feat: 新增操作日志、任务分页查询与模型失败重试优化

- 新增操作日志表（asynch_op_log）及对应DAO，记录任务创建等操作的审计信息
- 新增任务分页查询接口（ListTask）及对应DTO、Service和DAO方法
- 优化模型调用失败重试逻辑：支持配置重试排队策略（插队到队首或队尾）
- 新增临时文件存储机制，当模型调用成功但OSS上传失败时，下次仅重试OSS上传
- 模型配置新增retry_queue_max_seconds字段，控制失败重试排队策略
- 更新数据库表结构（asynch_models、asynch_task、新增asynch_op_log）及同步更新SQL
- 配置文件调整：超时单位改为秒，更新服务地址和轮询间隔
- 修复模型列表查询支持按名称模糊搜索

dao/model_dao.go

@@ -73,8 +73,11 @@ func (d *modelDao) GetByID(ctx context.Context, id int64) (m *entity.AsynchModel
 	return
 }
 
-func (d *modelDao) List(ctx context.Context, pageNum, pageSize int) (list []*entity.AsynchModel, total int64, err error) {
-	model := gfdb.DB(ctx).Model(ctx, public.TableNameModel).OrderDesc(entity.AsynchModelCol.CreatedAt)
+func (d *modelDao) List(ctx context.Context, pageNum, pageSize int, modelNameLike string) (list []*entity.AsynchModel, total int64, err error) {
+	model := gfdb.DB(ctx).Model(ctx, public.TableNameModel).Where("deleted_at IS NULL").OrderDesc(entity.AsynchModelCol.CreatedAt)
+	if modelNameLike != "" {
+		model = model.WhereLike(entity.AsynchModelCol.ModelName, "%"+modelNameLike+"%")
+	}
 	if pageNum > 0 && pageSize > 0 {
 		model = model.Page(pageNum, pageSize)
 	}
@@ -86,4 +89,3 @@ func (d *modelDao) List(ctx context.Context, pageNum, pageSize int) (list []*ent
 	err = r.Structs(&list)
 	return
 }
-

dao/op_log_dao.go

@@ -0,0 +1,22 @@
+package dao
+
+import (
+	"context"
+
+	"model-asynch/consts/public"
+	"model-asynