模型管理" summary:"获取模型是否为聊天模型" dc:"根据模型ID获取是否为聊天模型"`
+}
+
+type GetIsChatModelRes struct {
+	Model any `json:"model" dc:"模型详情"`
+}
+
+// NodeFormField 节点表单
+type NodeFormField struct {
+	Value           any            `json:"value" dc:"字段值"`
+	Field           string         `json:"field" dc:"字段标识"`
+	Label           string         `json:"label" dc:"字段标签"`
+	Type            string         `json:"type" dc:"字段类型"`
+	Required        bool           `json:"required" dc:"是否必填"`
+	Default         any            `json:"default,omitempty" dc:"默认值"`
+	Options         []SelectOption `json:"options" dc:"下拉选项列表"`
+	FieldConstraint any            `json:"fieldConstraint" dc:"字段约束"`
+}
+
+type SelectOption struct {
+	Label string `json:"label" dc:"选项标签"`
+	Value string `json:"value" dc:"选项值"`
+}

model/dto/model_gateway_task_dto.go

@@ -1,26 +1,60 @@
 package dto
 
-import "github.com/gogf/gf/v2/frame/g"
+import (
+	"github.com/gogf/gf/v2/frame/g"
+)
 
 // CreateTaskReq 创建异步任务
 type CreateTaskReq struct {
 	g.Meta         `path:"/createTask" method:"post" tags:"任务管理" summary:"创建异步任务" dc:"创建异步任务并返回任务ID；创建成功后会立即异步尝试执行当前任务，执行成功后按回调配置触发钩子"`
-	ModelName      string `p:"modelName" json:"modelName" v:"required#modelName不能为空" dc:"模型名称"`
-	BizName        string `p:"bizName" json:"bizName" dc:"业务名称（调用方模块/系统，用于统计）"`
-	CallbackUrl    string `p:"callbackUrl" json:"callbackUrl" dc:"回调地址（可选，用于后续业务通知）"`
-	InputRef       string `p:"inputRef" json:"inputRef" dc:"输入引用（如OSS/文件引用等）"`
-	RequestPayload any    `p:"requestPayload" json:"requestPayload" dc:"请求负载（透传给模型服务）"`
-	EpicycleId     int64  `json:"epicycleId" dc:"轮次ID"`
+	ModelName      string         `p:"modelName" json:"modelName" v:"required#modelName不能为空" dc:"模型名称"`
+	BizName        string         `p:"bizName" json:"bizName" dc:"业务名称（调用方模块/系统，用于统计）"`
+	CallbackUrl    string         `p:"callbackUrl" json:"callbackUrl" dc:"回调地址（可选，用于后续业务通知）"`
+	RequestPayload map[string]any `p:"requestPayload" json:"requestPayload" dc:"请求负载（透传给模型服务）"`
+	EpicycleId     int64          `json:"epicycleId" dc:"轮次ID"`
+	BuildType      int64          `json:"buildType" dc:"构建类型：1-提示词构建 2-节点构建"`
 }
 
 type CreateTaskRes struct {
 	TaskID string `json:"taskId" dc:"任务ID"`
 }
 
+type ModelTaskCallbackReq struct {
+	g.Meta  `path:"/modelCallback" method:"post" tags:"异步任务" summary:"模型任务回调通知"`
+	TaskID  string         `json:"id" dc:"任务ID"`
+	Status  string         `json:"status" dc:"queued/running/succeeded/failed/expired"`
+	Content map[string]any `json:"content,omitempty" dc:"任务结果内容"`
+	Usage   map[string]any `json:"usage,omitempty" dc:"token用量"`
+}
+
+type ModelTaskCallbackRes struct {
+	Success bool `json:"success" dc:"是否接收成功"`
+}
+
+// QueryPendingTasksReq 批量轮询请求
+type QueryPendingTasksReq struct {
+	g.Meta `path:"/queryPending" method:"get" tags:"异步任务" summary:"批量轮询进行中的任务"`
+	Limit  int `p:"limit" json:"limit" dc:"查询数量，默认10"`
+}
+
+// QueryPendingTasksRes 批量轮询响应
+type QueryPendingTasksRes struct {
+	Total   int             `json:"total" dc:"本次查询数量"`
+	Results []QueryTaskItem `json:"results" dc:"查询结果列表"`
+}
+
+// QueryTaskItem 单个任务查询结果
+type QueryTaskItem struct {
+	TaskID  string         `json:"taskId" dc:"任务ID"`
+	Status  string         `json:"status" dc:"任务状态"`
+	Content map[string]any `json:"content,omitempty" dc:"结果内容"`
+	Usage   map[string]any `json:"usage,omitempty" dc:"token用量"`
+}
+
 // GetTaskResultReq 获取结果（只返回 oss 地址）
 type GetTaskResultReq struct {
 	g.Meta `path:"/getTaskResult" method:"get" tags:"任务管理" summary:"获取任务结果" dc:"根据任务ID获取结果（只返回OSS地址）"`
-	TaskID string `p:"taskId" json:"taskId" v:"required#taskId不能为空" dc:"任务ID"`
+	TaskID string `p:"taskId" json:"taskId" v:"required#taskwId不能为空" dc:"任务ID"`
 }
 
 type GetTaskResultRes struct {
@@ -34,24 +68,26 @@ type GetTaskBatchReq struct {
 	TaskIDs []string `p:"taskIds" json:"taskIds" v:"required#taskIds不能为空" dc:"任务ID列表"`
 }
 
-type GetTaskBatchItem struct {
-	TaskID  string `json:"taskId" dc:"任务ID"`
-	State   int    `json:"state" dc:"任务状态"`
-	OssFile string `json:"ossFile" dc:"结果文件OSS地址"`
-}
-
 type GetTaskBatchRes struct {
 	List []GetTaskBatchItem `json:"list" dc:"任务列表"`
 }
 
+type GetTaskBatchItem struct {
+	TaskID     string         `json:"taskId" dc:"任务ID"`
+	State      int            `json:"state" dc:"任务状态"`
+	OssFile    string         `json:"ossFile" dc:"结果文件OSS地址"`
+	TextResult map[string]any `json:"textResult" dc:"文本结果"`
+}
+
 // ListTaskReq 任务列表分页查询
 type ListTaskReq struct {
 	g.Meta    `path:"/listTask" method:"get" tags:"任务管理" summary:"任务列表" dc:"分页查询任务列表，支持按状态/模型名称/task_id过滤"`
 	PageNum   int    `p:"pageNum" json:"pageNum" dc:"页码（默认1）"`
 	PageSize  int    `p:"pageSize" json:"pageSize" dc:"每页条数（默认10）"`
 	ModelName string `p:"modelName" json:"modelName" dc:"模型名称（模糊匹配）"`
+	BizName   string `p:"bizName" json:"bizName" dc:"业务名称"`
 	TaskID    string `p:"taskId" json:"taskId" dc:"任务ID（模糊匹配）"`
-	State     *int   `p:"state" json:"state" dc:"任务状态（0/1/2/3/4，可选）"`
+	State     int    `p:"state" json:"state" dc:"任务状态（0/1/2/3/4，可选）"`
 }
 
 type ListTaskRes struct {
@@ -69,12 +105,3 @@ type RunWorkReq struct {
 type RunWorkRes struct {
 	Claimed int `json:"claimed" dc:"本次抢占并处理的任务数"`
 }
-
-// CleanWorkReq 手动触发 cleaner 执行一次（由上层定时任务调用）
-type CleanWorkReq struct {
-	g.Meta `path:"/cleanWork" method:"post" tags:"任务管理" summary:"执行一次Cleaner" dc:"手动触发一次清理/重试（用于由上层定时任务控制）"`
-}
-
-type CleanWorkRes struct {
-	Ok bool `json:"ok" dc:"是否执行成功"`
-}

model/dto/stat_dto.go

@@ -1,20 +0,0 @@
-package dto
-
-import "github.com/gogf/gf/v2/frame/g"
-
-// ListModelStatReq 统计列表
-type ListModelStatReq struct {
-	g.Meta    `path:"/listModelStat" method:"get" tags:"统计" summary:"模型请求统计列表" dc:"按天统计模型请求次数，支持分页与条件筛选"`
-	PageNum   int         `p:"pageNum" json:"pageNum" dc:"页码（默认1）"`
-	PageSize  int         `p:"pageSize" json:"pageSize" dc:"每页条数（默认10）"`
-	StartDay  string      `p:"startDay" json:"startDay" dc:"开始日期（YYYY-MM-DD，可选）"`
-	EndDay    string      `p:"endDay" json:"endDay" dc:"结束日期（YYYY-MM-DD，可选）"`
-	TenantID  *int64      `p:"tenantId" json:"tenantId" dc:"租户ID（可选）"`
-	Creator   string      `p:"creator" json:"creator" dc:"创建人（可选，模糊匹配）"`
-	ModelName string      `p:"modelName" json:"modelName" dc:"模型名称（可选，模糊匹配）"`
-}
-
-type ListModelStatRes struct {
-	List  any   `json:"list" dc:"列表数据"`
-	Total int64 `json:"total" dc:"总数"`
-}

model/entity/asynch_model.go

@@ -1,94 +0,0 @@
-package entity
-
-import "gitea.com/red-future/common/beans"
-
-type asynchModelCol struct {
-	beans.SQLBaseCol
-	ModelName          string
-	ModelType          string
-	BaseURL            string
-	HttpMethod         string
-	HeadMsg            string
-	FormJSON           string
-	RequestMapping     string
-	ResponseMapping    string
-	ResponseBody       string
-	ResponseTokenField string
-	Prompt             string
-	IsPrivate          string
-	IsChatModel        string
-	ApiKey             string
-	Enabled            string
-	MaxConcurrency     string
-	QueueLimit         string
-	TimeoutSeconds     string
-	ExpectedSeconds    string
-	RetryTimes         string
-	RetryQueueMaxSecs  string
-	AutoCleanSeconds   string
-	Remark             string
-	IsOwner            string
-	OperatorName       string
-	TokenConfig        string
-}
-
-var AsynchModelCol = asynchModelCol{
-	SQLBaseCol:         beans.DefSQLBaseCol,
-	ModelName:          "model_name",
-	ModelType:          "model_type",
-	BaseURL:            "base_url",
-	HttpMethod:         "http_method",
-	HeadMsg:            "head_msg",
-	FormJSON:           "form_json",
-	RequestMapping:     "request_mapping",
-	ResponseMapping:    "response_mapping",
-	ResponseBody:       "response_body",
-	ResponseTokenField: "response_token_field",
-	Prompt:             "prompt",
-	IsPrivate:          "is_private",
-	IsChatModel:        "is_chat_model",
-	ApiKey:             "api_key",
-	Enabled:            "enabled",
-	MaxConcurrency:     "max_concurrency",
-	QueueLimit:         "queue_limit",
-	TimeoutSeconds:     "timeout_seconds",
-	ExpectedSeconds:    "expected_seconds",
-	RetryTimes:         "retry_times",
-	RetryQueueMaxSecs:  "retry_queue_max_seconds",
-	AutoCleanSeconds:   "auto_clean_seconds",
-	Remark:             "remark",
-	IsOwner:            "is_owner",
-	OperatorName:       "operator_name",
-	TokenConfig:        "token_config",
-}
-
-// AsynchModel 异步模型配置
-type AsynchModel struct {
-	beans.SQLBaseDO      `orm:",inline"`
-	ModelName            string `orm:"model_name" json:"modelName"`
-	ModelType            int    `orm:"model_type" json:"modelType"`
-	BaseURL              string `orm:"base_url" json:"baseUrl"`
-	HttpMethod           string `orm:"http_method" json:"httpMethod"`
-	HeadMsg              string `orm:"head_msg" json:"headMsg"`
-	Form                 any    `orm:"form_json" json:"form"`
-	RequestMapping       any    `orm:"request_mapping" json:"requestMapping"`
-	ResponseMapping      any    `orm:"response_mapping" json:"responseMapping"`
-	ResponseBody         any    `orm:"response_body" json:"responseBody"`
-	ResponseTokenField   string `orm:"response_token_field" json:"responseTokenField"`
-	Prompt               string `orm:"prompt" json:"prompt"`
-	IsPrivate            *int   `orm:"is_private" json:"isPrivate"`
-	IsChatModel          *int   `orm:"is_chat_model" json:"isChatModel"`
-	ApiKey               string `orm:"api_key" json:"apiKey"`
-	Enabled              *int   `orm:"enabled" json:"enabled"`
-	MaxConcurrency       int    `orm:"max_concurrency" json:"maxConcurrency"`
-	QueueLimit           int    `orm:"queue_limit" json:"queueLimit"`
-	TimeoutSeconds       int    `orm:"timeout_seconds" json:"timeoutSeconds"`
-	ExpectedSeconds      int    `orm:"expected_seconds" json:"expectedSeconds"`
-	RetryTimes           int    `orm:"retry_times" json:"retryTimes"`
-	RetryQueueMaxSeconds int    `orm:"retry_queue_max_seconds" json:"retryQueueMaxSeconds"`
-	AutoCleanSeconds     int    `orm:"auto_clean_seconds" json:"autoCleanSeconds"`
-	Remark               string `orm:"remark" json:"remark"`
-	IsOwner              *int   `json:"isOwner" orm:"is_owner"`
-	OperatorName         string `orm:"operator_name" json:"operatorName"`
-	TokenConfig          any    `orm:"token_config" json:"tokenConfig"`
-}

model/entity/asynch_task.go

@@ -1,89 +0,0 @@
-package entity
-
-import (
-	"gitea.com/red-future/common/beans"
-	"github.com/gogf/gf/v2/os/gtime"
-)
-
-type asynchTaskCol struct {
-	beans.SQLBaseCol
-	ModelName       string
-	TaskID          string
-	BizName         string
-	CallbackURL     string
-	ModelKey        string
-	State           string
-	OssFile         string
-	FileType        string
-	FileSize        string
-	ErrorMsg        string
-	StartedAt       string
-	FinishedAt      string
-	DurationSeconds string
-	ExpireAt        string
-	RetryCount      string
-	EnqueueAt       string
-	Phase           string
-	TmpFile         string
-	InputRef        string
-	RequestPayload  string
-	TextResult      string
-	EpicycleId      string
-	ExpendTokens    string
-}
-
-var AsynchTaskCol = asynchTaskCol{
-	SQLBaseCol:      beans.DefSQLBaseCol,
-	ModelName:       "model_name",
-	TaskID:          "task_id",
-	BizName:         "biz_name",
-	CallbackURL:     "callback_url",
-	ModelKey:        "model_key",
-	State:           "state",
-	OssFile:         "oss_file",
-	FileType:        "file_type",
-	FileSize:        "file_size",
-	ErrorMsg:        "error_msg",
-	StartedAt:       "started_at",
-	FinishedAt:      "finished_at",
-	DurationSeconds: "duration_seconds",
-	ExpireAt:        "expire_at",
-	RetryCount:      "retry_count",
-	EnqueueAt:       "enqueue_at",
-	Phase:           "phase",
-	TmpFile:         "tmp_file",
-	InputRef:        "input_ref",
-	RequestPayload:  "request_payload",
-	TextResult:      "text_result",
-	EpicycleId:      "epicycle_id",
-	ExpendTokens:    "expend_tokens",
-}
-
-// AsynchTask 异步任务
-type AsynchTask struct {
-	beans.SQLBaseDO      `orm:",inline"`
-	ModelName            string      `orm:"model_name" json:"modelName"`
-	TaskID               string      `orm:"task_id" json:"taskId"`
-	BizName              string      `orm:"biz_name" json:"bizName"`
-	CallbackURL          string      `orm:"callback_url" json:"callbackUrl"`
-	ModelKey             string      `orm:"model_key" json:"modelKey"`
-	State                int         `orm:"state" json:"state"` // 0排队中/1执行中/2成功/3失败/4已下载
-	OssFile              string      `orm:"oss_file" json:"ossFile"`
-	FileType             string      `orm:"file_type" json:"fileType"`
-	FileSize             int64       `orm:"file_size" json:"fileSize"`
-	ErrorMsg             string      `orm:"error_msg" json:"errorMsg"`
-	StartedAt            *gtime.Time `orm:"started_at" json:"startedAt"`
-	FinishedAt           *gtime.Time `orm:"finished_at" json:"finishedAt"`
-	DurationSeconds      int64       `orm:"duration_seconds" json:"durationSeconds"`
-	ExpireAt             *gtime.Time `orm:"expire_at" json:"expireAt"` // 已下载(state=4)后的过期时间
-	RetryCount           int         `orm:"retry_count" json:"retryCount"`
-	EnqueueAt            *gtime.Time `orm:"enqueue_at" json:"enqueueAt"`
-	Phase                int         `orm:"phase" json:"phase"`      // 0模型阶段/1OSS阶段
-	TmpFile              string      `orm:"tmp_file" json:"tmpFile"` // 临时结果文件路径
-	InputRef             string      `orm:"input_ref" json:"inputRef"`
-	RequestPayload       any         `orm:"request_payload" json:"requestPayload"`
-	TextResult           string      `orm:"text_result" json:"text"`
-	EpicycleId           int64       `orm:"epicycle_id" json:"epicycleId"`     // 轮次ID（用于标识同一轮次的任务）
-	ExpendTokens         int64       `orm:"expend_tokens" json:"expendTokens"` // 消耗 token 数
-	RetryQueueMaxSeconds int         `orm:"retry_queue_max_seconds" json:"-"`
-}

model/entity/logs_model_op.go

@@ -1,57 +0,0 @@
-package entity
-
-import (
-	"gitea.com/red-future/common/beans"
-)
-
-type LogsModelPpCol struct {
-	beans.SQLBaseCol
-	IP              string
-	UserAgent       string
-	APIPath         string
-	HttpMethod      string
-	BizName         string
-	ModelName       string
-	TaskID          string
-	OpType          string
-	Success         string
-	ErrorMsg        string
-	CostMs          string
-	RequestPayload  string
-	ResponsePayload string
-}
-
-var LogsModelOpCol = LogsModelPpCol{
-	SQLBaseCol:      beans.DefSQLBaseCol,
-	IP:              "ip",
-	UserAgent:       "user_agent",
-	APIPath:         "api_path",
-	HttpMethod:      "http_method",
-	BizName:         "biz_name",
-	ModelName:       "model_name",
-	TaskID:          "task_id",
-	OpType:          "op_type",
-	Success:         "success",
-	ErrorMsg:        "error_msg",
-	CostMs:          "cost_ms",
-	RequestPayload:  "request_payload",
-	ResponsePayload: "response_payload",
-}
-
-// LogsModelOp 操作日志（创建任务等）
-type LogsModelOp struct {
-	beans.SQLBaseDO `orm:",inline"`
-	IP              string `orm:"ip" json:"ip"`
-	UserAgent       string `orm:"user_agent" json:"userAgent"`
-	APIPath         string `orm:"api_path" json:"apiPath"`
-	HttpMethod      string `orm:"http_method" json:"httpMethod"`
-	BizName         string `orm:"biz_name" json:"bizName"`
-	ModelName       string `orm:"model_name" json:"modelName"`
-	TaskID          string `orm:"task_id" json:"taskId"`
-	OpType          string `orm:"op_type" json:"opType"`
-	Success         int    `orm:"success" json:"success"`
-	ErrorMsg        string `orm:"error_msg" json:"errorMsg"`
-	CostMs          int64  `orm:"cost_ms" json:"costMs"`
-	RequestPayload  any    `orm:"request_payload" json:"requestPayload"`
-	ResponsePayload any    `orm:"response_payload" json:"responsePayload"`
-}

model/entity/logs_model_stat.go

@@ -1,38 +0,0 @@
-package entity
-
-import (
-	"github.com/gogf/gf/v2/os/gtime"
-)
-
-// LogsModelStatCol 字段常量
-type LogsModelStatCol struct {
-	Day          string
-	TenantId     string
-	Creator      string
-	ModelName    string
-	RequestCount string
-	CreatedAt    string
-	UpdatedAt    string
-}
-
-var LogsModelStatCols = LogsModelStatCol{
-	Day:          "day",
-	TenantId:     "tenant_id",
-	Creator:      "creator",
-	ModelName:    "model_name",
-	RequestCount: "request_count",
-	CreatedAt:    "created_at",
-	UpdatedAt:    "updated_at",
-}
-
-// LogsModelStat 按天统计：某天/租户/创建人/模型的请求次数
-// 注：这里不走通用 SQLBaseDO，采用联合唯一键（day,tenant_id,creator,model_name）做 UPSERT 原子累加。
-type LogsModelStat struct {
-	Day          *gtime.Time `orm:"day" json:"day"`                    // 日期（建议仅使用日期部分）
-	TenantId     int64       `orm:"tenant_id" json:"tenantId"`         // 租户ID
-	Creator      string      `orm:"creator" json:"creator"`            // 创建人/操作人
-	ModelName    string      `orm:"model_name" json:"modelName"`       // 模型名称
-	RequestCount int64       `orm:"request_count" json:"requestCount"` // 请求次数
-	CreatedAt    *gtime.Time `orm:"created_at" json:"createdAt"`       // 创建时间
-	UpdatedAt    *gtime.Time `orm:"updated_at" json:"updatedAt"`       // 更新时间
-}

model/entity/model_gateway_logs_op.go

@@ -0,0 +1,56 @@
+package entity
+
+import "gitea.redpowerfuture.com/red-future/common/beans"
+
+// ModelGatewayLogsOpCol 字段常量
+type modelGatewayLogsOpCol struct {
+	beans.SQLBaseCol
+	IP              string
+	UserAgent       string
+	APIPath         string
+	HttpMethod      string
+	BizName         string
+	ModelName       string
+	TaskID          string
+	OpType          string
+	Success         string
+	ErrorMsg        string
+	CostMs          string
+	RequestPayload  string
+	ResponsePayload string
+}
+
+var ModelGatewayLogsOpCol = modelGatewayLogsOpCol{
+	SQLBaseCol:      beans.DefSQLBaseCol,
+	IP:              "ip",
+	UserAgent:       "user_agent",
+	APIPath:         "api_path",
+	HttpMethod:      "http_method",
+	BizName:         "biz_name",
+	ModelName:       "model_name",
+	TaskID:          "task_id",
+	OpType:          "op_type",
+	Success:         "success",
+	ErrorMsg:        "error_msg",
+	CostMs:          "cost_ms",
+	RequestPayload:  "request_payload",
+	ResponsePayload: "response_payload",
+}
+
+// ModelGatewayLogsOp 操作日志
+type ModelGatewayLogsOp struct {
+	beans.SQLBaseDO `orm:",inline"`
+	IP              string          `orm:"ip" json:"ip"`
+	UserAgent       string          `orm:"user_agent" json:"userAgent"`
+	APIPath         string          `orm:"api_path" json:"apiPath"`
+	HttpMethod      string          `orm:"http_method" json:"httpMethod"`
+	BizName         string          `orm:"biz_name" json:"bizName"`
+	ModelName       string          `orm:"model_name" json:"modelName"`
+	TaskID          string          `orm:"task_id" json:"taskId"`
+	OpType          string          `orm:"op_type" json:"opType"`
+	Success         int             `orm:"success" json:"success"`
+	ErrorMsg        string          `orm:"error_msg" json:"errorMsg"`
+	CostMs          int64           `orm:"cost_ms" json:"costMs"`
+	RequestPayload  *RequestPayload `orm:"request_payload" json:"requestPayload"`
+	ResponsePayload map[string]any  `orm:"response_payload" json:"responsePayload"`
+}

model/entity/model_gateway_logs_stat.go

@@ -0,0 +1,35 @@
+package entity
+
+import "github.com/gogf/gf/v2/os/gtime"
+
+// ModelGatewayLogsStatCol 字段常量
+type ModelGatewayLogsStatCol struct {
+	Day          string
+	TenantId     string
+	Creator      string
+	ModelName    string
+	RequestCount string
+	CreatedAt    string
+	UpdatedAt    string
+}
+
+var ModelGatewayLogsStatCols = ModelGatewayLogsStatCol{
+	Day:          "day",
+	TenantId:     "tenant_id",
+	Creator:      "creator",
+	ModelName:    "model_name",
+	RequestCount: "request_count",
+	CreatedAt:    "created_at",
+	UpdatedAt:    "updated_at",
+}
+
+// ModelGatewayLogsStat 按天统计
+type ModelGatewayLogsStat struct {
+	Day          *gtime.Time `orm:"day" json:"day"`
+	TenantId     uint64      `orm:"tenant_id" json:"tenantId"`
+	Creator      string      `orm:"creator" json:"creator"`
+	ModelName    string      `orm:"model_name" json:"modelName"`
+	RequestCount int64       `orm:"request_count" json:"requestCount"`
+	CreatedAt    *gtime.Time `orm:"created_at" json:"createdAt"`
+	UpdatedAt    *gtime.Time `orm:"updated_at" json:"updatedAt"`
+}

model/entity/model_gateway_model.go

@@ -0,0 +1,99 @@
+package entity
+
+import "gitea.redpowerfuture.com/red-future/common/beans"
+
+type modelGatewayModelCol struct {
+	beans.SQLBaseCol
+	ModelName        string
+	ModelType        string
+	BaseURL          string
+	HttpMethod       string
+	HeadMsg          string
+	FormJSON         string
+	RequestMapping   string
+	ResponseMapping  string
+	ResponseBody     string
+	RequiredFields   string
+	IsPrivate        string
+	IsChatModel      string
+	CallMode         string
+	ApiKey           string
+	Enabled          string
+	MaxConcurrency   string
+	TimeoutSeconds   string
+	RetryTimes       string
+	AutoCleanSeconds string
+	IsOwner          string
+	OperatorName     string
+	TokenConfig      string
+	ExtendMapping    string
+	QueryConfig      string
+	StreamConfig     string
+	FirstFrame       string
+	LastFrame        string
+}
+
+var ModelGatewayModelCol = modelGatewayModelCol{
+	SQLBaseCol:       beans.DefSQLBaseCol,
+	ModelName:        "model_name",
+	ModelType:        "model_type",
+	BaseURL:          "base_url",
+	HttpMethod:       "http_method",
+	HeadMsg:          "head_msg",
+	FormJSON:         "form_json",
+	RequestMapping:   "request_mapping",
+	ResponseMapping:  "response_mapping",
+	RequiredFields:   "required_fields",
+	IsPrivate:        "is_private",
+	IsChatModel:      "is_chat_model",
+	CallMode:         "call_mode",
+	ApiKey:           "api_key",
+	Enabled:          "enabled",
+	MaxConcurrency:   "max_concurrency",
+	TimeoutSeconds:   "timeout_seconds",
+	RetryTimes:       "retry_times",
+	AutoCleanSeconds: "auto_clean_seconds",
+	IsOwner:          "is_owner",
+	OperatorName:     "operator_name",
+	TokenConfig:      "token_config",
+	ExtendMapping:    "extend_mapping",
+	QueryConfig:      "query_config",
+	StreamConfig:     "stream_config",
+	FirstFrame:       "first_frame",
+	LastFrame:        "last_frame",
+}
+
+type ModelGatewayModel struct {
+	beans.SQLBaseDO  `orm:",inline"`
+	ModelName        string           `orm:"model_name" json:"modelName"`
+	ModelType        int              `orm:"model_type" json:"modelType"`
+	BaseURL          string           `orm:"base_url" json:"baseUrl"`
+	HttpMethod       string           `orm:"http_method" json:"httpMethod"`
+	HeadMsg          map[string]any   `orm:"head_msg" json:"headMsg"`
+	Form             []map[string]any `orm:"form_json" json:"form"`
+	RequestMapping   map[string]any   `orm:"request_mapping" json:"requestMapping"`
+	ResponseMapping  map[string]any   `orm:"response_mapping" json:"responseMapping"`
+	RequiredFields   []string         `orm:"required_fields" json:"requiredFields"`
+	IsPrivate        *int             `orm:"is_private" json:"isPrivate"`
+	IsChatModel      *int             `orm:"is_chat_model" json:"isChatModel"`
+	CallMode         *int             `orm:"call_mode" json:"callMode"`
+	ApiKey           string           `orm:"api_key" json:"apiKey"`
+	Enabled          *int             `orm:"enabled" json:"enabled"`
+	MaxConcurrency   int              `orm:"max_concurrency" json:"maxConcurrency"`
+	TimeoutSeconds   int              `orm:"timeout_seconds" json:"timeoutSeconds"`
+	RetryTimes       int              `orm:"retry_times" json:"retryTimes"`
+	AutoCleanSeconds int              `orm:"auto_clean_seconds" json:"autoCleanSeconds"`
+	IsOwner          *int             `orm:"is_owner" json:"isOwner"`
+	OperatorName     string           `orm:"operator_name" json:"operatorName"`
+	TokenConfig      map[string]any   `orm:"token_config" json:"tokenConfig"`
+	ExtendMapping    map[string]any   `orm:"extend_mapping" json:"extendMapping"`
+	QueryConfig      map[string]any   `orm:"query_config" json:"queryConfig"`
+	StreamConfig     map[string]any   `orm:"stream_config" json:"streamConfig"`
+	FirstFrame       string           `orm:"first_frame" json:"firstFrame"`
+	LastFrame        string           `orm:"last_frame" json:"lastFrame"`
+}
+
+const ( //ResponseMapping 下的字段
+	ResponseBody = "response_body" //返回主体
+	TotalTokens  = "total_tokens"  //总token数
+)

model/entity/model_gateway_task.go

@@ -0,0 +1,76 @@
+package entity
+
+import (
+	"gitea.redpowerfuture.com/red-future/common/beans"
+)
+
+type modelGatewayTaskCol struct {
+	beans.SQLBaseCol
+	ModelName       string
+	TaskID          string
+	BizName         string
+	CallbackURL     string
+	State           string
+	Phase           string
+	ErrorMsg        string
+	ResultFile      string
+	TextResult      string
+	ExpendTokens    string
+	DurationSeconds string
+	RetryCount      string
+	TmpFile         string
+	RequestPayload  string
+	EpicycleId      string
+}
+
+var ModelGatewayTaskCol = modelGatewayTaskCol{
+	SQLBaseCol:      beans.DefSQLBaseCol,
+	ModelName:       "model_name",
+	TaskID:          "task_id",
+	BizName:         "biz_name",
+	CallbackURL:     "callback_url",
+	State:           "state",
+	Phase:           "phase",
+	ErrorMsg:        "error_msg",
+	ResultFile:      "result_file",
+	TextResult:      "text_result",
+	ExpendTokens:    "expend_tokens",
+	DurationSeconds: "duration_seconds",
+	RetryCount:      "retry_count",
+	TmpFile:         "tmp_file",
+	RequestPayload:  "request_payload",
+	EpicycleId:      "epicycle_id",
+}
+
+// ModelGatewayTask 模型网关任务
+type ModelGatewayTask struct {
+	beans.SQLBaseDO `orm:",inline"`
+	ModelName       string          `orm:"model_name" json:"modelName"`
+	TaskID          string          `orm:"task_id" json:"taskId"`
+	BizName         string          `orm:"biz_name" json:"bizName"`
+	CallbackURL     string          `orm:"callback_url" json:"callbackUrl"`
+	State           int             `orm:"state" json:"state"`
+	Phase           int             `orm:"phase" json:"phase"`
+	ErrorMsg        string          `orm:"error_msg" json:"errorMsg"`
+	ResultFile      *ResultFile     `orm:"result_file" json:"resultFile"`
+	TextResult      map[string]any  `orm:"text_result" json:"text"`
+	ExpendTokens    int64           `orm:"expend_tokens" json:"expendTokens"`
+	DurationSeconds int64           `orm:"duration_seconds" json:"durationSeconds"`
+	RetryCount      int             `orm:"retry_count" json:"retryCount"`
+	TmpFile         string          `orm:"tmp_file" json:"tmpFile"`
+	RequestPayload  *RequestPayload `orm:"request_payload" json:"requestPayload"`
+	EpicycleId      int64           `orm:"epicycle_id" json:"epicycleId"`
+}
+
+// ResultFile OSS 结果文件
+type ResultFile struct {
+	OssFile  string `json:"ossFile"`
+	FileType string `json:"fileType"`
+	FileSize int64  `json:"fileSize"`
+}
+
+// RequestPayload 请求参数结构体
+type RequestPayload struct {
+	Headers map[string]string `json:"headers"`
+	Body    map[string]any    `json:"body"`
+}

service/cleaner.go

@@ -1,97 +0,0 @@
-package service
-
-import (
-	"context"
-	"model-gateway/model/dto"
-	"os"
-	"time"
-
-	"model-gateway/dao"
-
-	"github.com/gogf/gf/v2/frame/g"
-)
-
-var Cleaner = &cleaner{}
-
-type cleaner struct{}
-
-// RunOnce 由上层定时任务触发：执行一次清理/重试
-func (c *cleaner) RunOnce(ctx context.Context) (res *dto.CleanWorkRes, err error) {
-	// 1) 清理已下载(state=4)且过期的任务（硬删除 + OSS）
-	expired, err := dao.Task.ListExpiredDownloadedGlobal(ctx, 200)
-	if err != nil {
-		g.Log().Errorf(ctx, "[cleaner] list expired(downloaded) error: %v", err)
-	} else {
-		for _, t := range expired {
-			_ = os.Remove(t.TmpFile)
-			_ = dao.Task.HardDeleteByIDGlobal(ctx, t.Id)
-		}
-		g.Log().Infof(ctx, "[cleaner] expired(downloaded) cleaned, count=%d", len(expired))
-	}
-
-	// 2) 超时任务标失败
-	list, err := dao.Task.ListTimeoutTasksGlobal(ctx, 200)
-	if err != nil {
-		g.Log().Errorf(ctx, "[cleaner] list timeout error: %v", err)
-	} else {
-		for _, t := range list {
-			_ = dao.Task.UpdateFailedGlobal(ctx, t.Id, "任务超时自动失败")
-			ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-		}
-		g.Log().Infof(ctx, "[cleaner] timeout cleaned, count=%d", len(list))
-	}
-
-	// 3) 失败(state=3)的任务按模型配置 retry_times 重新入队（放到队尾）
-	retryable, err := dao.Task.ListFailedRetryableGlobal(ctx, 200)
-	if err != nil {
-		g.Log().Errorf(ctx, "[cleaner] list failed retryable error: %v", err)
-	} else {
-		for _, t := range retryable {
-			// 失败任务重新入队（state=3 -> 0）前，先严格占用 queue_limit slot；占用失败则留在失败态，下一轮再尝试
-			// 获取模型配置以得到 queue_limit / expected_seconds
-			m, err := dao.Model.GetByModelNameForTenant(ctx, t.TenantId, t.ModelName)
-			if err != nil || m == nil {
-				continue
-			}
-			limit := GetRuntimeQueueLimit(ctx, t.ModelName, m.QueueLimit)
-			if limit > 0 {
-				ok, _ := AcquireQueueSlot(ctx, t.ModelName, t.TaskID, limit, m.ExpectedSeconds)
-				if !ok {
-					continue
-				}
-			}
-			// retry_queue_max_seconds 控制失败重试的排队策略：
-			// - =0：失败重试插队到队首
-			// - >0：当任务从创建到现在的排队时长 >= maxSeconds，则插队到队首；否则仍放到队尾
-			now := time.Now()
-			enqueueAt := now
-			maxSeconds := t.RetryQueueMaxSeconds
-			if maxSeconds == 0 {
-				enqueueAt = now.Add(-100 * 365 * 24 * time.Hour)
-			} else if maxSeconds > 0 && t.CreatedAt != nil {
-				if now.Sub(t.CreatedAt.Time) >= time.Duration(maxSeconds)*time.Second {
-					enqueueAt = now.Add(-100 * 365 * 24 * time.Hour)
-				}
-			}
-			_ = dao.Task.RequeueForRetryGlobal(ctx, t.Id, enqueueAt)
-		}
-		g.Log().Infof(ctx, "[cleaner] failed retryable cleaned, count=%d", len(retryable))
-	}
-
-	// 4) 超过重试次数仍失败(state=3)的任务：硬删除
-	exhausted, err := dao.Task.ListFailedExhaustedGlobal(ctx, 200)
-	if err != nil {
-		g.Log().Errorf(ctx, "[cleaner] list failed exhausted error: %v", err)
-	} else {
-		for _, t := range exhausted {
-			_ = os.Remove(t.TmpFile)
-			// 重试耗尽硬删除：释放闸门占位（兜底，若此前已释放则幂等）
-			ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-			_ = dao.Task.HardDeleteByIDGlobal(ctx, t.Id)
-		}
-		g.Log().Infof(ctx, "[cleaner] failed exhausted cleaned, count=%d", len(exhausted))
-	}
-	return &dto.CleanWorkRes{
-		Ok: true,
-	}, nil
-}

service/file_detect.go

@@ -1 +0,0 @@
-package service

service/gateway/gateway_http_service.go

@@ -4,18 +4,19 @@ import (
 	"bytes"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"mime/multipart"
 	"model-gateway/common/util"
 	"model-gateway/model/entity"
 	"time"
 
-	commonHttp "gitea.com/red-future/common/http"
+	commonHttp "gitea.redpowerfuture.com/red-future/common/http"
 	"github.com/gogf/gf/v2/frame/g"
 	"github.com/gogf/gf/v2/util/guid"
 )
 
-type uploadFileResponse struct {
+type UploadFileResponse struct {
 	FileURL           string `json:"fileURL"`           // 文件 URL
 	FileSize          int    `json:"fileSize"`          // 文件大小（字节）
 	FileName          string `json:"fileName"`          // 文件名
@@ -23,7 +24,8 @@ type uploadFileResponse struct {
 	FileAddressPrefix string `json:"fileAddressPrefix"` // 文件地址前缀
 }
 
-func UploadByTask(ctx context.Context, _ *entity.AsynchTask, data []byte, fileExt string, _ string) (ossURL string, err error) {
+// UploadByTask 通过任务上传文件
+func UploadByTask(ctx context.Context, data []byte, fileExt string) (oss *UploadFileResponse, err error) {
 	// multipart
 	body := &bytes.Buffer{}
 	writer := multipart.NewWriter(body)
@@ -39,42 +41,51 @@ func UploadByTask(ctx context.Context, _ *entity.AsynchTask, data []byte, fileEx
 	filename := fmt.Sprintf("asynch_%d_%s%s", time.Now().Unix(), guid.S(), ext)
 	part, err := writer.CreateFormFile("file", filename)
 	if err != nil {
-		return "", err
+		return nil, err
 	}
 	if _, err := part.Write(data); err != nil {
-		return "", err
+		return nil, err
 	}
 	contentType := writer.FormDataContentType()
 	if err = writer.Close(); err != nil {
-		return "", err
+		return nil, err
 	}
 
 	headers := util.ForwardHeaders(ctx)
 	headers["Content-Type"] = contentType
-	//fullURL := "oss/file/uploadFile"
 	fullURL := "oss/file/uploadFile"
 	g.Log().Infof(ctx, "[OSS] upload start url=%s filename=%s size=%d", fullURL, filename, len(data))
 
-	var resp uploadFileResponse
+	var resp UploadFileResponse
 	if err = commonHttp.Post(ctx, fullURL, headers, &resp, body.Bytes()); err != nil {
-		return "", err
+		return nil, err
 	}
-	g.Log().Infof(ctx, "[OSS] upload success url=%s size=%d format=%s", resp.FileURL, resp.FileSize, resp.FileFormat)
-	return resp.FileURL, nil
+	if &resp == nil {
+		return nil, errors.New("[OSS] 上传文件失败")
+	}
+	g.Log().Infof(ctx, "[OSS] 上传成功 url=%s size=%d format=%s", resp.FileURL, resp.FileSize, resp.FileFormat)
+	return &resp, nil
 }
 
-// TriggerCallback 任务成功后的回调：
-// - JSON body 参数：task_id/state/oss_file/file_type/text（可选）
-func TriggerCallback(ctx context.Context, t *entity.AsynchTask) {
+// CallbackPayload 回调请求体
+type CallbackPayload struct {
+	TaskId   string `json:"task_id"`
+	State    int    `json:"state"`
+	OssFile  string `json:"oss_file"`
+	FileType string `json:"file_type"`
+	ErrorMsg string `json:"error_msg"`
+}
+
+// TriggerCallback 任务的回调
+func TriggerCallback(ctx context.Context, t *entity.ModelGatewayTask) {
 	headers := util.ForwardHeaders(ctx)
-	var req struct{}
-	payload := map[string]interface{}{
-		"task_id":   t.TaskID,
-		"state":     t.State,
-		"oss_file":  t.OssFile,
-		"file_type": t.FileType,
-		"text":      t.TextResult,
-		"error_msg": t.ErrorMsg,
+	var resp struct{}
+	payload := CallbackPayload{
+		TaskId:   t.TaskID,
+		State:    t.State,
+		OssFile:  t.ResultFile.OssFile,
+		FileType: t.ResultFile.FileType,
+		ErrorMsg: t.ErrorMsg,
 	}
 	jsonData, err := json.Marshal(payload)
 	if err != nil {
@@ -84,7 +95,7 @@ func TriggerCallback(ctx context.Context, t *entity.AsynchTask) {
 	g.Log().Infof(ctx, "[回调] 开始发送 taskId=%s 回调地址=%s 请求头数量=%d 消息体大小=%d字节",
 		t.TaskID, t.CallbackURL, len(headers), len(jsonData))
 
-	err = commonHttp.Post(ctx, t.CallbackURL, headers, &req, jsonData)
+	err = commonHttp.Post(ctx, t.CallbackURL, headers, &resp, jsonData)
 	if err != nil {
 		g.Log().Warningf(ctx, "[回调] 发送失败 taskId=%s 回调地址=%s 错误=%v", t.TaskID, t.CallbackURL, err)
 		return
@@ -92,15 +103,20 @@ func TriggerCallback(ctx context.Context, t *entity.AsynchTask) {
 	g.Log().Infof(ctx, "[回调] 发送成功 taskId=%s 回调地址=%s 消息体大小=%d字节", t.TaskID, t.CallbackURL, len(jsonData))
 }
 
+// PromptsCallbackPayload 提示词回调请求体
+type PromptsCallbackPayload struct {
+	EpicycleId int64          `json:"epicycleId"`
+	Messages   map[string]any `json:"messages"`
+}
+
 // TriggerPromptsCallback 任务成功后的提示词回调
-// - JSON body 参数：epicycleId（轮次id）/textResult（模型回答消息）
-func TriggerPromptsCallback(ctx context.Context, t *entity.AsynchTask, epicycleId int64) {
-	callbackURL := "prompts-core/session/sessionCallback"
+func TriggerPromptsCallback(ctx context.Context, t *entity.ModelGatewayTask, epicycleId int64) {
+	callbackURL := "prompts-core/session/callback"
 	headers := util.ForwardHeaders(ctx)
-	var req struct{}
-	payload := map[string]interface{}{
-		"epicycleId": epicycleId,
-		"text":       t.TextResult,
+	var resp struct{}
+	payload := PromptsCallbackPayload{
+		EpicycleId: epicycleId,
+		Messages:   t.TextResult,
 	}
 	jsonData, err := json.Marshal(payload)
 	if err != nil {
@@ -110,7 +126,7 @@ func TriggerPromptsCallback(ctx context.Context, t *entity.AsynchTask, epicycleI
 	g.Log().Infof(ctx, "[提示词回调] 开始发送 epicycleId=%d 回调地址=%s 请求头数量=%d 消息体大小=%d字节",
 		t.EpicycleId, callbackURL, len(headers), len(jsonData))
 
-	err = commonHttp.Post(ctx, callbackURL, headers, &req, jsonData)
+	err = commonHttp.Post(ctx, callbackURL, headers, &resp, jsonData)
 	if err != nil {
 		g.Log().Warningf(ctx, "[提示词回调] 发送失败 epicycleId=%d 回调地址=%s 错误=%v", t.EpicycleId, callbackURL, err)
 		return

service/model/model_service.go

@@ -0,0 +1,256 @@
+package model
+
+import (
+	"context"
+	"errors"
+	"model-gateway/common/util"
+	"model-gateway/consts/public"
+	"model-gateway/dao"
+	"model-gateway/model/dto"
+	"model-gateway/model/entity"
+	"model-gateway/service/gateway"
+
+	"gitea.redpowerfuture.com/red-future/common/beans"
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
+	"gitea.redpowerfuture.com/red-future/common/utils"
+	"github.com/gogf/gf/v2/database/gdb"
+	"github.com/gogf/gf/v2/frame/g"
+	"github.com/gogf/gf/v2/util/gconv"
+)
+
+var ModelGatewayModels = &modelService{}
+
+type modelService struct{}
+
+// Create 创建模型
+func (s *modelService) Create(ctx context.Context, req *dto.CreateModelReq) (*dto.CreateModelRes, error) {
+	// 1）如果设为会话模型，先把该用户旧会话模型取消
+	if !g.IsEmpty(req.IsChatModel) && *req.IsChatModel == 1 {
+		if err := s.clearUserChatModel(ctx); err != nil {
+			return nil, err
+		}
+	}
+	// 2）判断是否超管，决定 isOwner
+	req.IsOwner = gconv.PtrInt(1)
+	if isAdmin, _ := gateway.IsSuperAdmin(ctx); isAdmin {
+		req.IsOwner = gconv.PtrInt(0)
+	}
+
+	// 3）入库
+	id, err := dao.ModelGatewayModels.Insert(ctx, util.ConvertTo[entity.ModelGatewayModel](req))
+	if err != nil {
+		return nil, err
+	}
+	return &dto.CreateModelRes{ID: id}, nil
+}
+
+// Update 更新模型配置
+func (s *modelService) Update(ctx context.Context, req *dto.UpdateModelReq) error {
+	// 1）会话模型唯一性校验
+	if req.IsChatModel != nil && *req.IsChatModel == 1 {
+		if err := s.checkChatModelUnique(ctx); err != nil {
+			return err
+		}
+	}
+	// 2）超管创建/普通用户更新
+	req.IsOwner = gconv.PtrInt(1)
+	if isAdmin, _ := gateway.IsSuperAdmin(ctx); isAdmin {
+		req.IsOwner = gconv.PtrInt(0)
+		_, err := dao.ModelGatewayModels.Update(ctx, util.ConvertTo[entity.ModelGatewayModel](req))
+		return err
+	}
+	// 3）跨租户判断：超管的模型不允许直接修改，走插入新记录
+	model, err := dao.ModelGatewayModels.GetByAcrossTenant(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO: beans.SQLBaseDO{Id: req.ID},
+	})
+	if err != nil {
+		return err
+	}
+	if model.TenantId == 1 {
+		_, err = dao.ModelGatewayModels.Insert(ctx, util.ConvertTo[entity.ModelGatewayModel](req))
+		return err
+	}
+	_, err = dao.ModelGatewayModels.Update(ctx, util.ConvertTo[entity.ModelGatewayModel](req))
+	return err
+}
+
+// Delete 删除模型
+func (s *modelService) Delete(ctx context.Context, req *dto.DeleteModelReq) error {
+	_, err := dao.ModelGatewayModels.Delete(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO: beans.SQLBaseDO{Id: req.ID},
+	})
+	return err
+}
+
+// Get 获取模型详情
+func (s *modelService) Get(ctx context.Context, req *dto.GetModelReq) (*dto.GetModelRes, error) {
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return nil, err
+	}
+	if g.IsEmpty(req.ID) {
+		req.Creator = user.UserName
+	}
+	model, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO: beans.SQLBaseDO{
+			Id:      req.ID,
+			Creator: user.UserName,
+		},
+		ModelName:   req.ModelName,
+		IsChatModel: req.IsChatModel,
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &dto.GetModelRes{
+		Model: model,
+	}, nil
+}
+
+// List 获取模型列表
+func (s *modelService) List(ctx context.Context, req *dto.ListModelReq) (*dto.ListModelRes, error) {
+	// 1）判断超管
+	req.IsOwner = gconv.PtrInt(1)
+	if isAdmin, _ := gateway.IsSuperAdmin(ctx); isAdmin {
+		req.IsOwner = gconv.PtrInt(0)
+	}
+
+	// 2）获取当前用户
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return nil, err
+	}
+	req.Creator = user.UserName
+
+	// 3）查询
+	models, total, err := dao.ModelGatewayModels.GetByCreatorAndPlatform(ctx, req)
+	if err != nil {
+		return nil, err
+	}
+
+	return &dto.ListModelRes{List: models, Total: total}, nil
+}
+
+// UpdateChatModel 设置会话模型
+func (s *modelService) UpdateChatModel(ctx context.Context, req *dto.UpdateChatModelReq) error {
+	// 1）校验新模型存在
+	newModel, err := dao.ModelGatewayModels.GetByAcrossTenant(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO: beans.SQLBaseDO{Id: req.Id},
+	})
+	if err != nil || newModel == nil {
+		return errors.New("新会话模型不存在")
+	}
+
+	// 2）获取当前用户的会话模型
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return err
+	}
+	currentModel, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO:   beans.SQLBaseDO{Creator: user.UserName},
+		IsChatModel: gconv.PtrInt(1),
+	})
+	if err != nil {
+		return err
+	}
+
+	// 3）事务：取消旧的 + 设置新的
+	return gfdb.DB(ctx).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
+		if !g.IsEmpty(currentModel) {
+			if currentModel.ModelType != public.ModelTypeInference {
+				return errors.New("当前模型为非推理模型，不能设置为会话模型")
+			}
+			if currentModel.Id != req.Id {
+				_, err = dao.ModelGatewayModels.Update(ctx, &entity.ModelGatewayModel{
+					SQLBaseDO:   beans.SQLBaseDO{Id: currentModel.Id},
+					IsChatModel: gconv.PtrInt(0),
+				})
+				if err != nil {
+					return err
+				}
+			}
+		}
+
+		_, err = dao.ModelGatewayModels.Update(ctx, &entity.ModelGatewayModel{
+			SQLBaseDO:   beans.SQLBaseDO{Id: req.Id},
+			IsChatModel: gconv.PtrInt(1),
+		})
+		return err
+	})
+}
+
+// GetIsChatModel 获取当前用户会话模型
+func (s *modelService) GetIsChatModel(ctx context.Context) (*dto.GetIsChatModelRes, error) {
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return nil, err
+	}
+	model, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO:   beans.SQLBaseDO{Creator: user.UserName},
+		IsChatModel: gconv.PtrInt(1),
+	})
+	if err != nil || model == nil {
+		return nil, err
+	}
+	return &dto.GetIsChatModelRes{Model: model}, nil
+}
+
+// ==================== 辅助方法 ====================
+
+// clearUserChatModel 清除当前用户旧会话模型
+func (s *modelService) clearUserChatModel(ctx context.Context) error {
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return err
+	}
+	model, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO:   beans.SQLBaseDO{Creator: user.UserName},
+		IsChatModel: gconv.PtrInt(1),
+	})
+	if err != nil || model == nil {
+		return nil
+	}
+	_, err = dao.ModelGatewayModels.Update(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO:   beans.SQLBaseDO{Id: model.Id},
+		IsChatModel: gconv.PtrInt(0),
+	})
+	return err
+}
+
+// checkChatModelUnique 校验用户是否已有会话模型
+func (s *modelService) checkChatModelUnique(ctx context.Context) error {
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return err
+	}
+	model, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO:   beans.SQLBaseDO{Creator: user.UserName},
+		IsChatModel: gconv.PtrInt(1),
+	})
+	if err != nil {
+		return err
+	}
+	if model != nil {
+		return errors.New("用户已存在会话模型")
+	}
+	return nil
+}
+
+// GetModelTypesFromConfig 从配置文件读取模型类型
+func GetModelTypesFromConfig() (res *dto.TypeItem, err error) {
+	// 返回副本，避免外部修改
+	types := make(map[int]string, len(public.ModelTypeName))
+	for k, v := range public.ModelTypeName {
+		types[k] = v
+	}
+	return &dto.TypeItem{
+		Type: types,
+	}, nil
+}
+
+// GetOperatorList 获取运营商列表
+func GetOperatorList() (res *dto.ListOperatorRes, err error) {
+	return &dto.ListOperatorRes{
+		List: public.OperatorList,
+	}, nil
+}

service/model_invoker.go

@@ -1,417 +0,0 @@
-package service
-
-import (
-	"bytes"
-	"context"
-	"encoding/json"
-	"fmt"
-	"io"
-	"net/http"
-	"net/url"
-	"strings"
-	"time"
-
-	"model-gateway/model/entity"
-
-	"github.com/gogf/gf/v2/container/gvar"
-	"github.com/gogf/gf/v2/frame/g"
-	"github.com/tidwall/gjson"
-	"github.com/tidwall/sjson"
-)
-
-// parseHeadMsgHeaders 支持多个 header 绑定，逗号分隔：
-// 示例：
-// - X-API-Key:qwen3-tts-key,operation:true,count:123
-// - X-API-Key:"qwen3-tts-key",operation:"true"
-//
-// 说明：
-// - HTTP Header 最终都是字符串，这里做的是“值的字符串化表达”。
-// - 若 value 用双引号包裹，会去掉外层引号再注入，便于在配置中区分字符串/布尔/数字等表达（以及避免值中包含特殊字符时歧义）。
-func parseHeadMsgHeaders(headMsg string) map[string]string {
-	headMsg = strings.TrimSpace(headMsg)
-	if headMsg == "" {
-		return nil
-	}
-	out := map[string]string{}
-	parts := strings.Split(headMsg, ",")
-	for _, p := range parts {
-		p = strings.TrimSpace(p)
-		if p == "" {
-			continue
-		}
-		// HeaderName:HeaderValue（推荐） / HeaderName=HeaderValue（兼容）
-		if strings.Contains(p, ":") {
-			kv := strings.SplitN(p, ":", 2)
-			k := strings.TrimSpace(kv[0])
-			v := strings.TrimSpace(kv[1])
-			v = strings.Trim(v, "\"")
-			if k != "" && v != "" {
-				out[k] = v
-			}
-			continue
-		}
-		if strings.Contains(p, "=") {
-			kv := strings.SplitN(p, "=", 2)
-			k := strings.TrimSpace(kv[0])
-			v := strings.TrimSpace(kv[1])
-			v = strings.Trim(v, "\"")
-			if k != "" && v != "" {
-				out[k] = v
-			}
-			continue
-		}
-	}
-	if len(out) == 0 {
-		return nil
-	}
-	return out
-}
-
-func payloadToQuery(payload any) (url.Values, error) {
-	if payload == nil {
-		return url.Values{}, nil
-	}
-	// 统一转成 map[string]any
-	b, err := json.Marshal(payload)
-	if err != nil {
-		return nil, err
-	}
-	m := map[string]any{}
-	if err := json.Unmarshal(b, &m); err != nil {
-		return nil, err
-	}
-	q := url.Values{}
-	for k, v := range m {
-		if v == nil {
-			continue
-		}
-		// 复杂类型直接 json 字符串化
-		switch vv := v.(type) {
-		case string:
-			q.Set(k, vv)
-		case float64, bool, int, int64, uint64:
-			q.Set(k, fmt.Sprintf("%v", vv))
-		default:
-			bs, _ := json.Marshal(v)
-			q.Set(k, string(bs))
-		}
-	}
-	return q, nil
-}
-
-// InvokeModel 调用模型服务，返回二进制结果
-// modelKey 用于覆盖/补充模型配置 head_msg（例如每次请求携带不同的 X-API-Key）。
-func InvokeModel(ctx context.Context, m *entity.AsynchModel, payload any, modelKey string) ([]byte, error) {
-	if m == nil || m.BaseURL == "" {
-		return nil, fmt.Errorf("模型配置不完整")
-	}
-
-	// ============ 新增：请求参数映射 ============
-	mappedPayload, err := mapRequestPayload(m.RequestMapping, payload)
-	if err != nil {
-		return nil, fmt.Errorf("请求参数映射失败: %w", err)
-	}
-
-	url := strings.TrimRight(m.BaseURL, "/")
-	timeout := time.Duration(m.TimeoutSeconds) * time.Second
-	if timeout <= 0 {
-		timeout = 60 * time.Second
-	}
-	client := &http.Client{Timeout: timeout}
-
-	method := strings.ToUpper(strings.TrimSpace(m.HttpMethod))
-	if method == "" {
-		method = http.MethodPost
-	}
-
-	var (
-		req *http.Request
-	)
-	switch method {
-	case http.MethodGet:
-		q, err := payloadToQuery(mappedPayload) // 使用映射后的payload
-		if err != nil {
-			return nil, err
-		}
-		if len(q) > 0 {
-			if strings.Contains(url, "?") {
-				url = url + "&" + q.Encode()
-			} else {
-				url = url + "?" + q.Encode()
-			}
-		}
-		req, err = http.NewRequestWithContext(ctx, http.MethodGet, url, nil)
-	default:
-		bodyBytes, err := json.Marshal(mappedPayload) // 使用映射后的payload
-		if err != nil {
-			return nil, err
-		}
-		req, err = http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(bodyBytes))
-	}
-	if err != nil {
-		return nil, err
-	}
-
-	// 先注入模型配置 head_msg（静态头部，适合公共模型固定 API Key）
-	for hk, hv := range parseHeadMsgHeaders(m.HeadMsg) {
-		req.Header.Set(hk, hv)
-	}
-
-	// 最后注入动态 modelKey（允许覆盖/补充静态 head_msg），适合按请求动态传密钥。
-	for hk, hv := range parseHeadMsgHeaders(modelKey) {
-		req.Header.Set(hk, hv)
-	}
-
-	if method != http.MethodGet {
-		req.Header.Set("Content-Type", "application/json")
-	}
-
-	resp, err := client.Do(req)
-	if err != nil {
-		return nil, err
-	}
-	defer resp.Body.Close()
-
-	b, err := io.ReadAll(resp.Body)
-	if err != nil {
-		return nil, err
-	}
-	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
-		msg := string(b)
-		if len(msg) > 2000 {
-			msg = msg[:2000]
-		}
-		return nil, fmt.Errorf("模型服务返回非2xx: %d, body=%s", resp.StatusCode, msg)
-	}
-
-	// ============ 新增：响应参数映射 ============
-	mappedResponse, err := mapResponsePayload(m.ResponseMapping, b)
-	if err != nil {
-		// 响应映射失败不阻塞，返回原始数据
-		g.Log().Warningf(ctx, "响应参数映射失败: %v，返回原始数据", err)
-		return b, nil
-	}
-	// =========================================
-
-	return mappedResponse, nil
-}
-
-// ============================================
-// 映射相关函数
-// ============================================
-
-// mapRequestPayload 将标准请求映射为模型特定格式
-func mapRequestPayload(mappingAny any, payload any) (any, error) {
-	// 1. 解析请求映射配置（值是any类型，支持bool、number等）
-	mapping, err := parseRequestMapping(mappingAny)
-	if err != nil {
-		return nil, err
-	}
-
-	// 如果没有映射配置，直接返回原始payload
-	if len(mapping) == 0 {
-		return payload, nil
-	}
-
-	// 2. 将payload转为map
-	var payloadMap map[string]any
-	switch v := payload.(type) {
-	case map[string]any:
-		payloadMap = v
-	case []map[string]any:
-		// 如果传进来的是纯messages数组，包装成标准格式
-		payloadMap = map[string]any{
-			"messages": v,
-		}
-	default:
-		// 通过JSON转换
-		jsonBytes, err := json.Marshal(payload)
-		if err != nil {
-			return nil, fmt.Errorf("序列化payload失败: %w", err)
-		}
-		if err := json.Unmarshal(jsonBytes, &payloadMap); err != nil {
-			return nil, fmt.Errorf("反序列化payload失败: %w", err)
-		}
-	}
-
-	// 3. 用数据库固定参数覆盖/补充
-	for key, value := range mapping {
-		if existingValue, exists := payloadMap[key]; !exists || isEmptyValue(existingValue) {
-			payloadMap[key] = value
-		}
-	}
-
-	return payloadMap, nil
-}
-
-// mapResponsePayload 将模型响应映射为标准格式
-func mapResponsePayload(mappingAny any, responseBytes []byte) ([]byte, error) {
-	mapping, err := parseResponseMapping(mappingAny)
-	if err != nil {
-		return nil, err
-	}
-	if len(mapping) == 0 {
-		return responseBytes, nil
-	}
-
-	responseStr := string(responseBytes)
-	resultStr := `{}`
-
-	for standardField, modelPath := range mapping {
-		value := gjson.Get(responseStr, modelPath)
-		if !value.Exists() {
-			continue
-		}
-
-		resultStr, err = sjson.SetRaw(resultStr, standardField, value.Raw)
-		if err != nil {
-			return nil, fmt.Errorf("提取字段 %s <- %s 失败: %w", standardField, modelPath, err)
-		}
-	}
-
-	return []byte(resultStr), nil
-}
-
-func parseRequestMapping(mappingAny any) (map[string]any, error) {
-	if mappingAny == nil {
-		return nil, nil
-	}
-
-	result := make(map[string]any)
-
-	switch v := mappingAny.(type) {
-	case *gvar.Var:
-		if v == nil || v.IsNil() || v.IsEmpty() {
-			return nil, nil
-		}
-		// 尝试转成 map
-		if m := v.Map(); m != nil {
-			for k, val := range m {
-				result[k] = val
-			}
-			return result, nil
-		}
-		// 尝试转成 string
-		if s := v.String(); s != "" && s != "{}" && s != "null" {
-			if err := json.Unmarshal([]byte(s), &result); err != nil {
-				return nil, fmt.Errorf("解析请求映射字符串失败: %w", err)
-			}
-			return result, nil
-		}
-		return nil, nil
-	// =======================================================
-
-	case map[string]interface{}:
-		result = v
-
-	case string:
-		if v == "" || v == "{}" || v == "null" {
-			return nil, nil
-		}
-		if err := json.Unmarshal([]byte(v), &result); err != nil {
-			return nil, fmt.Errorf("解析请求映射字符串失败: %w", err)
-		}
-
-	case []byte:
-		if len(v) == 0 {
-			return nil, nil
-		}
-		if err := json.Unmarshal(v, &result); err != nil {
-			return nil, fmt.Errorf("解析请求映射字节失败: %w", err)
-		}
-
-	default:
-		jsonBytes, err := json.Marshal(mappingAny)
-		if err != nil {
-			return nil, fmt.Errorf("序列化映射配置失败: %w", err)
-		}
-		if err := json.Unmarshal(jsonBytes, &result); err != nil {
-			return nil, fmt.Errorf("解析映射配置失败: %w", err)
-		}
-	}
-
-	return result, nil
-}
-
-// parseResponseMapping 解析响应映射配置
-// 返回值类型为 map[string]string，值都是JSON路径字符串
-func parseResponseMapping(mappingAny any) (map[string]string, error) {
-	if mappingAny == nil {
-		return nil, nil
-	}
-
-	mapping := make(map[string]string)
-
-	switch v := mappingAny.(type) {
-	case *gvar.Var:
-		if v == nil || v.IsNil() || v.IsEmpty() {
-			return nil, nil
-		}
-		if m := v.Map(); m != nil {
-			for k, val := range m {
-				if strVal, ok := val.(string); ok {
-					mapping[k] = strVal
-				}
-			}
-			return mapping, nil
-		}
-		if s := v.String(); s != "" && s != "{}" && s != "null" {
-			if err := json.Unmarshal([]byte(s), &mapping); err != nil {
-				return nil, fmt.Errorf("解析响应映射字符串失败: %w", err)
-			}
-			return mapping, nil
-		}
-		return nil, nil
-	case string:
-		if v == "" || v == "{}" || v == "null" {
-			return nil, nil
-		}
-		if err := json.Unmarshal([]byte(v), &mapping); err != nil {
-			return nil, fmt.Errorf("解析响应映射字符串失败: %w", err)
-		}
-
-	case map[string]interface{}:
-		// 数据库JSONB直接返回的map
-		for k, val := range v {
-			if strVal, ok := val.(string); ok {
-				mapping[k] = strVal
-			}
-		}
-
-	case []byte:
-		if len(v) == 0 {
-			return nil, nil
-		}
-		if err := json.Unmarshal(v, &mapping); err != nil {
-			return nil, fmt.Errorf("解析响应映射字节失败: %w", err)
-		}
-
-	default:
-		jsonBytes, err := json.Marshal(mappingAny)
-		if err != nil {
-			return nil, fmt.Errorf("序列化响应映射配置失败: %w", err)
-		}
-		if err := json.Unmarshal(jsonBytes, &mapping); err != nil {
-			return nil, fmt.Errorf("解析响应映射配置失败: %w", err)
-		}
-	}
-
-	return mapping, nil
-}
-
-// isEmptyValue 判断值是否为空
-func isEmptyValue(v any) bool {
-	if v == nil {
-		return true
-	}
-	switch val := v.(type) {
-	case string:
-		return val == ""
-	case []any:
-		return len(val) == 0
-	case map[string]any:
-		return len(val) == 0
-	default:
-		return false
-	}
-}

service/model_service.go

@@ -1,392 +0,0 @@
-package service
-
-import (
-	"context"
-	"errors"
-	"model-gateway/common/util"
-	"model-gateway/consts/public"
-	"model-gateway/dao"
-	"model-gateway/model/dto"
-	"model-gateway/model/entity"
-	"model-gateway/service/gateway"
-
-	"gitea.com/red-future/common/beans"
-	"gitea.com/red-future/common/db/gfdb"
-	"gitea.com/red-future/common/utils"
-	"github.com/gogf/gf/v2/database/gdb"
-	"github.com/gogf/gf/v2/frame/g"
-	"github.com/gogf/gf/v2/util/gconv"
-)
-
-var Model = &modelService{}
-
-type modelService struct{}
-
-func (s *modelService) Create(ctx context.Context, req *dto.CreateModelReq) (res *dto.CreateModelRes, err error) {
-	// 获取当前会话模型
-	if !g.IsEmpty(req.IsChatModel) && *req.IsChatModel == 1 {
-		var user *beans.User
-		user, err = utils.GetUserInfo(ctx)
-		if err != nil {
-			return nil, err
-		}
-		// 获取当前用户会话模型
-		var model *entity.AsynchModel
-		model, err = dao.Model.Get(ctx, &entity.AsynchModel{
-			SQLBaseDO: beans.SQLBaseDO{
-				Creator: user.UserName,
-			},
-			IsChatModel: new(1),
-		})
-		if err != nil {
-			return nil, err
-		}
-		// 如果有会话模型，那就改变为 0
-		if model != nil {
-			_, err = dao.Model.Update(ctx, &entity.AsynchModel{
-				SQLBaseDO:   beans.SQLBaseDO{Id: model.Id},
-				IsChatModel: gconv.PtrInt(0),
-			})
-			if err != nil {
-				return nil, err
-			}
-		}
-	}
-
-	req.IsOwner = gconv.PtrInt(1)
-	admin, err := gateway.IsSuperAdmin(ctx)
-	if err != nil {
-		return
-	}
-	if admin {
-		req.IsOwner = gconv.PtrInt(0)
-	}
-	id, err := dao.Model.Insert(ctx, &entity.AsynchModel{
-		ModelName:            req.ModelName,
-		ModelType:            req.ModelType,
-		BaseURL:              req.BaseURL,
-		HttpMethod:           req.HttpMethod,
-		HeadMsg:              req.HeadMsg,
-		Form:                 req.Form,
-		RequestMapping:       req.RequestMapping,
-		ResponseMapping:      req.ResponseMapping,
-		ResponseBody:         req.ResponseBody,
-		ResponseTokenField:   req.ResponseTokenField,
-		IsPrivate:            req.IsPrivate,
-		IsChatModel:          req.IsChatModel,
-		ApiKey:               req.ApiKey,
-		Enabled:              req.Enabled,
-		MaxConcurrency:       req.MaxConcurrency,
-		QueueLimit:           req.QueueLimit,
-		TimeoutSeconds:       req.TimeoutSeconds,
-		ExpectedSeconds:      req.ExpectedSeconds,
-		RetryTimes:           req.RetryTimes,
-		RetryQueueMaxSeconds: req.RetryQueueMaxSeconds,
-		AutoCleanSeconds:     req.AutoCleanSeconds,
-		Remark:               req.Remark,
-		IsOwner:              req.IsOwner,
-		OperatorName:         req.OperatorName,
-		TokenConfig:          req.TokenConfig,
-	})
-	if err != nil {
-		return nil, err
-	}
-	return &dto.CreateModelRes{ID: id}, nil
-}
-
-func (s *modelService) Update(ctx context.Context, req *dto.UpdateModelReq) error {
-	//根据当前 isChatModel 来判断是否更新模型
-	if req.IsChatModel == gconv.PtrInt(1) {
-		user, err := utils.GetUserInfo(ctx)
-		if err != nil {
-			return err
-		}
-		// 获取当前用户会话模型
-		model, err := dao.Model.Get(ctx, &entity.AsynchModel{
-			SQLBaseDO: beans.SQLBaseDO{
-				Creator: user.UserName,
-			},
-			IsChatModel: new(1),
-		})
-		if err != nil {
-			return err
-		}
-		if model != nil {
-			return errors.New("用户已存在会话模型，不能创建")
-		}
-	}
-
-	req.IsOwner = gconv.PtrInt(1)
-	admin, err := gateway.IsSuperAdmin(ctx)
-	if err != nil {
-		return err
-	}
-	if admin {
-		req.IsOwner = gconv.PtrInt(0)
-		_, err = dao.Model.Update(ctx, &entity.AsynchModel{
-			SQLBaseDO:            beans.SQLBaseDO{Id: req.ID},
-			ModelName:            req.ModelName,
-			ModelType:            req.ModelType,
-			BaseURL:              req.BaseURL,
-			HttpMethod:           req.HttpMethod,
-			HeadMsg:              req.HeadMsg,
-			Form:                 req.Form,
-			RequestMapping:       req.RequestMapping,
-			ResponseMapping:      req.ResponseMapping,
-			ResponseBody:         req.ResponseBody,
-			ResponseTokenField:   req.ResponseTokenField,
-			IsPrivate:            req.IsPrivate,
-			IsChatModel:          req.IsChatModel,
-			ApiKey:               req.ApiKey,
-			Enabled:              req.Enabled,
-			MaxConcurrency:       req.MaxConcurrency,
-			QueueLimit:           req.QueueLimit,
-			TimeoutSeconds:       req.TimeoutSeconds,
-			ExpectedSeconds:      req.ExpectedSeconds,
-			RetryTimes:           req.RetryTimes,
-			RetryQueueMaxSeconds: req.RetryQueueMaxSeconds,
-			AutoCleanSeconds:     req.AutoCleanSeconds,
-			Remark:               req.Remark,
-			IsOwner:              req.IsOwner,
-			OperatorName:         req.OperatorName,
-			TokenConfig:          req.TokenConfig,
-		})
-		if err != nil {
-			return err
-		}
-		return nil
-	}
-	// 判断当前传过来的模型id的模型是否是超级管理员的。如果是超管的进行创建，否则更新
-	model, err := dao.Model.GetByAcrossTenant(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{Id: req.ID},
-	})
-	if err != nil {
-		return err
-	}
-	if model.TenantId == 1 {
-		insertDto := new(dto.CreateModelReq)
-		err = gconv.Struct(req, insertDto)
-		if err != nil {
-			return err
-		}
-		_, err = dao.Model.Insert(ctx, &entity.AsynchModel{
-			ModelName:            req.ModelName,
-			ModelType:            req.ModelType,
-			BaseURL:              req.BaseURL,
-			HttpMethod:           req.HttpMethod,
-			HeadMsg:              req.HeadMsg,
-			Form:                 req.Form,
-			RequestMapping:       req.RequestMapping,
-			ResponseMapping:      req.ResponseMapping,
-			ResponseBody:         req.ResponseBody,
-			ResponseTokenField:   req.ResponseTokenField,
-			IsPrivate:            req.IsPrivate,
-			IsChatModel:          req.IsChatModel,
-			ApiKey:               req.ApiKey,
-			Enabled:              req.Enabled,
-			MaxConcurrency:       req.MaxConcurrency,
-			QueueLimit:           req.QueueLimit,
-			TimeoutSeconds:       req.TimeoutSeconds,
-			ExpectedSeconds:      req.ExpectedSeconds,
-			RetryTimes:           req.RetryTimes,
-			RetryQueueMaxSeconds: req.RetryQueueMaxSeconds,
-			AutoCleanSeconds:     req.AutoCleanSeconds,
-			Remark:               req.Remark,
-			IsOwner:              req.IsOwner,
-			OperatorName:         req.OperatorName,
-			TokenConfig:          req.TokenConfig,
-		})
-		return err
-	}
-	_, err = dao.Model.Update(ctx, &entity.AsynchModel{
-		SQLBaseDO:            beans.SQLBaseDO{Id: req.ID},
-		ModelName:            req.ModelName,
-		ModelType:            req.ModelType,
-		BaseURL:              req.BaseURL,
-		HttpMethod:           req.HttpMethod,
-		HeadMsg:              req.HeadMsg,
-		Form:                 req.Form,
-		RequestMapping:       req.RequestMapping,
-		ResponseMapping:      req.ResponseMapping,
-		ResponseBody:         req.ResponseBody,
-		ResponseTokenField:   req.ResponseTokenField,
-		IsPrivate:            req.IsPrivate,
-		IsChatModel:          req.IsChatModel,
-		ApiKey:               req.ApiKey,
-		Enabled:              req.Enabled,
-		MaxConcurrency:       req.MaxConcurrency,
-		QueueLimit:           req.QueueLimit,
-		TimeoutSeconds:       req.TimeoutSeconds,
-		ExpectedSeconds:      req.ExpectedSeconds,
-		RetryTimes:           req.RetryTimes,
-		RetryQueueMaxSeconds: req.RetryQueueMaxSeconds,
-		AutoCleanSeconds:     req.AutoCleanSeconds,
-		Remark:               req.Remark,
-		IsOwner:              req.IsOwner,
-		OperatorName:         req.OperatorName,
-		TokenConfig:          req.TokenConfig,
-	})
-	return err
-}
-
-func (s *modelService) Delete(ctx context.Context, req *dto.DeleteModelReq) error {
-	_, err := dao.Model.Delete(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{Id: req.ID},
-	})
-	return err
-}
-
-func (s *modelService) Get(ctx context.Context, req *dto.GetModelReq) (*dto.GetModelRes, error) {
-	model, err := dao.Model.Get(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{Id: req.ID},
-	})
-	if err != nil {
-		return nil, err
-	}
-	model.Form = util.ParseJSONField(model.Form)
-	model.RequestMapping = util.ParseJSONField(model.RequestMapping)
-	model.ResponseMapping = util.ParseJSONField(model.ResponseMapping)
-	model.ResponseBody = util.ParseJSONField(model.ResponseBody)
-	model.TokenConfig = util.ParseJSONField(model.TokenConfig)
-	return &dto.GetModelRes{
-		Model: model,
-	}, nil
-}
-
-func (s *modelService) List(ctx context.Context, req *dto.ListModelReq) (res *dto.ListModelRes, err error) {
-	var models []*entity.AsynchModel
-
-	req.IsOwner = gconv.PtrInt(1)
-	admin, err := gateway.IsSuperAdmin(ctx)
-	if err != nil {
-		return
-	}
-	if admin {
-		req.IsOwner = gconv.PtrInt(0)
-	}
-
-	var user *beans.User
-	user, err = utils.GetUserInfo(ctx)
-	if err != nil {
-		return nil, err
-	}
-	req.Creator = user.UserName
-
-	models, total, err := dao.Model.GetByCreatorAndPlatform(ctx, req)
-	if err != nil {
-		return
-	}
-
-	// 处理列表中每条记录的 JSONB 字段
-	for _, m := range models {
-		m.Form = util.ParseJSONField(m.Form)
-		m.RequestMapping = util.ParseJSONField(m.RequestMapping)
-		m.ResponseMapping = util.ParseJSONField(m.ResponseMapping)
-		m.ResponseBody = util.ParseJSONField(m.ResponseBody)
-		m.TokenConfig = util.ParseJSONField(m.TokenConfig)
-	}
-	return &dto.ListModelRes{
-		List:  models,
-		Total: total,
-	}, nil
-}
-
-// GetModelTypesFromConfig 从配置文件读取模型类型
-func GetModelTypesFromConfig() (res *dto.TypeItem, err error) {
-	// 返回副本，避免外部修改
-	types := make(map[int]string, len(public.ModelTypeName))
-	for k, v := range public.ModelTypeName {
-		types[k] = v
-	}
-	return &dto.TypeItem{
-		Type: types,
-	}, nil
-}
-
-// GetOperatorList 获取运营商列表
-func GetOperatorList() (res *dto.ListOperatorRes, err error) {
-	return &dto.ListOperatorRes{
-		List: public.OperatorList,
-	}, nil
-}
-
-func (s *modelService) UpdateChatModel(ctx context.Context, req *dto.UpdateChatModelReq) error {
-	// 校验新会话模型是否存在
-	newModel, err := dao.Model.GetByAcrossTenant(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{Id: req.Id},
-	})
-	if err != nil {
-		return err
-	}
-	if newModel == nil {
-		return errors.New("新会话模型不存在")
-	}
-	var user *beans.User
-	user, err = utils.GetUserInfo(ctx)
-	if err != nil {
-		return err
-	}
-	// 获取当前用户会话模型
-	currentModel, err := dao.Model.Get(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{
-			Creator: user.UserName,
-		},
-		IsChatModel: new(1),
-	})
-	if err != nil {
-		return err
-	}
-	err = gfdb.DB(ctx).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
-		if !g.IsEmpty(currentModel) {
-			if currentModel.ModelType != public.ModelTypeInference {
-				return errors.New("当前模型为非推理模型,不能设置为会话模型")
-			}
-
-			// 如果点击的就是当前会话模型（已经是1），取消它（设为0）
-			if currentModel.Id != req.Id {
-				_, err = dao.Model.Update(ctx, &entity.AsynchModel{
-					SQLBaseDO:   beans.SQLBaseDO{Id: currentModel.Id},
-					IsChatModel: gconv.PtrInt(0),
-				})
-				if err != nil {
-					return err
-				}
-			}
-		}
-
-		// 设置当前为会话模型（设为1）
-		_, err = dao.Model.Update(ctx, &entity.AsynchModel{
-			SQLBaseDO:   beans.SQLBaseDO{Id: req.Id},
-			IsChatModel: gconv.PtrInt(1),
-		})
-		return err
-	})
-	return err
-}
-
-func (s *modelService) GetIsChatModel(ctx context.Context) (*dto.GetIsChatModelRes, error) {
-	user, err := utils.GetUserInfo(ctx)
-	if err != nil {
-		return nil, err
-	}
-	model, err := dao.Model.Get(ctx, &entity.AsynchModel{
-		SQLBaseDO: beans.SQLBaseDO{
-			Creator: user.UserName,
-		},
-		IsChatModel: new(1),
-	})
-	if err != nil {
-		return nil, err
-	}
-	if model == nil {
-		return nil, nil
-	}
-	model.Form = util.ParseJSONField(model.Form)
-	model.RequestMapping = util.ParseJSONField(model.RequestMapping)
-	model.ResponseMapping = util.ParseJSONField(model.ResponseMapping)
-	model.ResponseBody = util.ParseJSONField(model.ResponseBody)
-	model.TokenConfig = util.ParseJSONField(model.TokenConfig)
-	return &dto.GetIsChatModelRes{
-		Model: model,
-	}, nil
-}

service/queue/auto_tune.go

@@ -1,4 +1,4 @@
-package service
+package queue
 
 import (
 	"context"
@@ -10,7 +10,7 @@ import (
 	"model-gateway/consts/public"
 	"model-gateway/model/entity"
 
-	"gitea.com/red-future/common/db/gfdb"
+	"gitea.redpowerfuture.com/red-future/common/db/gfdb"
 	"github.com/gogf/gf/v2/frame/g"
 )
 
@@ -28,7 +28,6 @@ type AutoTuneResult struct {
 	OldQueueLimit int `json:"oldQueueLimit"` // 调参前运行时值（Redis），若无则等于 cap
 	NewQueueLimit int `json:"newQueueLimit"` // 本次计算出的运行时值（将写入 Redis），受 ±50% 约束且不超过 cap
 
-	ExpectedSeconds int `json:"expectedSeconds"` // 模型预计执行时间（秒）：asynch_models.expected_seconds（用于 queue_limit 计算绑定）
 }
 
 // AutoTune 由上层定时任务通过接口触发：
@@ -44,14 +43,14 @@ func AutoTune(ctx context.Context, req *dto.AutoTuneReq) (res *dto.AutoTuneRes,
 		req.WindowSeconds = 3600 // 默认1小时
 	}
 	// 1) 读取模型配置（cap），按 model_name 聚合去重（如果表里有多租户重复数据，取较大上限）
-	var modelRows []*entity.AsynchModel
+	var modelRows []*entity.ModelGatewayModel
 	if err := gfdb.DB(ctx).Model(ctx, public.TableNameModel).
 		Where("deleted_at IS NULL").
-		Where(entity.AsynchModelCol.Enabled, 1).
+		Where(entity.ModelGatewayModelCol.Enabled, 1).
 		Scan(&modelRows); err != nil {
 		return nil, err
 	}
-	modelMap := make(map[string]*entity.AsynchModel)
+	modelMap := make(map[string]*entity.ModelGatewayModel)
 	for _, m := range modelRows {
 		if m == nil || m.ModelName == "" {
 			continue
@@ -65,11 +64,11 @@ func AutoTune(ctx context.Context, req *dto.AutoTuneReq) (res *dto.AutoTuneRes,
 		if m.MaxConcurrency > cur.MaxConcurrency {
 			cur.MaxConcurrency = m.MaxConcurrency
 		}
-		if m.QueueLimit > cur.QueueLimit {
-			cur.QueueLimit = m.QueueLimit
+		if m.MaxConcurrency*2 > cur.MaxConcurrency*2 {
+			cur.MaxConcurrency = m.MaxConcurrency
 		}
-		if m.ExpectedSeconds > cur.ExpectedSeconds {
-			cur.ExpectedSeconds = m.ExpectedSeconds
+		if m.TimeoutSeconds > cur.TimeoutSeconds {
+			cur.TimeoutSeconds = m.TimeoutSeconds
 		}
 	}
 	if len(modelMap) == 0 {
@@ -113,7 +112,7 @@ SELECT model_name,
 	for modelName, m := range modelMap {
 		s := statMap[modelName]
 		capMax := m.MaxConcurrency
-		capQueue := m.QueueLimit
+		capQueue := m.MaxConcurrency * 2
 		oldMax := GetRuntimeMaxConcurrency(ctx, modelName, capMax)
 		oldQueue := GetRuntimeQueueLimit(ctx, modelName, capQueue)
 
@@ -129,7 +128,6 @@ SELECT model_name,
 				CapQueueLimit:     capQueue,
 				OldQueueLimit:     oldQueue,
 				NewQueueLimit:     oldQueue,
-				ExpectedSeconds:   m.ExpectedSeconds,
 			})
 			continue
 		}
@@ -155,7 +153,7 @@ SELECT model_name,
 		setRuntimeInt(ctx, runtimeMaxConcurrencyKey(modelName), newMax)
 
 		// queue_limit：W_target = expected_seconds * queueFactor
-		exp := m.ExpectedSeconds
+		exp := m.TimeoutSeconds
 		if exp <= 0 {
 			exp = 60
 		}
@@ -190,7 +188,6 @@ SELECT model_name,
 			CapQueueLimit:     capQueue,
 			OldQueueLimit:     oldQueue,
 			NewQueueLimit:     newQueue,
-			ExpectedSeconds:   m.ExpectedSeconds,
 		})
 	}
 

service/queue/queue_gate.go

@@ -1,4 +1,4 @@
-package service
+package queue
 
 import (
 	"context"

service/queue/runtime_tune.go

@@ -1,4 +1,4 @@
-package service
+package queue
 
 import (
 	"context"
@@ -11,9 +11,9 @@ import (
 // 上层每小时调用 /model/autoTune 写入运行时值；Worker/CreateTask 读取运行时值生效。
 
 const (
-	runtimeMaxCKeyPrefix   = "asynch:runtime:max_concurrency:" // + model_name
-	runtimeQueueKeyPrefix  = "asynch:runtime:queue_limit:"     // + model_name
-	runtimeTTLSeconds      = 2 * 3600                           // 2小时，避免一次调参失败导致立即回退
+	runtimeMaxCKeyPrefix  = "asynch:runtime:max_concurrency:" // + model_name
+	runtimeQueueKeyPrefix = "asynch:runtime:queue_limit:"     // + model_name
+	runtimeTTLSeconds     = 2 * 3600                          // 2小时，避免一次调参失败导致立即回退
 )
 
 func runtimeMaxConcurrencyKey(modelName string) string {
@@ -80,4 +80,3 @@ func clampInt(v, minV, maxV int) int {
 	}
 	return v
 }
-

service/queue/semaphore.go

@@ -1,4 +1,4 @@
-package service
+package queue
 
 import (
 	"context"
@@ -34,7 +34,8 @@ end
 return 1
 `
 
-func acquireSemaphore(ctx context.Context, key string, max int, ttlSeconds int64) (bool, error) {
+// AcquireSemaphore 获取并发令牌
+func AcquireSemaphore(ctx context.Context, key string, max int, ttlSeconds int64) (bool, error) {
 	if max <= 0 {
 		// 不限制
 		return true, nil
@@ -49,8 +50,8 @@ func acquireSemaphore(ctx context.Context, key string, max int, ttlSeconds int64
 	return gconv.Int(r) == 1, nil
 }
 
-func releaseSemaphore(ctx context.Context, key string) error {
+// ReleaseSemaphore 释放并发令牌
+func ReleaseSemaphore(ctx context.Context, key string) error {
 	_, err := g.Redis().Do(ctx, "EVAL", releaseLua, 1, key)
 	return err
 }
-

service/stat/stat_service.go

@@ -0,0 +1,34 @@
+package stat
+
+import (
+	"context"
+	"model-gateway/model/entity"
+
+	"model-gateway/dao"
+	"model-gateway/model/dto"
+)
+
+var ModelGatewayLogsStat = &logsStatService{}
+
+type logsStatService struct{}
+
+func (s *logsStatService) List(ctx context.Context, req *dto.ListModelStatReq) (*dto.ListModelStatRes, error) {
+	if req == nil {
+		req = &dto.ListModelStatReq{}
+	}
+	if req.PageNum <= 0 {
+		req.PageNum = 1
+	}
+	if req.PageSize <= 0 {
+		req.PageSize = 10
+	}
+
+	list, total, err := dao.ModelGatewayLogsStat.List(ctx, req.PageNum, req.PageSize, &entity.ModelGatewayLogsStat{
+		Creator:   req.Creator,
+		ModelName: req.ModelName,
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &dto.ListModelStatRes{List: list, Total: total}, nil
+}

service/stat_service.go

@@ -1,39 +0,0 @@
-package service
-
-import (
-	"context"
-
-	"model-gateway/dao"
-	"model-gateway/model/dto"
-)
-
-type statService struct{}
-
-var Stat = &statService{}
-
-func (s *statService) List(ctx context.Context, req *dto.ListModelStatReq) (res *dto.ListModelStatRes, err error) {
-	pageNum, pageSize := 1, 10
-	if req != nil {
-		if req.PageNum > 0 {
-			pageNum = req.PageNum
-		}
-		if req.PageSize > 0 {
-			pageSize = req.PageSize
-		}
-	}
-	startDay, endDay := "", ""
-	var tenantID *int64
-	creator, modelName := "", ""
-	if req != nil {
-		startDay = req.StartDay
-		endDay = req.EndDay
-		tenantID = req.TenantID
-		creator = req.Creator
-		modelName = req.ModelName
-	}
-	list, total, err := dao.Stat.List(ctx, pageNum, pageSize, startDay, endDay, tenantID, creator, modelName)
-	if err != nil {
-		return nil, err
-	}
-	return &dto.ListModelStatRes{List: list, Total: total}, nil
-}

service/task/task_service.go

@@ -0,0 +1,283 @@
+package task
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"model-gateway/common/util"
+	"model-gateway/consts/public"
+	"model-gateway/service/queue"
+	"time"
+
+	"model-gateway/dao"
+	"model-gateway/model/dto"
+	"model-gateway/model/entity"
+
+	"gitea.redpowerfuture.com/red-future/common/beans"
+	"gitea.redpowerfuture.com/red-future/common/utils"
+	"github.com/gogf/gf/v2/database/gdb"
+	"github.com/gogf/gf/v2/frame/g"
+	"github.com/gogf/gf/v2/util/gconv"
+	"github.com/google/uuid"
+)
+
+var ModelGatewayTask = &taskService{}
+
+type taskService struct{}
+
+// Create 创建任务
+func (s *taskService) Create(ctx context.Context, req *dto.CreateTaskReq) (res *dto.CreateTaskRes, err error) {
+	taskID := uuid.NewString()
+
+	// 1) 检查模型配置，并且获取模型
+	userInfo, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return nil, err
+	}
+	model, err := dao.ModelGatewayModels.Get(ctx, &entity.ModelGatewayModel{
+		SQLBaseDO: beans.SQLBaseDO{
+			TenantId: userInfo.TenantId,
+			Creator:  userInfo.UserName,
+		},
+		ModelName: req.ModelName,
+	})
+	if err != nil {
+		return nil, err
+	}
+	if model == nil || (model.Enabled != nil && *model.Enabled != 1) {
+		return nil, errors.New("模型不存在或未启用")
+	}
+
+	// 2) 排队上限（严格控制：Redis 原子闸门）
+	limit := queue.GetRuntimeQueueLimit(ctx, req.ModelName, model.MaxConcurrency*2)
+	if limit > 0 {
+		ok, err := queue.AcquireQueueSlot(ctx, req.ModelName, taskID, limit, model.TimeoutSeconds)
+		if err != nil {
+			return nil, err
+		}
+		if !ok {
+			return nil, errors.New("任务排队已满，请稍后再试")
+		}
+	}
+
+	// 3) 插入任务记录
+	requestPayload := entity.RequestPayload{
+		Body:    req.RequestPayload,
+		Headers: util.ParseHeadMsgHeaders(model.HeadMsg),
+	}
+	id, err := dao.ModelGatewayTask.Insert(ctx, &entity.ModelGatewayTask{
+		ModelName:      req.ModelName,
+		TaskID:         taskID,
+		State:          public.TaskStatusPending,
+		BizName:        req.BizName,
+		CallbackURL:    req.CallbackUrl,
+		RequestPayload: &requestPayload,
+		EpicycleId:     req.EpicycleId,
+	})
+	if err != nil { // 入库失败：回滚闸门占位
+		queue.ReleaseQueueSlot(ctx, req.ModelName, taskID)
+		return nil, err
+	}
+
+	// 4) 写操作日志（不影响主流程，失败忽略）
+	ip := ""
+	ua := ""
+	apiPath := "/task/createTask"
+	httpMethod := "POST"
+	if r := g.RequestFromCtx(ctx); r != nil {
+		ip = utils.GetLocalIP()
+		ua = r.UserAgent()
+		apiPath = r.URL.Path
+		httpMethod = r.Method
+	}
+	_, _ = dao.ModelGatewayLogsOp.Insert(ctx, &entity.ModelGatewayLogsOp{
+		IP:             ip,
+		UserAgent:      ua,
+		APIPath:        apiPath,
+		HttpMethod:     httpMethod,
+		BizName:        req.BizName,
+		ModelName:      req.ModelName,
+		TaskID:         taskID,
+		OpType:         "createTask",
+		Success:        1,
+		CostMs:         time.Since(time.Now()).Milliseconds(),
+		RequestPayload: &requestPayload,
+		ResponsePayload: gdb.Map{
+			"taskId": taskID,
+		},
+	})
+
+	// 5) 获取任务信息
+	task, err := dao.ModelGatewayTask.ClaimByID(ctx, id)
+	if err != nil {
+		return nil, err
+	}
+
+	// 5) 创建成功后立即异步尝试执行当前任务
+	go AsyncWorker.handleOne(util.AsyncCtx(ctx), task, model, req)
+
+	return &dto.CreateTaskRes{TaskID: taskID}, nil
+}
+
+// GetResult 获取任务结果
+func (s *taskService) GetResult(ctx context.Context, taskID string) (res *dto.GetTaskResultRes, err error) {
+	t, err := dao.ModelGatewayTask.Get(ctx, &entity.ModelGatewayTask{
+		TaskID: taskID,
+	})
+	if err != nil {
+		return nil, err
+	}
+	if t == nil {
+		return nil, errors.New("任务不存在")
+	}
+	return &dto.GetTaskResultRes{
+		OssFile: t.ResultFile.OssFile,
+		State:   t.State,
+	}, nil
+}
+
+// GetBatch 批量查询任务；将成功(state=2)的任务更新为已下载(state=4)，并写入过期时间
+func (s *taskService) GetBatch(ctx context.Context, req *dto.GetTaskBatchReq) (res *dto.GetTaskBatchRes, err error) {
+	if req == nil || len(req.TaskIDs) == 0 {
+		return &dto.GetTaskBatchRes{List: []dto.GetTaskBatchItem{}}, nil
+	}
+	// 1) 先查当前租户下的任务列表
+	list, err := dao.ModelGatewayTask.ListByTaskIDs(ctx, req.TaskIDs)
+	if err != nil {
+		return nil, err
+	}
+
+	// 2) 对成功(state=2)的任务：标记为已下载(state=4)
+	for _, t := range list {
+		if t == nil {
+			continue
+		}
+		if t.State != public.BuildTypeNode {
+			continue
+		}
+		_ = dao.ModelGatewayTask.MarkDownloadedByID(ctx, t.Id)
+
+		// 为了本次返回一致性，内存里也更新
+		t.State = public.TaskStatusDownloaded
+	}
+
+	// 3) 组装返回
+	items := make([]dto.GetTaskBatchItem, 0, len(list))
+	for _, t := range list {
+		if t == nil {
+			continue
+		}
+		items = append(items, dto.GetTaskBatchItem{
+			TaskID:     t.TaskID,
+			State:      t.State,
+			OssFile:    t.ResultFile.OssFile,
+			TextResult: t.TextResult,
+		})
+	}
+	return &dto.GetTaskBatchRes{List: items}, nil
+}
+
+// List 获取任务列表
+func (s *taskService) List(ctx context.Context, req *dto.ListTaskReq) (*dto.ListTaskRes, error) {
+	if req.PageNum <= 0 {
+		req.PageNum = 1
+	}
+	if req.PageSize <= 0 {
+		req.PageSize = 10
+	}
+	user, err := utils.GetUserInfo(ctx)
+	if err != nil {
+		return nil, err
+	}
+	list, total, err := dao.ModelGatewayTask.List(ctx, req.PageNum, req.PageSize, &entity.ModelGatewayTask{
+		SQLBaseDO: beans.SQLBaseDO{
+			Creator: user.UserName,
+		},
+		ModelName: req.ModelName,
+		BizName:   req.BizName,
+		State:     req.State,
+		TaskID:    req.TaskID,
+	})
+	if err != nil {
+		return nil, err
+	}
+	return &dto.ListTaskRes{List: list, Total: total}, nil
+}
+
+// ModelTaskCallback 模型异步任务的回调通知
+func (s *taskService) ModelTaskCallback(ctx context.Context, req *dto.ModelTaskCallbackReq) (*dto.ModelTaskCallbackRes, error) {
+	g.Log().Infof(ctx, "[模型回调] 收到通知 taskID=%s status=%s", req.TaskID, req.Status)
+	// 1. 查本地任务
+	task, err := dao.ModelGatewayTask.Get(ctx, &entity.ModelGatewayTask{
+		TaskID: req.TaskID,
+	})
+	if err != nil || task == nil {
+		return nil, fmt.Errorf("任务不存在: %s", req.TaskID)
+	}
+
+	// 2. 成功：取 video_url 和 usage
+	if req.Status == "succeeded" {
+		result := map[string]any{
+			"video_url": req.Content["video_url"],
+			"usage":     req.Usage,
+		}
+		NotifyAsyncResult(req.TaskID, result, nil)
+		return &dto.ModelTaskCallbackRes{Success: true}, nil
+	}
+
+	// 3. 失败/过期
+	if req.Status == "failed" || req.Status == "expired" {
+		NotifyAsyncResult(req.TaskID, nil, fmt.Errorf(req.Status))
+		return &dto.ModelTaskCallbackRes{Success: true}, nil
+	}
+
+	return &dto.ModelTaskCallbackRes{Success: true}, nil
+}
+
+// QueryPendingTasks 批量轮询进行中的异步任务
+func (s *taskService) QueryPendingTasks(ctx context.Context, req *dto.QueryPendingTasksReq) (*dto.QueryPendingTasksRes, error) {
+	limit := req.Limit
+	if limit <= 0 {
+		limit = g.Cfg().MustGet(ctx, "asynch.queryPending.limit", 10).Int()
+	}
+
+	// 1. 查 state=1（执行中）的异步任务
+	tasks, err := dao.ModelGatewayTask.GetPendingAsyncTasks(ctx, limit)
+	if err != nil {
+		return nil, err
+	}
+
+	// 2. 逐个查询
+	var results []dto.QueryTaskItem
+	for _, t := range tasks {
+		// 拿到模型配置
+		model, err := dao.ModelGatewayModels.GetByModelNameForTenant(ctx, t.TenantId, t.ModelName)
+		if err != nil || model == nil || model.QueryConfig == nil {
+			continue
+		}
+		result, err := util.PullTaskResult(ctx, nil, model.QueryConfig, model.HeadMsg)
+		if err != nil {
+			g.Log().Warningf(ctx, "[轮询] 查询失败 taskID=%s err=%v", t.TaskID, err)
+			continue
+		}
+
+		status := gconv.String(result["status"])
+		item := dto.QueryTaskItem{
+			TaskID:  t.TaskID,
+			Status:  status,
+			Content: result["content"].(map[string]any),
+			Usage:   result["usage"].(map[string]any),
+		}
+		results = append(results, item)
+
+		// 如果任务完成，通知等待通道
+		if status == "succeeded" || status == "failed" || status == "expired" {
+			NotifyAsyncResult(t.TaskID, result["content"].(map[string]any), nil)
+		}
+	}
+
+	return &dto.QueryPendingTasksRes{
+		Total:   len(results),
+		Results: results,
+	}, nil
+}

service/task/worker.go

@@ -0,0 +1,501 @@
+package task
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"os"
+	"strings"
+	"sync"
+	"time"
+	"unicode/utf8"
+
+	"model-gateway/common/util"
+	"model-gateway/consts/public"
+	"model-gateway/dao"
+	"model-gateway/model/dto"
+	"model-gateway/model/entity"
+	"model-gateway/service/gateway"
+	"model-gateway/service/queue"
+
+	"gitea.redpowerfuture.com/red-future/common/beans"
+	"github.com/gogf/gf/v2/encoding/gjson"
+	"github.com/gogf/gf/v2/frame/g"
+	"github.com/gogf/gf/v2/util/gconv"
+)
+
+var AsyncWorker = &asyncWorker{}
+
+type asyncWorker struct {
+}
+
+// handleOne 执行一次完整的任务
+func (w *asyncWorker) handleOne(ctx context.Context, task *entity.ModelGatewayTask, model *entity.ModelGatewayModel, req *dto.CreateTaskReq) {
+	var (
+		body      = task.RequestPayload.Body
+		maxRetry  = model.RetryTimes
+		startTime = time.Now()
+		result    map[string]any
+		err       error
+	)
+	g.Log().Infof(ctx, "[执行任务][开始] taskId=%s model=%s", task.TaskID, task.ModelName)
+
+	// ============================================
+	// 1) 分布式并发控制
+	// ============================================
+	semKey := fmt.Sprintf("asynch:sem:%s", task.ModelName)
+	maxC := queue.GetRuntimeMaxConcurrency(ctx, task.ModelName, model.MaxConcurrency)
+	acquired, err := queue.AcquireSemaphore(ctx, semKey, maxC, 3600)
+	if err != nil {
+		task.DurationSeconds = int64(time.Since(startTime).Seconds())
+		w.failTask(ctx, task, startTime, err.Error())
+		return
+	}
+	if !acquired {
+		_, _ = dao.ModelGatewayTask.Update(ctx, &entity.ModelGatewayTask{
+			SQLBaseDO: beans.SQLBaseDO{Id: task.Id},
+			State:     public.TaskStatusPending,
+		})
+		g.Log().Infof(ctx, "[执行任务][排队] 并发已满，放回队列 taskId=%s", task.TaskID)
+		return
+	}
+	defer func() { _ = queue.ReleaseSemaphore(ctx, semKey) }()
+
+	// ============================================
+	// 2) 调用模型
+	// ============================================
+	switch {
+	case model.CallMode != nil && *model.CallMode == public.CallModeStream:
+		rawBytes, streamErr := w.callModelStream(ctx, task, model, body)
+		if streamErr != nil {
+			w.failTask(ctx, task, startTime, streamErr.Error())
+			return
+		}
+		result, err = util.ParseStreamResponse(rawBytes, model.StreamConfig)
+	case model.CallMode != nil && *model.CallMode == public.CallModeAsync:
+		result, err = w.callModel(ctx, task, model, body)
+		if err == nil {
+			result, err = util.PullTaskResult(ctx, result, model.QueryConfig, model.HeadMsg)
+		}
+	default:
+		result, err = w.callModel(ctx, task, model, body)
+	}
+	if err != nil {
+		w.failTask(ctx, task, startTime, err.Error())
+		return
+	}
+
+	// ============================================
+	// 3) 缓存临时文件
+	// ============================================
+	if tmpPath, tmpErr := util.SaveTempFileByType(task.TaskID, result, task.TmpFile); tmpErr == nil && tmpPath != "" {
+		task.TmpFile = tmpPath
+		task.Phase = 1
+		_, _ = dao.ModelGatewayTask.Update(ctx, task)
+	}
+
+	// ============================================
+	// 4) 解析校验 + 响应映射（可重试）
+	// ============================================
+	result, err = w.parseAndRetry(ctx, result, task, model, req, maxRetry, startTime)
+	if err != nil {
+		task.TextResult = result
+		w.failTask(ctx, task, startTime, err.Error())
+		return
+	}
+
+	// ============================================
+	// 5) 上传 OSS（可重试）
+	// ============================================
+	var oss *gateway.UploadFileResponse
+	for attempt := 0; attempt <= maxRetry; attempt++ {
+		if attempt > 0 {
+			g.Log().Infof(ctx, "[执行任务][重试] OSS上传 第%d/%d次 taskId=%s", attempt, maxRetry, task.TaskID)
+		}
+		oss, err = gateway.UploadByTask(ctx, gjson.New(result).MustToJson(), "json")
+		if err == nil {
+			break
+		}
+		g.Log().Errorf(ctx, "[执行任务][失败] OSS上传失败 taskId=%s attempt=%d/%d err=%v", task.TaskID, attempt, maxRetry, err)
+		if attempt == maxRetry {
+			task.State = public.TaskStatusFailed
+			task.ErrorMsg = err.Error()
+			task.Phase = 1
+			_, _ = dao.ModelGatewayTask.Update(ctx, task)
+			w.failTask(ctx, task, startTime, fmt.Sprintf("OSS上传重试耗尽: %v", err))
+			return
+		}
+	}
+
+	// ============================================
+	// 6) 成功收尾
+	// ============================================
+	task.State = public.TaskStatusSuccess
+	task.DurationSeconds = int64(time.Since(startTime).Seconds())
+	task.ResultFile = &entity.ResultFile{
+		OssFile:  oss.FileAddressPrefix + oss.FileURL,
+		FileType: oss.FileFormat,
+		FileSize: int64(oss.FileSize),
+	}
+	task.TextResult = result
+	if _, err = dao.ModelGatewayTask.Update(ctx, task); err != nil {
+		g.Log().Errorf(ctx, "[执行任务][失败] 更新数据库失败 taskId=%s err=%v", task.TaskID, err)
+		return
+	}
+
+	queue.ReleaseQueueSlot(ctx, task.ModelName, task.TaskID)
+	go gateway.TriggerCallback(context.WithoutCancel(ctx), task)
+	if req.EpicycleId != 0 {
+		go gateway.TriggerPromptsCallback(context.WithoutCancel(ctx), task, req.EpicycleId)
+	}
+
+	g.Log().Infof(ctx, "[执行任务][成功] taskId=%s duration=%ds fileType=%s",
+		task.TaskID, task.DurationSeconds, oss.FileFormat)
+
+	_ = os.Remove(task.TmpFile)
+}
+
+// callModelStream 调用模型，返回原始字节（不做响应映射，用于流式输出）
+func (w *asyncWorker) callModelStream(ctx context.Context, task *entity.ModelGatewayTask, model *entity.ModelGatewayModel, body map[string]any) ([]byte, error) {
+	var data []byte
+	var err error
+
+	if task.Phase == 1 && strings.TrimSpace(task.TmpFile) != "" {
+		data, err = os.ReadFile(task.TmpFile)
+		if err != nil || len(data) == 0 {
+			data = nil
+		}
+	}
+
+	if data == nil {
+		data, err = InvokeModel(ctx, model, body)
+		if err != nil {
+			return nil, err
+		}
+		tmpPath, tmpErr := util.SaveTmpResult(task.TaskID, data, "")
+		if tmpErr == nil && tmpPath != "" {
+			task.TmpFile = tmpPath
+			task.Phase = 1
+			_, err = dao.ModelGatewayTask.Update(ctx, task)
+			if err != nil {
+				g.Log().Errorf(ctx, "[执行任务][失败] 临时文件保存失败 taskId=%s err=%v", task.TaskID, tmpErr)
+			}
+		}
+	}
+
+	return data, nil
+}
+
+// asyncResult 异步任务结果
+type asyncResult struct {
+	result map[string]any
+	err    error
+}
+
+// asyncTaskChan 全局异步任务等待通道
+var asyncTaskChan = sync.Map{} // taskID → chan asyncResult
+
+func (w *asyncWorker) callModelAsync(ctx context.Context, task *entity.ModelGatewayTask, model *entity.ModelGatewayModel, body map[string]any) (map[string]any, error) {
+	// 1. 提交异步任务
+	body, err := w.callModel(ctx, task, model, body)
+	if err != nil {
+		return nil, err
+	}
+	// 2. 拿到 task_id
+	taskID := gjson.New(body).Get(entity.ResponseBody).String()
+
+	// 3. 创建等待通道
+	ch := make(chan asyncResult, 1)
+	asyncTaskChan.Store(taskID, ch)
+	defer func() {
+		asyncTaskChan.Delete(taskID)
+		close(ch)
+	}()
+
+	// 4. 阻塞等待回调或超时
+	timeout := time.Duration(model.TimeoutSeconds) * time.Second
+	ctx, cancel := context.WithTimeout(ctx, timeout)
+	defer cancel()
+
+	g.Log().Infof(ctx, "[异步任务] 开始等待结果 taskID=%s timeout=%v", taskID, timeout)
+
+	select {
+	case res, ok := <-ch:
+		if !ok {
+			return nil, fmt.Errorf("异步任务通道已关闭: taskID=%s", taskID)
+		}
+		g.Log().Infof(ctx, "[异步任务] 获取结果成功 taskID=%s", taskID)
+		return res.result, res.err
+	case <-ctx.Done():
+		return nil, fmt.Errorf("异步任务超时: taskID=%s", taskID)
+	}
+}
+
+// NotifyAsyncResult 回调接口调用此方法通知结果
+func NotifyAsyncResult(taskID string, result map[string]any, err error) {
+	if ch, ok := asyncTaskChan.Load(taskID); ok {
+		ch.(chan asyncResult) <- asyncResult{result: result, err: err}
+	}
+}
+
+// callModel 调用模型 + 检测文件类型 + 保存临时文件
+// 返回: 解析后的响应体, error
+func (w *asyncWorker) callModel(ctx context.Context, task *entity.ModelGatewayTask, model *entity.ModelGatewayModel, body map[string]any) (map[string]any, error) {
+	var data []byte
+	var err error
+
+	// 1) 如果已有临时文件且 phase=1，直接读取
+	if task.Phase == 1 && strings.TrimSpace(task.TmpFile) != "" {
+		data, err = os.ReadFile(task.TmpFile)
+		if err != nil || len(data) == 0 {
+			g.Log().Warningf(ctx, "[callModel] 读取临时文件失败，重新调用模型 taskId=%s err=%v", task.TaskID, err)
+			data = nil
+		}
+	}
+
+	// 2) 没有可用数据，调用模型
+	if data == nil {
+		data, err = InvokeModel(ctx, model, body)
+		if err != nil {
+			return nil, err
+		}
+
+		// 3) 检测文件类型，保存临时文件
+		_, ext := util.DetectFileType(data)
+		tmpPath, tmpErr := util.SaveTmpResult(task.TaskID, data, ext)
+		if tmpErr == nil && tmpPath != "" {
+			task.TmpFile = tmpPath
+			task.Phase = 1
+			_, err = dao.ModelGatewayTask.Update(ctx, task)
+			if err != nil {
+				g.Log().Errorf(ctx, "[执行任务][失败] 临时文件保存失败 taskId=%s err=%v", task.TaskID, tmpErr)
+			}
+		}
+	}
+
+	// 4) 检测文件类型，提取文本结果
+	contentType, _ := util.DetectFileType(data)
+	var textResult string
+	if utf8.Valid(data) && (strings.HasPrefix(contentType, "text/") || contentType == "application/json") {
+		textResult = string(data)
+	}
+
+	// 5) 非文本内容，返回错误
+	if textResult == "" {
+		return nil, fmt.Errorf("模型返回非文本内容，contentType=%s", contentType)
+	}
+
+	// 6) 解析并返回
+	return gjson.New(textResult).Map(), nil
+}
+
+// parseAndRetry 解析模型返回结果，并重试
+func (w *asyncWorker) parseAndRetry(ctx context.Context, body map[string]any, task *entity.ModelGatewayTask, model *entity.ModelGatewayModel, req *dto.CreateTaskReq, maxRetry int, startTime time.Time) (map[string]any, error) {
+	for attempt := 0; attempt <= maxRetry; attempt++ {
+		if attempt > 0 {
+			g.Log().Infof(ctx, "[执行任务][重试] JSON解析 第%d/%d次 taskId=%s", attempt, maxRetry, task.TaskID)
+		}
+
+		// 1) 响应映射
+		mapped, err := util.MapResponsePayload(model.ResponseMapping, body)
+		if err != nil {
+			g.Log().Warningf(ctx, "[执行任务][映射失败] taskId=%s attempt=%d/%d err=%v", task.TaskID, attempt, maxRetry, err)
+			if attempt == maxRetry {
+				return nil, fmt.Errorf("响应映射重试耗尽: %w", err)
+			}
+			continue
+		}
+
+		// 2) 先存 token 到数据库，防止后续失败丢失
+		if _, ok := mapped[entity.TotalTokens]; ok {
+			task.ExpendTokens = gconv.Int64(mapped[entity.TotalTokens])
+			_, err = dao.ModelGatewayTask.Update(ctx, &entity.ModelGatewayTask{
+				SQLBaseDO:    beans.SQLBaseDO{Id: task.Id},
+				ExpendTokens: task.ExpendTokens,
+			})
+		}
+
+		// 3) 解析 + 校验
+		var parsed map[string]any
+		switch req.BuildType {
+		case public.BuildTypePrompt, public.BuildTypeNode:
+			parsed, err = util.ParseAndValidate(mapped, model)
+			if err == nil {
+				return parsed, nil
+			}
+		case public.BuildTypeStruct:
+			parsed = util.ParseStructResult(mapped, entity.ResponseBody)
+			return parsed, nil
+		default:
+			return mapped, nil
+		}
+
+		g.Log().Warningf(ctx, "[执行任务][解析失败] taskId=%s attempt=%d/%d err=%v", task.TaskID, attempt, maxRetry, err)
+
+		if attempt == maxRetry {
+			return nil, fmt.Errorf("JSON解析重试耗尽: %w", err)
+		}
+
+		// 4) 重新调模型（直接调，不走缓存）
+		task.RetryCount++
+		_, _ = dao.ModelGatewayTask.Update(ctx, task)
+		rawData, callErr := InvokeModel(ctx, model, task.RequestPayload.Body)
+
+		if callErr != nil {
+			g.Log().Warningf(ctx, "[执行任务][重调模型失败] taskId=%s attempt=%d/%d err=%v", task.TaskID, attempt, maxRetry, callErr)
+			continue
+		}
+
+		// 5) 解析原始响应，覆盖 body 进入下一轮
+		var rawResp map[string]any
+		if err = json.Unmarshal(rawData, &rawResp); err != nil {
+			g.Log().Warningf(ctx, "[执行任务][Unmarshal失败] taskId=%s err=%v", task.TaskID, err)
+			continue
+		}
+		body = rawResp
+	}
+
+	return body, nil
+}
+
+// InvokeModel 调用模型服务，返回二进制结果
+// modelKey 用于覆盖/补充模型配置 head_msg（例如每次请求携带不同的 X-API-Key）
+func InvokeModel(ctx context.Context, model *entity.ModelGatewayModel, body map[string]any) ([]byte, error) {
+	// 1) 记录模型调用次数
+	_ = dao.ModelGatewayLogsStat.IncRequestCount(ctx, time.Now(), model.TenantId, model.Creator, model.ModelName)
+
+	// 2）请求参数映射：将标准 payload 按模型配置的 requestMapping 转为模型需要的格式
+	//—— 请求映射实际处理为提示词构建请求，因为有附加字段及其他字段的拼接。这里不方便做请求映射
+	//mappedPayload := util.ReverseMap(model.RequestMapping, payload)
+
+	// 3）构建请求 URL 和超时
+	baseURL := strings.TrimRight(model.BaseURL, "/")
+	timeout := time.Duration(model.TimeoutSeconds) * time.Second
+	client := &http.Client{Timeout: timeout}
+	method := strings.ToUpper(strings.TrimSpace(model.HttpMethod))
+
+	// 4）构建 HTTP 请求
+	var req *http.Request
+	switch method {
+	case http.MethodGet:
+		q, err := util.BodyToQuery(body)
+		if err != nil {
+			return nil, err
+		}
+		if len(q) > 0 {
+			if strings.Contains(baseURL, "?") {
+				baseURL = baseURL + "&" + q.Encode()
+			} else {
+				baseURL = baseURL + "?" + q.Encode()
+			}
+		}
+		req, err = http.NewRequestWithContext(ctx, http.MethodGet, baseURL, nil)
+	default:
+		bodyBytes, err := json.Marshal(body)
+		if err != nil {
+			return nil, err
+		}
+		req, err = http.NewRequestWithContext(ctx, http.MethodPost, baseURL, bytes.NewReader(bodyBytes))
+	}
+
+	// 5）注入请求头：先模型静态配置，再动态 modelKey（后者可覆盖前者）
+	for hk, hv := range util.ParseHeadMsgHeaders(model.HeadMsg) {
+		req.Header.Set(hk, hv)
+	}
+	if model.ApiKey != "" {
+		req.Header.Set("Authorization", "Bearer "+model.ApiKey)
+	}
+	if method != http.MethodGet {
+		req.Header.Set("Content-Type", "application/json")
+	}
+
+	// 6）发送请求
+	resp, err := client.Do(req)
+	if err != nil {
+		return nil, err
+	}
+	defer resp.Body.Close()
+
+	// 7）读取响应体
+	b, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return nil, err
+	}
+
+	// 8）检查 HTTP 状态码
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		msg := string(b)
+		return nil, fmt.Errorf("模型服务返回非2xx: %d, body=%s", resp.StatusCode, msg)
+	}
+	return b, nil
+}
+
+// // InvokeModel 调用模型服务，返回二进制结果
+//
+//	func InvokeModel(ctx context.Context, m *entity.AsynchModel, payload any, modelKey string) ([]byte, error) {
+//		if m == nil || m.BaseURL == "" {
+//			return nil, fmt.Errorf("模型配置不完整")
+//		}
+//		// 请求参数映射
+//		mappedPayload, err := mapRequestPayload(m.RequestMapping, payload)
+//		if err != nil {
+//			return nil, fmt.Errorf("请求参数映射失败: %w", err)
+//		}
+//		// 合并请求头
+//		headers := util.ForwardHeaders(ctx)
+//		for hk, hv := range parseHeadMsgHeaders(m.HeadMsg) {
+//			headers[hk] = hv
+//		}
+//		for hk, hv := range parseHeadMsgHeaders(modelKey) {
+//			headers[hk] = hv
+//		}
+//
+//		// 设置超时
+//		timeout := time.Duration(m.TimeoutSeconds) * time.Second
+//		if timeout <= 0 {
+//			timeout = 600 * time.Second
+//		}
+//		ctx, cancel := context.WithTimeout(ctx, timeout)
+//		defer cancel()
+//
+//		invokeUrl := strings.TrimRight(m.BaseURL, "/")
+//		method := strings.ToUpper(strings.TrimSpace(m.HttpMethod))
+//		if method == "" {
+//			method = http.MethodPost
+//		}
+//
+//		var respBytes []byte
+//
+//		switch method {
+//		case http.MethodGet:
+//			err = commonHttp.Get(ctx, invokeUrl, headers, &respBytes, mappedPayload)
+//		default:
+//			err = commonHttp.Post(ctx, invokeUrl, headers, &respBytes, mappedPayload)
+//		}
+//		if err != nil {
+//			return nil, err
+//		}
+//		// 响应参数映射
+//		mappedResponse, err := mapResponsePayload(m.ResponseMapping, respBytes)
+//		if err != nil {
+//			g.Log().Warningf(ctx, "响应参数映射失败: %v，返回原始数据", err)
+//			return respBytes, nil
+//		}
+//		return mappedResponse, nil
+//	}
+
+// failTask 任务失败统一处理：更新数据库 + 释放排队 + 回调
+func (w *asyncWorker) failTask(ctx context.Context, t *entity.ModelGatewayTask, startTime time.Time, errMsg string) {
+	t.State = 3
+	t.ErrorMsg = errMsg
+	t.DurationSeconds = int64(time.Since(startTime).Seconds())
+	_, err := dao.ModelGatewayTask.Update(ctx, t)
+	if err != nil {
+		g.Log().Warningf(ctx, "[执行任务][更新数据库失败] taskId=%s err=%v", t.TaskID, err)
+	}
+	queue.ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
+	go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
+}

service/task_service.go

@@ -1,270 +0,0 @@
-package service
-
-import (
-	"context"
-	"errors"
-	"model-gateway/common/util"
-	"time"
-
-	"model-gateway/dao"
-	"model-gateway/model/dto"
-	"model-gateway/model/entity"
-
-	"github.com/gogf/gf/v2/database/gdb"
-	"github.com/gogf/gf/v2/frame/g"
-	"github.com/gogf/gf/v2/os/gtime"
-	"github.com/google/uuid"
-)
-
-var Task = &taskService{}
-
-type taskService struct{}
-
-func (s *taskService) Create(ctx context.Context, req *dto.CreateTaskReq) (res *dto.CreateTaskRes, err error) {
-	startAt := time.Now()
-	// 固化 token/user 等信息
-	ctx = util.AsyncCtx(ctx)
-	// 1) 检查模型配置
-	m, err := dao.Model.Get(ctx, &entity.AsynchModel{
-		ModelName: req.ModelName,
-	})
-	if err != nil {
-		return nil, err
-	}
-	if m == nil || (m.Enabled != nil && *m.Enabled != 1) {
-		return nil, errors.New("模型不存在或未启用")
-	}
-
-	taskID := uuid.NewString()
-	// 2) 排队上限（严格控制：Redis 原子闸门）
-	limit := GetRuntimeQueueLimit(ctx, req.ModelName, m.QueueLimit)
-	if limit > 0 {
-		ok, err := AcquireQueueSlot(ctx, req.ModelName, taskID, limit, m.ExpectedSeconds)
-		if err != nil {
-			return nil, err
-		}
-		if !ok {
-			return nil, errors.New("任务排队已满，请稍后再试")
-		}
-	}
-
-	// 将调用模型的 payload 与透传头信息一起存入 request_payload，供后台 worker 使用
-	storedPayload := map[string]any{
-		"payload": req.RequestPayload,
-		"headers": util.ForwardHeaders(ctx),
-	}
-
-	t := &entity.AsynchTask{
-		ModelName:      req.ModelName,
-		TaskID:         taskID,
-		State:          0,
-		BizName:        req.BizName,
-		CallbackURL:    req.CallbackUrl,
-		ModelKey:       m.ApiKey,
-		InputRef:       req.InputRef,
-		RequestPayload: storedPayload,
-		EpicycleId:     req.EpicycleId,
-	}
-	_, err = dao.Task.Insert(ctx, t)
-	if err != nil {
-		// 入库失败：回滚闸门占位
-		ReleaseQueueSlot(ctx, req.ModelName, taskID)
-		return nil, err
-	}
-
-	// 3) 写操作日志（尽量不影响主流程，失败忽略）
-	ip := ""
-	ua := ""
-	apiPath := "/task/createTask"
-	httpMethod := "POST"
-	if r := g.RequestFromCtx(ctx); r != nil {
-		ip = r.GetClientIp()
-		ua = r.UserAgent()
-		apiPath = r.URL.Path
-		httpMethod = r.Method
-	}
-	_, _ = dao.OpLog.Insert(ctx, &entity.LogsModelOp{
-		IP:             ip,
-		UserAgent:      ua,
-		APIPath:        apiPath,
-		HttpMethod:     httpMethod,
-		BizName:        req.BizName,
-		ModelName:      req.ModelName,
-		TaskID:         taskID,
-		OpType:         "createTask",
-		Success:        1,
-		ErrorMsg:       "",
-		CostMs:         time.Since(startAt).Milliseconds(),
-		RequestPayload: storedPayload,
-		ResponsePayload: gdb.Map{
-			"taskId": taskID,
-		},
-	})
-
-	// 4) 创建成功后立即异步尝试执行当前任务，并仅在任务仍处于 pending(state=0) 时做定向轮询。
-	// 一旦任务进入 running/success/failed/downloaded，就停止轮询，避免一直空转。
-	go s.pollAndRunUntilPicked(context.WithoutCancel(ctx), taskID, req.EpicycleId)
-	return &dto.CreateTaskRes{TaskID: taskID}, nil
-}
-
-// pollAndRunUntilPicked 用于 createTask 创建后的“轻量级定向轮询”：
-// - 目标：尽快把刚创建的任务拉起来执行
-// - 只在任务仍为 pending(state=0) 时继续尝试抢占
-// - 一旦任务进入 running(1) / success(2) / failed(3) / downloaded(4)，立即停止
-// - 这样不会无限轮询；runWork 仍负责处理积压队列和未处理到的任务
-func (s *taskService) pollAndRunUntilPicked(ctx context.Context, taskID string, epicycleId int64) {
-	if taskID == "" {
-		return
-	}
-	interval := g.Cfg().MustGet(ctx, "asynch.worker.intervalSeconds").Int()
-	if interval <= 0 {
-		interval = 5
-	}
-	g.Log().Infof(ctx, "[task-auto-run][start] taskId=%s interval=%ds", taskID, interval)
-
-	ticker := time.NewTicker(time.Duration(interval) * time.Second)
-	defer ticker.Stop()
-
-	tryRun := func() bool {
-		t, err := dao.Task.Get(ctx, &entity.AsynchTask{
-			TaskID: taskID,
-		})
-		if err != nil {
-			g.Log().Warningf(ctx, "[task-auto-run][stop] taskId=%s reason=query_failed err=%v", taskID, err)
-			return true
-		}
-		if t == nil {
-			g.Log().Warningf(ctx, "[task-auto-run][stop] taskId=%s reason=task_not_found", taskID)
-			return true
-		}
-		switch t.State {
-		case 0:
-			if err = AsyncWorker.RunByTaskID(ctx, taskID, epicycleId); err != nil {
-				g.Log().Warningf(ctx, "[task-auto-run][retry] taskId=%s state=0 err=%v", taskID, err)
-			} else {
-				g.Log().Infof(ctx, "[task-auto-run][triggered] taskId=%s state=0", taskID)
-			}
-			return false
-		case 1:
-			g.Log().Infof(ctx, "[task-auto-run][stop] taskId=%s reason=running", taskID)
-			return true
-		case 2, 3, 4:
-			g.Log().Infof(ctx, "[task-auto-run][stop] taskId=%s reason=terminal state=%d", taskID, t.State)
-			return true
-		default:
-			g.Log().Infof(ctx, "[task-auto-run][stop] taskId=%s reason=unknown_state state=%d", taskID, t.State)
-			return true
-		}
-	}
-
-	// 先立即尝试一次
-	if stop := tryRun(); stop {
-		return
-	}
-	for {
-		select {
-		case <-ctx.Done():
-			g.Log().Infof(ctx, "[task-auto-run][stop] taskId=%s reason=context_done", taskID)
-			return
-		case <-ticker.C:
-			if stop := tryRun(); stop {
-				return
-			}
-		}
-	}
-}
-
-func (s *taskService) GetResult(ctx context.Context, taskID string) (res *dto.GetTaskResultRes, err error) {
-	t, err := dao.Task.Get(ctx, &entity.AsynchTask{
-		TaskID: taskID,
-	})
-	if err != nil {
-		return nil, err
-	}
-	if t == nil {
-		return nil, errors.New("任务不存在")
-	}
-	return &dto.GetTaskResultRes{
-		OssFile: t.OssFile,
-		State:   t.State,
-	}, nil
-}
-
-// GetBatch 批量查询任务；将成功(state=2)的任务更新为已下载(state=4)，并写入过期时间
-func (s *taskService) GetBatch(ctx context.Context, req *dto.GetTaskBatchReq) (res *dto.GetTaskBatchRes, err error) {
-	if req == nil || len(req.TaskIDs) == 0 {
-		return &dto.GetTaskBatchRes{List: []dto.GetTaskBatchItem{}}, nil
-	}
-	// 1) 先查当前租户下的任务列表
-	list, err := dao.Task.ListByTaskIDs(ctx, req.TaskIDs)
-	if err != nil {
-		return nil, err
-	}
-
-	// 2) 对成功(state=2)的任务：标记为已下载(state=4)并写入 expire_at
-	now := time.Now()
-	for _, t := range list {
-		if t == nil {
-			continue
-		}
-		if t.State != 2 {
-			continue
-		}
-		// 按模型配置决定保留时间
-		m, err := dao.Model.Get(ctx, &entity.AsynchModel{
-			ModelName: t.ModelName,
-		})
-		if err != nil {
-			return nil, err
-		}
-		retainSeconds := 86400
-		if m != nil && m.AutoCleanSeconds > 0 {
-			retainSeconds = m.AutoCleanSeconds
-		}
-		expireAt := gtime.New(now.Add(time.Duration(retainSeconds) * time.Second))
-		_ = dao.Task.MarkDownloadedByID(ctx, t.Id, expireAt)
-
-		// 为了本次返回一致性，内存里也更新
-		t.State = 4
-		t.ExpireAt = expireAt
-	}
-
-	// 3) 组装返回
-	items := make([]dto.GetTaskBatchItem, 0, len(list))
-	for _, t := range list {
-		if t == nil {
-			continue
-		}
-		items = append(items, dto.GetTaskBatchItem{
-			TaskID:  t.TaskID,
-			State:   t.State,
-			OssFile: t.OssFile,
-		})
-	}
-	return &dto.GetTaskBatchRes{List: items}, nil
-}
-
-func (s *taskService) List(ctx context.Context, req *dto.ListTaskReq) (res *dto.ListTaskRes, err error) {
-	pageNum, pageSize := 1, 10
-	if req != nil {
-		if req.PageNum > 0 {
-			pageNum = req.PageNum
-		}
-		if req.PageSize > 0 {
-			pageSize = req.PageSize
-		}
-	}
-	modelName := ""
-	taskID := ""
-	var state *int
-	if req != nil {
-		modelName = req.ModelName
-		taskID = req.TaskID
-		state = req.State
-	}
-	list, total, err := dao.Task.List(ctx, pageNum, pageSize, modelName, taskID, state)
-	if err != nil {
-		return nil, err
-	}
-	return &dto.ListTaskRes{List: list, Total: total}, nil
-}

service/worker.go

@@ -1,271 +0,0 @@
-package service
-
-import (
-	"context"
-	"errors"
-	"fmt"
-	"model-gateway/common/util"
-	"model-gateway/model/dto"
-	"model-gateway/service/gateway"
-	"os"
-	"path/filepath"
-	"strings"
-	"time"
-	"unicode/utf8"
-
-	"model-gateway/dao"
-	"model-gateway/model/entity"
-
-	"github.com/gogf/gf/v2/frame/g"
-	"github.com/gogf/gf/v2/os/grpool"
-	"github.com/tidwall/gjson"
-)
-
-var AsyncWorker = &asyncWorker{}
-
-type asyncWorker struct {
-}
-
-// RunOnce 由上层定时任务触发：一次性抢占并处理一批任务
-// - batchSize: 本次抢占数量
-// - goroutines: 本次并发数（协程池大小）
-func (w *asyncWorker) RunOnce(ctx context.Context, req *dto.RunWorkReq) (res *dto.RunWorkRes, err error) {
-	if req.BatchSize <= 0 {
-		req.BatchSize = 10
-	}
-	if req.Goroutines <= 0 {
-		req.Goroutines = 1
-	}
-	tasks, err := dao.Task.ClaimPendingGlobal(ctx, req.BatchSize)
-	if err != nil {
-		return nil, err
-	}
-	if len(tasks) == 0 {
-		return nil, errors.New("no task to run")
-	}
-	pool := grpool.New(req.Goroutines)
-	defer pool.Close()
-	claimed := len(tasks)
-	done := make(chan struct{}, claimed)
-	for _, t := range tasks {
-		task := t
-		_ = pool.AddWithRecover(ctx, func(ctx context.Context) {
-			w.handleOne(ctx, task, 0)
-			done <- struct{}{}
-		}, func(ctx context.Context, e error) {
-			if e != nil {
-				_ = dao.Task.UpdateFailedGlobal(ctx, task.Id, fmt.Sprintf("worker panic: %v", e))
-				ReleaseQueueSlot(ctx, task.ModelName, task.TaskID)
-			}
-			done <- struct{}{}
-		})
-	}
-	for i := 0; i < claimed; i++ {
-		<-done
-	}
-	return &dto.RunWorkRes{
-		Claimed: claimed,
-	}, nil
-}
-
-// RunByTaskID 创建任务后立即异步尝试执行当前任务：
-// - 只定向抢占当前 taskId 对应的 pending 任务
-// - 若任务已被其它 worker 抢走/已不在 pending，则直接返回
-func (w *asyncWorker) RunByTaskID(ctx context.Context, taskID string, epicycleId int64) error {
-	task, err := dao.Task.ClaimPendingByTaskIDGlobal(ctx, taskID)
-	if err != nil {
-		return err
-	}
-	if task == nil {
-		return nil
-	}
-	w.handleOne(ctx, task, epicycleId)
-	return nil
-}
-
-func (w *asyncWorker) handleOne(ctx context.Context, t *entity.AsynchTask, epicycleId int64) {
-	// 从任务入库的 request_payload 里恢复 payload + headers
-	payload, headers := util.ParseStoredPayload(t.RequestPayload)
-	if len(headers) > 0 {
-		ctx = util.SetTaskHeadersToCtx(ctx, headers)
-	}
-
-	// 1) 拉取模型配置
-	m, err := dao.Model.GetByModelNameForTenant(ctx, t.TenantId, t.ModelName)
-	if err != nil {
-		_ = dao.Task.UpdateFailedGlobal(ctx, t.Id, err.Error())
-		ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-		// ============ 失败回调 ============
-		t.State = 3
-		t.ErrorMsg = err.Error()
-		go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
-		// ================================
-		return
-	}
-	if m == nil || (m.Enabled != nil && *m.Enabled != 1) {
-		errMsg := "模型不存在或未启用"
-		_ = dao.Task.UpdateFailedGlobal(ctx, t.Id, errMsg)
-		ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-		// ============ 失败回调 ============
-		t.State = 3
-		t.ErrorMsg = errMsg
-		go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
-		// ================================
-		return
-	}
-
-	// 2) 分布式并发限制
-	semKey := fmt.Sprintf("asynch:sem:%s", t.ModelName)
-	leaseSeconds := int64(3600)
-	maxC := GetRuntimeMaxConcurrency(ctx, t.ModelName, m.MaxConcurrency)
-	acquired, err := acquireSemaphore(ctx, semKey, maxC, leaseSeconds)
-	if err != nil {
-		_ = dao.Task.UpdateFailedGlobal(ctx, t.Id, err.Error())
-		ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-		// ============ 失败回调 ============
-		t.State = 3
-		t.ErrorMsg = err.Error()
-		go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
-		// ================================
-		return
-	}
-	if !acquired {
-		// 并发满了：放回排队，不回调（不是失败）
-		_ = w.rollbackToPending(ctx, t.Id)
-		return
-	}
-	defer func() {
-		_ = releaseSemaphore(ctx, semKey)
-	}()
-
-	// 3) 调用模型服务
-	if payload == nil {
-		payload = map[string]any{
-			"taskId":   t.TaskID,
-			"inputRef": t.InputRef,
-		}
-	}
-	var (
-		data        []byte
-		contentType string
-		ext         string
-		textResult  string
-	)
-
-	// phase=1 表示模型已成功但 OSS 上传失败：优先从临时文件加载
-	if t.Phase == 1 && strings.TrimSpace(t.TmpFile) != "" {
-		data, err = os.ReadFile(t.TmpFile)
-		if err == nil && len(data) > 0 {
-			contentType, ext = util.DetectFileType(data)
-		} else {
-			data = nil
-		}
-	}
-	if data == nil {
-		// 统计
-		_ = dao.Stat.IncRequestCount(ctx, time.Now(), int64(t.TenantId), t.Creator, t.ModelName)
-		// 核心调用
-		data, err = InvokeModel(ctx, m, payload, t.ModelKey)
-		if err != nil {
-			_ = dao.Task.UpdateFailedGlobal(ctx, t.Id, err.Error())
-			ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-			// ============ 失败回调 ============
-			t.State = 3
-			t.ErrorMsg = err.Error()
-			go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
-			// ================================
-			return
-		}
-		contentType, ext = util.DetectFileType(data)
-		if utf8.Valid(data) && (strings.HasPrefix(contentType, "text/") || contentType == "application/json") {
-			textResult = string(data)
-		}
-		tmpPath, err := saveTmpResult(t.TaskID, data, ext)
-		if err == nil && tmpPath != "" {
-			t.TmpFile = tmpPath
-			t.Phase = 1
-			_ = dao.Task.UpdateTmpAfterModelGlobal(ctx, t.Id, tmpPath)
-		}
-	}
-
-	// 4) 存储 OSS
-	ossURL, err := gateway.UploadByTask(ctx, t, data, ext, contentType)
-	if err != nil {
-		// OSS 阶段失败：保留临时文件，下一轮仅重试 OSS
-		_ = dao.Task.UpdateFailedKeepTmpGlobal(ctx, t.Id, err.Error())
-		ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-		// ============ OSS失败不回调（还会重试） ============
-		// 注意：OSS失败保留临时文件，下次重试，所以这里不触发最终回调
-		// 如果已经重试多次还没成功，需要在任务超时或超过最大重试次数时才回调失败
-		return
-	}
-
-	// 5) 更新任务状态成功
-	fileType := strings.TrimPrefix(ext, ".")
-	if fileType == "" {
-		fileType = contentType
-	}
-	if err = dao.Task.UpdateSuccessGlobal(
-		ctx,
-		t.Id,
-		ossURL,
-		fileType,
-		textResult,
-		int64(len(data)),
-		nil,
-		GetExpendTokens(m.ResponseTokenField, textResult),
-	); err != nil {
-		g.Log().Errorf(ctx, "[worker] update success failed: %v", err)
-		return
-	}
-
-	// 成功/失败均不再占用 queue_limit
-	ReleaseQueueSlot(ctx, t.ModelName, t.TaskID)
-
-	// 6) 成功回调
-	t.State = 2
-	t.OssFile = ossURL
-	t.FileType = fileType
-	t.TextResult = textResult
-	g.Log().Infof(ctx, "[CALLBACK][DISPATCH] taskId=%s bizName=%s callbackUrl=%s", t.TaskID, t.BizName, t.CallbackURL)
-	go gateway.TriggerCallback(context.WithoutCancel(ctx), t)
-	// ============ 如果有 epicycleId，也触发业务回调 ============
-	if epicycleId != 0 {
-		go gateway.TriggerPromptsCallback(context.WithoutCancel(ctx), t, epicycleId)
-	}
-
-	// 成功后清理临时文件
-	_ = os.Remove(t.TmpFile)
-}
-
-// saveTmpResult 将模型输出写入临时文件，用于 OSS 上传失败后的“仅重试 OSS”。
-func saveTmpResult(taskID string, data []byte, ext string) (string, error) {
-	dir := filepath.Join(os.TempDir(), "model-asynch")
-	if err := os.MkdirAll(dir, 0o755); err != nil {
-		return "", err
-	}
-	if ext == "" {
-		ext = ".bin"
-	}
-	if ext[0] != '.' {
-		ext = "." + ext
-	}
-	path := filepath.Join(dir, fmt.Sprintf("%s%s", taskID, ext))
-	if err := os.WriteFile(path, data, 0o644); err != nil {
-		return "", err
-	}
-	return path, nil
-}
-
-func (w *asyncWorker) rollbackToPending(ctx context.Context, id int64) error {
-	return dao.Task.RollbackToPendingGlobal(ctx, id)
-}
-
-// GetExpendTokens 根据映射路径从 textResult 中提取消耗 token 值
-func GetExpendTokens(tokenMapping string, textResult string) int {
-	value := gjson.Get(textResult, tokenMapping)
-	if value.Exists() {
-		return int(value.Int())
-	}
-	return len(textResult)
-}

timezone/timezone

@@ -1 +0,0 @@
-Asia/Shanghai

update.sql

@@ -1,282 +1,230 @@
--- model-asynch 核心表(pgsql)
--- 1) asynch_models：模型配置
--- 2) asynch_task：异步任务
--- 3) logs_model_op：操作日志(统计用)
--- 4) logs_model_stat：按天模型请求统计(限流/监控用)
-
 -- =========================
--- 1) asynch_models
+-- model_gateway_models
 -- =========================
-CREATE TABLE IF NOT EXISTS asynch_models (
-    -- 基础字段
-    id                         BIGINT PRIMARY KEY,                 -- 主键ID(非自增)
-    tenant_id                  BIGINT NOT NULL DEFAULT 0,          -- 租户ID
-    creator                    VARCHAR(64) NOT NULL,               -- 创建人
-    created_at                 TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- 创建时间
-    updater                    VARCHAR(64) NOT NULL,               -- 更新人
-    updated_at                 TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- 更新时间
-    deleted_at                 TIMESTAMP(6),                       -- 删除时间(软删)
-    -- 业务字段
-    model_name                 VARCHAR(128) NOT NULL,              -- 模型名称
-    model_type                SMALLINT NOT NULL DEFAULT 0,        -- 模型类型
-    base_url                   VARCHAR(256) NOT NULL,              -- 模型地址
-    http_method                VARCHAR(8) NOT NULL DEFAULT 'POST', -- 请求方式 GET/POST
-    head_msg                   VARCHAR(1024) DEFAULT '',           -- 请求头绑定(支持多个,逗号分隔)示例 X-API:xxx,operation:true
-    is_private                 SMALLINT NOT NULL DEFAULT 0,        -- 是否私有化 0-私有 1-公共
-    enabled                    SMALLINT NOT NULL DEFAULT 1,        -- 是否启用 0停用 1-启用
-    is_chat_model              SMALLINT NOT NULL DEFAULT 0,        -- 是否为对话模型 0-否 1-是
-    is_owner              SMALLINT NOT NULL DEFAULT 99,        -- 1=当前用户创建的，0=超级管理员的
-    api_key                    VARCHAR(256) NOT NULL DEFAULT '',   -- 调用凭证,密钥
-    prompt                     TEXT NOT NULL DEFAULT '',           -- 提示词内容(文本)
-    form_json                  JSONB NOT NULL DEFAULT '{}'::jsonb, -- 表单结构(用于前端渲染)
-    request_mapping            JSONB NOT NULL DEFAULT '{}'::jsonb  -- 请求映射
-    response_mapping           JSONB NOT NULL DEFAULT '{}'::jsonb, -- 返回映射
-    response_body              JSONB NOT NULL DEFAULT '{}'::jsonb, -- 返回主体
-    max_concurrency            INT NOT NULL DEFAULT 10,            -- 单模型最大并发
-    queue_limit                INT NOT NULL DEFAULT 1000,          -- 排队上限(近似控制)
-    timeout_seconds            INT NOT NULL DEFAULT 600,           -- 调用模型服务超时(秒)
-    expected_seconds           INT NOT NULL DEFAULT 600,           -- 模型预计执行时间(秒)
-    retry_times                SMALLINT NOT NULL DEFAULT 3,        -- 失败重试次数
-    retry_queue_max_seconds    INT NOT NULL DEFAULT 600,           -- 失败重试最大排队时间(秒 0=插队到队首；>0=排队超过该时间后插队，否则仍到队尾)
-    auto_clean_seconds         INT NOT NULL DEFAULT 86400,         -- 已下载(state=4 后的保留时间(秒)，到期清理)
-    remark                     TEXT DEFAULT ''                     -- 备注
-    response_token_field              VARCHAR(128) NOT NULL DEFAULT '';   -- 响应中消耗token的字段映射
-    operator_name              VARCHAR(64) NOT NULL DEFAULT '',    -- 运营商名称
-    token_config               JSONB NOT NULL DEFAULT '{
-    "zh_ratio": 1.0,
-    "en_ratio": 1.3,
-    "space_ratio": 0.1,
-    "punctuation_ratio": 0.1,
-    "max_window_size": 8192,
-    "reserve_ratio": 0.2,
-    "min_reserve": 512,
-}'::jsonb  -- Token配置
+CREATE TABLE IF NOT EXISTS model_gateway_models (
+                                                    id              int8 PRIMARY KEY,
+                                                    tenant_id       int8 NOT NULL DEFAULT 0,
+                                                    creator         varchar(64) NOT NULL,
+    created_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updater         varchar(64) NOT NULL,
+    updated_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    deleted_at      timestamp(6),
+    model_name      varchar(128) NOT NULL,
+    model_type      int2 NOT NULL DEFAULT 0,
+    operator_name   varchar(64) NOT NULL DEFAULT '',
+    base_url        varchar(256) NOT NULL,
+    http_method     varchar(8) NOT NULL DEFAULT 'POST',
+    head_msg        jsonb NOT NULL DEFAULT '{}',
+    api_key         varchar(256) NOT NULL DEFAULT '',
+    is_private      int2 NOT NULL DEFAULT 0,
+    enabled         int2 NOT NULL DEFAULT 1,
+    is_chat_model   int2 NOT NULL DEFAULT 0,
+    is_owner        int2 NOT NULL DEFAULT 99,
+    form_json       jsonb NOT NULL DEFAULT '{}',
+    request_mapping  jsonb NOT NULL DEFAULT '{}',
+    response_mapping jsonb NOT NULL DEFAULT '{}',
+    response_body   varchar(128) NOT NULL DEFAULT '',
+    token_config    jsonb NOT NULL DEFAULT '{}',
+    extend_mapping  jsonb NOT NULL DEFAULT '{}',
+    query_config    jsonb NOT NULL DEFAULT '{}',
+    stream_config   jsonb NOT NULL DEFAULT '{}',
+    first_frame     varchar(128) NOT NULL DEFAULT '',
+    last_frame      varchar(128) NOT NULL DEFAULT '',
+    max_concurrency int4 NOT NULL DEFAULT 10,
+    timeout_seconds int4 NOT NULL DEFAULT 600,
+    retry_times     int2 NOT NULL DEFAULT 3,
+    response_token_field varchar(128) NOT NULL DEFAULT '',
+    call_mode       int2 NOT NULL DEFAULT 0,
+    required_fields jsonb NOT NULL DEFAULT '[]',
+    max_tokens      int4 DEFAULT 0
     );
-CREATE UNIQUE INDEX IF NOT EXISTS uk_asynch_models_tenant_creator_chat ON asynch_models(tenant_id, creator) WHERE is_chat_model = 1 AND deleted_at IS NULL;
-CREATE UNIQUE INDEX IF NOT EXISTS uk_asynch_models_tenant_model_name ON asynch_models(tenant_id, creator, model_name);
-CREATE INDEX IF NOT EXISTS idx_asynch_models_tenant_id ON asynch_models(tenant_id);
-CREATE INDEX IF NOT EXISTS idx_asynch_models_model_name ON asynch_models(model_name);
-CREATE INDEX IF NOT EXISTS idx_asynch_models_model_type ON asynch_models(model_type);
-CREATE INDEX IF NOT EXISTS idx_asynch_models_enabled ON asynch_models(enabled);
-CREATE INDEX IF NOT EXISTS idx_asynch_models_deleted_at ON asynch_models(deleted_at);
 
-COMMENT ON TABLE asynch_models IS '模型配置表';
-COMMENT ON COLUMN asynch_models.id IS '主键ID(非自增)';
-COMMENT ON COLUMN asynch_models.tenant_id IS '租户ID';
-COMMENT ON COLUMN asynch_models.creator IS '创建人';
-COMMENT ON COLUMN asynch_models.created_at IS '创建时间';
-COMMENT ON COLUMN asynch_models.updater IS '更新人';
-COMMENT ON COLUMN asynch_models.updated_at IS '更新时间';
-COMMENT ON COLUMN asynch_models.deleted_at IS '删除时间(软删)';
+CREATE UNIQUE INDEX IF NOT EXISTS uk_model_gateway_models_tenant_creator_model ON model_gateway_models (tenant_id, creator, model_name);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_models_model_name ON model_gateway_models (model_name);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_models_model_type ON model_gateway_models (model_type);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_models_tenant_id ON model_gateway_models (tenant_id);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_models_deleted_at ON model_gateway_models (deleted_at);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_models_enabled ON model_gateway_models (enabled);
 
-COMMENT ON COLUMN asynch_models.model_name IS '模型名称';
-COMMENT ON COLUMN asynch_models.model_type IS '模型类型';
-COMMENT ON COLUMN asynch_models.base_url IS '模型地址';
-COMMENT ON COLUMN asynch_models.http_method IS '请求方式 GET/POST';
-COMMENT ON COLUMN asynch_models.head_msg IS '请求头绑定(支持多个,逗号分隔)示例 X-API:xxx,operation:true';
-COMMENT ON COLUMN asynch_models.is_private IS '是否私有化 0-私有 1-公共';
-COMMENT ON COLUMN asynch_models.enabled IS '是否启用 0停用 1-启用';
-COMMENT ON COLUMN asynch_models.is_chat_model IS '是否为对话模型 0-否 1-是';
-COMMENT ON COLUMN asynch_models.is_owner IS '1=当前用户创建的，0=超级管理员的';
-COMMENT ON COLUMN asynch_models.api_key IS '调用凭证,密钥';
-COMMENT ON COLUMN asynch_models.prompt IS '提示词内容(文本)';
-COMMENT ON COLUMN asynch_models.form_json IS '表单结构(用于前端渲染，也用于后端校验)';
-COMMENT ON COLUMN asynch_models.request_mapping IS '请求映射';
-COMMENT ON COLUMN asynch_models.response_mapping IS '返回映射';
-COMMENT ON COLUMN asynch_models.response_body IS '返回主体';
-COMMENT ON COLUMN asynch_models.max_concurrency IS '单模型最大并发';
-COMMENT ON COLUMN asynch_models.queue_limit IS '排队上限(近似控制)';
-COMMENT ON COLUMN asynch_models.timeout_seconds IS '调用模型服务超时(秒)';
-COMMENT ON COLUMN asynch_models.expected_seconds IS '模型预计执行时间(秒)';
-COMMENT ON COLUMN asynch_models.retry_times IS '失败重试次数';
-COMMENT ON COLUMN asynch_models.retry_queue_max_seconds IS '失败重试最大排队时间(秒 0=插队到队首；>0=排队超过该时间后插队，否则仍到队尾)';
-COMMENT ON COLUMN asynch_models.auto_clean_seconds IS '已下载(state=4 后的保留时间(秒)，到期清理)';
-COMMENT ON COLUMN asynch_models.remark IS '备注';
-COMMENT ON COLUMN asynch_models.response_token_field IS '响应中消耗token的字段映射';
-COMMENT ON COLUMN asynch_models.operator_name IS '运营商名称';
-COMMENT ON COLUMN asynch_models.token_config IS '{
-  "zh_ratio": 1.0,          // 中文字符→token系数
-  "en_ratio": 1.3,          // 英文单词→token系数
-  "space_ratio": 0.1,       // 空格系数
-  "punctuation_ratio": 0.1, // 标点系数
-  "max_window_size": 8192,  // 模型最大窗口
-  "reserve_ratio": 0.2,     // 预留回复空间比例
-  "min_reserve": 512,       // 最少预留token数
-}';
+COMMENT ON TABLE model_gateway_models IS '模型配置表';
+COMMENT ON COLUMN model_gateway_models.id IS '主键ID(非自增)';
+COMMENT ON COLUMN model_gateway_models.tenant_id IS '租户ID';
+COMMENT ON COLUMN model_gateway_models.creator IS '创建人';
+COMMENT ON COLUMN model_gateway_models.created_at IS '创建时间';
+COMMENT ON COLUMN model_gateway_models.updater IS '更新人';
+COMMENT ON COLUMN model_gateway_models.updated_at IS '更新时间';
+COMMENT ON COLUMN model_gateway_models.deleted_at IS '删除时间(软删)';
 
-
--- =========================
--- 2) asynch_task
--- =========================
-CREATE TABLE IF NOT EXISTS asynch_task (
-    -- 基础字段
-    id                         BIGINT PRIMARY KEY,                 -- 主键ID(非自增)
-    tenant_id                  BIGINT NOT NULL DEFAULT 0,          -- 租户ID
-    creator                    VARCHAR(64) NOT NULL,               -- 创建人
-    created_at                 TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- 创建时间
-    updater                    VARCHAR(64) NOT NULL,               -- 更新人
-    updated_at                 TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- 更新时间
-    deleted_at                 TIMESTAMP(6),                       -- 删除时间(软删)
-
-    -- 业务字段
-    model_name                 VARCHAR(128) NOT NULL,              -- 模型名称
-    task_id                    VARCHAR(64) NOT NULL,               -- 任务ID(对外返回)
-    biz_name                   VARCHAR(128) NOT NULL DEFAULT '',   -- 业务名称(调用方模块/系统)
-    callback_url               VARCHAR(512) DEFAULT '',            -- 回调地址(可选，用于后续业务通知)
-    model_key                  VARCHAR(1024) DEFAULT '',           -- 动态请求头(用于覆盖/补充模型配置 head_msg)，如 X-API-Key:xxx
-    state                      SMALLINT NOT NULL DEFAULT 0,        -- 0排队中/1执行中/2成功/3失败/4已下载
-    oss_file                   VARCHAR(512) DEFAULT '',            -- 结果文件OSS地址
-    file_type                  VARCHAR(32) DEFAULT '',             -- 文件类型(mp3/mp4/png/...)
-    file_size                  BIGINT NOT NULL DEFAULT 0,          -- 文件大小(字节)
-    error_msg                  TEXT DEFAULT '',                    -- 错误信息
-    started_at                 TIMESTAMP,                          -- 开始执行时间
-    finished_at                TIMESTAMP,                          -- 执行结束时间
-    duration_seconds           BIGINT NOT NULL DEFAULT 0,          -- 耗时(秒)：从创建到完成(成功/失败)整体耗时
-    expire_at                  TIMESTAMP,                          -- state=4 后写入，用于清理
-    retry_count                INT NOT NULL DEFAULT 0,             -- 已重试次数(不含首次)
-    enqueue_at                 TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- 入队时间(用于排队顺序)
-    phase                      SMALLINT NOT NULL DEFAULT 0,        -- 0模型阶段/1OSS阶段
-    tmp_file                   TEXT DEFAULT '',                    -- 临时结果文件路径(phase=1 时仅重试 OSS 上传)
-    input_ref                  TEXT DEFAULT '',                    -- 输入引用(如OSS/业务资源ID等)
-    request_payload            JSONB,                              -- 请求参数(可选)
-    text_result                TEXT DEFAULT '',                    -- 文本类结果(可选，支持直接回调)
-    epicycle_id                VARCHAR(64) DEFAULT '',             -- 轮次ID
-    expend_tokens              BIGINT NOT NULL DEFAULT 0           -- 消耗 token 数
-);
-
-CREATE UNIQUE INDEX IF NOT EXISTS uk_asynch_task_tenant_task_id ON asynch_task(tenant_id, task_id);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_tenant_id ON asynch_task(tenant_id);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_model_name ON asynch_task(model_name);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_biz_name ON asynch_task(biz_name);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_model_key ON asynch_task(model_key);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_state ON asynch_task(state);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_enqueue_at ON asynch_task(enqueue_at);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_updated_at ON asynch_task(updated_at);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_expire_at ON asynch_task(expire_at);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_deleted_at ON asynch_task(deleted_at);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_epicycle_id ON asynch_task(epicycle_id);
-CREATE INDEX IF NOT EXISTS idx_asynch_task_expend_tokens ON asynch_task(expend_tokens);
-
-COMMENT ON TABLE asynch_task IS '异步任务表';
-COMMENT ON COLUMN asynch_task.id IS '主键ID(非自增)';
-COMMENT ON COLUMN asynch_task.tenant_id IS '租户ID';
-COMMENT ON COLUMN asynch_task.creator IS '创建人';
-COMMENT ON COLUMN asynch_task.created_at IS '创建时间';
-COMMENT ON COLUMN asynch_task.updater IS '更新人';
-COMMENT ON COLUMN asynch_task.updated_at IS '更新时间';
-COMMENT ON COLUMN asynch_task.deleted_at IS '删除时间(软删)';
-COMMENT ON COLUMN asynch_task.model_name IS '模型名称';
-COMMENT ON COLUMN asynch_task.task_id IS '任务ID(对外返回)';
-COMMENT ON COLUMN asynch_task.biz_name IS '业务名称(调用方模块/系统)';
-COMMENT ON COLUMN asynch_task.callback_url IS '回调地址(可选，用于后续业务通知)';
-COMMENT ON COLUMN asynch_task.model_key IS '动态请求头(用于覆盖/补充模型配置 head_msg)，如 X-API-Key:xxx';
-COMMENT ON COLUMN asynch_task.state IS '0排队中/1执行中/2成功/3失败/4已下载';
-COMMENT ON COLUMN asynch_task.oss_file IS '结果文件OSS地址';
-COMMENT ON COLUMN asynch_task.file_type IS '文件类型(mp3/mp4/png/...)';
-COMMENT ON COLUMN asynch_task.file_size IS '文件大小(字节)';
-COMMENT ON COLUMN asynch_task.error_msg IS '错误信息';
-COMMENT ON COLUMN asynch_task.started_at IS '开始执行时间';
-COMMENT ON COLUMN asynch_task.finished_at IS '执行结束时间';
-COMMENT ON COLUMN asynch_task.duration_seconds IS '耗时(秒)：从创建到完成(成功/失败)整体耗时';
-COMMENT ON COLUMN asynch_task.expire_at IS 'state=4 后写入，用于清理';
-COMMENT ON COLUMN asynch_task.retry_count IS '已重试次数(不含首次)';
-COMMENT ON COLUMN asynch_task.enqueue_at IS '入队时间(用于排队顺序)';
-COMMENT ON COLUMN asynch_task.phase IS '执行阶段 模型阶段/1OSS阶段(模型已成功,等待上传OSS)';
-COMMENT ON COLUMN asynch_task.tmp_file IS '临时结果文件路径(phase=1 时仅重试 OSS 上传)';
-COMMENT ON COLUMN asynch_task.input_ref IS '输入引用(如OSS/业务资源ID等)';
-COMMENT ON COLUMN asynch_task.request_payload IS '请求参数(可选,JSON)';
-COMMENT ON COLUMN asynch_task.text_result IS '文本类结果(可选，支持直接回调)';
-COMMENT ON COLUMN asynch_task.epicycle_id IS '轮次ID(用于标识同一轮次的任务)';
-COMMENT ON COLUMN asynch_task.expend_tokens IS '消耗 token 数';
+COMMENT ON COLUMN model_gateway_models.model_name IS '模型名称';
+COMMENT ON COLUMN model_gateway_models.model_type IS '模型类型';
+COMMENT ON COLUMN model_gateway_models.operator_name IS '运营商名称';
+COMMENT ON COLUMN model_gateway_models.base_url IS '模型地址';
+COMMENT ON COLUMN model_gateway_models.http_method IS '请求方式 GET/POST';
+COMMENT ON COLUMN model_gateway_models.head_msg IS '请求头信息';
+COMMENT ON COLUMN model_gateway_models.api_key IS '调用凭证/密钥';
+COMMENT ON COLUMN model_gateway_models.is_private IS '是否私有化：0-私有 1-公共';
+COMMENT ON COLUMN model_gateway_models.enabled IS '是否启用：0-停用 1-启用';
+COMMENT ON COLUMN model_gateway_models.is_chat_model IS '是否为对话模型：0-否 1-是';
+COMMENT ON COLUMN model_gateway_models.is_owner IS '1=当前用户创建 0=超级管理员';
+COMMENT ON COLUMN model_gateway_models.call_mode IS '调用模式：0-同步 1-异步 2-流式';
+COMMENT ON COLUMN model_gateway_models.form_json IS '动态表单结构';
+COMMENT ON COLUMN model_gateway_models.request_mapping IS '请求映射';
+COMMENT ON COLUMN model_gateway_models.response_mapping IS '返回映射';
+COMMENT ON COLUMN model_gateway_models.response_body IS '返回主体';
+COMMENT ON COLUMN model_gateway_models.token_config IS 'Token计算配置';
+COMMENT ON COLUMN model_gateway_models.extend_mapping IS '附加映射';
+COMMENT ON COLUMN model_gateway_models.query_config IS '查询/回调配置';
+COMMENT ON COLUMN model_gateway_models.stream_config IS '流式输出配置';
+COMMENT ON COLUMN model_gateway_models.first_frame IS '首帧图片参数';
+COMMENT ON COLUMN model_gateway_models.last_frame IS '尾帧图片参数';
+COMMENT ON COLUMN model_gateway_models.max_concurrency IS '最大并发数';
+COMMENT ON COLUMN model_gateway_models.timeout_seconds IS '调用模型超时(秒)';
+COMMENT ON COLUMN model_gateway_models.retry_times IS '失败重试次数';
+COMMENT ON COLUMN model_gateway_models.response_token_field IS '响应中消耗token的字段映射';
+COMMENT ON COLUMN model_gateway_models.required_fields IS '必选字段列表';
+COMMENT ON COLUMN model_gateway_models.max_tokens IS '最大 token 数，0 表示不传';
 
 
 
 -- =========================
--- 3) logs_model_op
+-- model_gateway_task
 -- =========================
-CREATE TABLE IF NOT EXISTS logs_model_op (
-    -- 基础字段
-    id BIGINT PRIMARY KEY,
-    tenant_id BIGINT NOT NULL DEFAULT 0,
-    creator VARCHAR(64) NOT NULL,
-    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    updater VARCHAR(64) NOT NULL,
-    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    deleted_at TIMESTAMP(6),
-    -- 基础审计信息
-    ip VARCHAR(64) DEFAULT '',
-    user_agent VARCHAR(256) DEFAULT '',
-    api_path VARCHAR(256) DEFAULT '',
-    http_method VARCHAR(16) DEFAULT '',
-    -- 业务信息
-    biz_name VARCHAR(128) NOT NULL DEFAULT '', -- 调用方业务模块/系统
-    model_name VARCHAR(128) NOT NULL DEFAULT '',
-    task_id VARCHAR(64) NOT NULL DEFAULT '',
-    -- 统计字段
-    op_type VARCHAR(64) NOT NULL DEFAULT 'createTask', -- 操作类型(默认创建任务)
-    success SMALLINT NOT NULL DEFAULT 1, -- 1成功/0失败
-    error_msg TEXT DEFAULT '',
-    cost_ms BIGINT NOT NULL DEFAULT 0, -- 耗时(毫秒)
-    -- 请求/响应 JSON(用于后期统计分析)
-    request_payload JSONB,
-    response_payload JSONB
-);
+CREATE TABLE IF NOT EXISTS model_gateway_task (
+                                                  id              int8 PRIMARY KEY,
+                                                  tenant_id       int8 NOT NULL DEFAULT 0,
+                                                  creator         varchar(64) NOT NULL,
+    created_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updater         varchar(64) NOT NULL,
+    updated_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    deleted_at      timestamp(6),
+    model_name      varchar(128) NOT NULL,
+    task_id         varchar(64) NOT NULL,
+    biz_name        varchar(128) NOT NULL DEFAULT '',
+    callback_url    varchar(512) DEFAULT '',
+    state           int2 NOT NULL DEFAULT 0,
+    retry_count     int4 NOT NULL DEFAULT 0,
+    phase           int2 NOT NULL DEFAULT 0,
+    tmp_file        text DEFAULT '',
+    error_msg       text DEFAULT '',
+    result_file     jsonb NOT NULL DEFAULT '{}',
+    request_payload jsonb NOT NULL DEFAULT '{}',
+    text_result     jsonb NOT NULL DEFAULT '{}',
+    expend_tokens   int8 NOT NULL DEFAULT 0,
+    duration_seconds int8 NOT NULL DEFAULT 0,
+    epicycle_id     varchar(64) NOT NULL DEFAULT ''
+    );
 
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_tenant_time ON logs_model_op(tenant_id, created_at);
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_model_name ON logs_model_op(model_name);
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_biz_name ON logs_model_op(biz_name);
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_task_id ON logs_model_op(task_id);
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_op_type ON logs_model_op(op_type);
-CREATE INDEX IF NOT EXISTS idx_logs_model_op_deleted_at ON logs_model_op(deleted_at);
+CREATE UNIQUE INDEX IF NOT EXISTS uk_model_gateway_task_tenant_creator_task_id ON model_gateway_task (tenant_id, creator, task_id);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_task_task_id ON model_gateway_task (task_id);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_task_state ON model_gateway_task (state);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_task_deleted_at ON model_gateway_task (deleted_at);
+
+COMMENT ON TABLE model_gateway_task IS '模型网关任务表';
+COMMENT ON COLUMN model_gateway_task.id IS '主键ID';
+COMMENT ON COLUMN model_gateway_task.tenant_id IS '租户ID';
+COMMENT ON COLUMN model_gateway_task.creator IS '创建人';
+COMMENT ON COLUMN model_gateway_task.created_at IS '创建时间';
+COMMENT ON COLUMN model_gateway_task.updater IS '更新人';
+COMMENT ON COLUMN model_gateway_task.updated_at IS '更新时间';
+COMMENT ON COLUMN model_gateway_task.deleted_at IS '删除时间（软删）';
+COMMENT ON COLUMN model_gateway_task.model_name IS '模型名称';
+COMMENT ON COLUMN model_gateway_task.task_id IS '任务ID（对外返回）';
+COMMENT ON COLUMN model_gateway_task.biz_name IS '业务名称（调用方模块/系统）';
+COMMENT ON COLUMN model_gateway_task.callback_url IS '回调地址';
+COMMENT ON COLUMN model_gateway_task.state IS '0排队中/1执行中/2成功/3失败/4已下载';
+COMMENT ON COLUMN model_gateway_task.retry_count IS '已重试次数';
+COMMENT ON COLUMN model_gateway_task.phase IS '执行阶段：0模型阶段/1OSS阶段';
+COMMENT ON COLUMN model_gateway_task.tmp_file IS '临时结果文件路径';
+COMMENT ON COLUMN model_gateway_task.error_msg IS '错误信息';
+COMMENT ON COLUMN model_gateway_task.result_file IS '结果文件：{oss_file, file_type, file_size}';
+COMMENT ON COLUMN model_gateway_task.request_payload IS '请求参数（JSON）';
+COMMENT ON COLUMN model_gateway_task.text_result IS '文本类结果';
+COMMENT ON COLUMN model_gateway_task.expend_tokens IS '消耗token数';
+COMMENT ON COLUMN model_gateway_task.duration_seconds IS '耗时（秒）';
+COMMENT ON COLUMN model_gateway_task.epicycle_id IS '轮次ID';
 
-COMMENT ON TABLE logs_model_op IS '操作记录日志表(创建任务等，用于统计)';
-COMMENT ON COLUMN logs_model_op.id IS '主键ID(非自增)';
-COMMENT ON COLUMN logs_model_op.tenant_id IS '租户ID';
-COMMENT ON COLUMN logs_model_op.creator IS '创建人';
-COMMENT ON COLUMN logs_model_op.created_at IS '创建时间';
-COMMENT ON COLUMN logs_model_op.updater IS '更新人';
-COMMENT ON COLUMN logs_model_op.updated_at IS '更新时间';
-COMMENT ON COLUMN logs_model_op.deleted_at IS '删除时间(软删)';
-COMMENT ON COLUMN logs_model_op.ip IS '客户端IP';
-COMMENT ON COLUMN logs_model_op.user_agent IS 'User-Agent';
-COMMENT ON COLUMN logs_model_op.api_path IS '接口路径';
-COMMENT ON COLUMN logs_model_op.http_method IS 'HTTP方法';
-COMMENT ON COLUMN logs_model_op.biz_name IS '业务名称(调用方模块/系统)';
-COMMENT ON COLUMN logs_model_op.model_name IS '模型名称';
-COMMENT ON COLUMN logs_model_op.task_id IS '任务ID';
-COMMENT ON COLUMN logs_model_op.op_type IS '操作类型(如 createTask/getTaskResult/getTaskBatch 等)';
-COMMENT ON COLUMN logs_model_op.success IS '是否成功：1成功/0失败';
-COMMENT ON COLUMN logs_model_op.error_msg IS '错误信息(失败时)';
-COMMENT ON COLUMN logs_model_op.cost_ms IS '耗时(毫秒)';
-COMMENT ON COLUMN logs_model_op.request_payload IS '请求 JSON';
-COMMENT ON COLUMN logs_model_op.response_payload IS '响应 JSON';
 
 
 -- =========================
--- 4) logs_model_stat
+-- model_gateway_log_stat
 -- =========================
-CREATE TABLE IF NOT EXISTS logs_model_stat (
-    day DATE NOT NULL,                     -- 天(YYYY-MM-DD)
-    tenant_id BIGINT NOT NULL DEFAULT 0,   -- 租户ID
-    creator VARCHAR(64) NOT NULL DEFAULT '', -- 创建人
-    model_name VARCHAR(128) NOT NULL DEFAULT '', -- 模型名称
-    request_count BIGINT NOT NULL DEFAULT 0, -- 请求次数
-    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
-    PRIMARY KEY(day, tenant_id, creator, model_name)
-);
+CREATE TABLE IF NOT EXISTS model_gateway_log_stat (
+                                                      day             date NOT NULL,
+                                                      tenant_id       int8 NOT NULL DEFAULT 0,
+                                                      creator         varchar(64) NOT NULL DEFAULT '',
+    model_name      varchar(128) NOT NULL DEFAULT '',
+    request_count   int8 NOT NULL DEFAULT 0,
+    created_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updated_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    PRIMARY KEY (day, tenant_id, creator, model_name)
+    );
 
--- 便于时间段/租户/人/模型过滤
-CREATE INDEX IF NOT EXISTS idx_logs_model_stat_tenant_day ON logs_model_stat(tenant_id, day);
-CREATE INDEX IF NOT EXISTS idx_logs_model_stat_day ON logs_model_stat(day);
-CREATE INDEX IF NOT EXISTS idx_logs_model_stat_model_name ON logs_model_stat(model_name);
-CREATE INDEX IF NOT EXISTS idx_logs_model_stat_creator ON logs_model_stat(creator);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_log_stat_day ON model_gateway_log_stat (day);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_log_stat_creator ON model_gateway_log_stat (creator);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_log_stat_model_name ON model_gateway_log_stat (model_name);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_log_stat_tenant_day ON model_gateway_log_stat (tenant_id, day);
 
-COMMENT ON TABLE logs_model_stat IS '按天模型请求统计(用于限流/监控)';
-COMMENT ON COLUMN logs_model_stat.day IS '天(YYYY-MM-DD)';
-COMMENT ON COLUMN logs_model_stat.tenant_id IS '租户ID';
-COMMENT ON COLUMN logs_model_stat.creator IS '创建人';
-COMMENT ON COLUMN logs_model_stat.model_name IS '模型名称';
-COMMENT ON COLUMN logs_model_stat.request_count IS '请求次数';
-COMMENT ON COLUMN logs_model_stat.created_at IS '创建时间';
-COMMENT ON COLUMN logs_model_stat.updated_at IS '更新时间';
+COMMENT ON TABLE model_gateway_log_stat IS '按天统计表';
+COMMENT ON COLUMN model_gateway_log_stat.day IS '天（YYYY-MM-DD）';
+COMMENT ON COLUMN model_gateway_log_stat.tenant_id IS '租户ID';
+COMMENT ON COLUMN model_gateway_log_stat.creator IS '创建人';
+COMMENT ON COLUMN model_gateway_log_stat.model_name IS '模型名称';
+COMMENT ON COLUMN model_gateway_log_stat.request_count IS '请求次数';
+COMMENT ON COLUMN model_gateway_log_stat.created_at IS '创建时间';
+COMMENT ON COLUMN model_gateway_log_stat.updated_at IS '更新时间';
+
+
+-- =========================
+-- model_gateway_logs_op
+-- =========================
+CREATE TABLE IF NOT EXISTS model_gateway_logs_op (
+                                                     id              int8 PRIMARY KEY,
+                                                     tenant_id       int8 NOT NULL DEFAULT 0,
+                                                     creator         varchar(64) NOT NULL,
+    created_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    updater         varchar(64) NOT NULL,
+    updated_at      timestamp(6) NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    deleted_at      timestamp(6),
+    ip              varchar(64) DEFAULT '',
+    user_agent      varchar(256) DEFAULT '',
+    api_path        varchar(256) DEFAULT '',
+    http_method     varchar(16) DEFAULT '',
+    biz_name        varchar(128) NOT NULL DEFAULT '',
+    model_name      varchar(128) NOT NULL DEFAULT '',
+    task_id         varchar(64) NOT NULL DEFAULT '',
+    op_type         varchar(64) NOT NULL DEFAULT 'createTask',
+    success         int2 NOT NULL DEFAULT 1,
+    error_msg       text DEFAULT '',
+    cost_ms         int8 NOT NULL DEFAULT 0,
+    request_payload  jsonb,
+    response_payload jsonb
+    );
+
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_task_id ON model_gateway_logs_op (task_id);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_biz_name ON model_gateway_logs_op (biz_name);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_model_name ON model_gateway_logs_op (model_name);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_op_type ON model_gateway_logs_op (op_type);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_deleted_at ON model_gateway_logs_op (deleted_at);
+CREATE INDEX IF NOT EXISTS idx_model_gateway_logs_op_tenant_time ON model_gateway_logs_op (tenant_id, created_at);
+
+COMMENT ON TABLE model_gateway_logs_op IS '操作日志表';
+COMMENT ON COLUMN model_gateway_logs_op.id IS '主键ID（非自增）';
+COMMENT ON COLUMN model_gateway_logs_op.tenant_id IS '租户ID';
+COMMENT ON COLUMN model_gateway_logs_op.creator IS '创建人';
+COMMENT ON COLUMN model_gateway_logs_op.created_at IS '创建时间';
+COMMENT ON COLUMN model_gateway_logs_op.updater IS '更新人';
+COMMENT ON COLUMN model_gateway_logs_op.updated_at IS '更新时间';
+COMMENT ON COLUMN model_gateway_logs_op.deleted_at IS '删除时间（软删）';
+COMMENT ON COLUMN model_gateway_logs_op.ip IS '客户端IP';
+COMMENT ON COLUMN model_gateway_logs_op.user_agent IS 'User-Agent';
+COMMENT ON COLUMN model_gateway_logs_op.api_path IS '接口路径';
+COMMENT ON COLUMN model_gateway_logs_op.http_method IS 'HTTP方法';
+COMMENT ON COLUMN model_gateway_logs_op.biz_name IS '业务名称（调用方模块/系统）';
+COMMENT ON COLUMN model_gateway_logs_op.model_name IS '模型名称';
+COMMENT ON COLUMN model_gateway_logs_op.task_id IS '任务ID';
+COMMENT ON COLUMN model_gateway_logs_op.op_type IS '操作类型';
+COMMENT ON COLUMN model_gateway_logs_op.success IS '是否成功：1成功/0失败';
+COMMENT ON COLUMN model_gateway_logs_op.error_msg IS '错误信息（失败时）';
+COMMENT ON COLUMN model_gateway_logs_op.cost_ms IS '耗时（毫秒）';
+COMMENT ON COLUMN model_gateway_logs_op.request_payload IS '请求 JSON';
+COMMENT ON COLUMN model_gateway_logs_op.response_payload IS '响应 JSON';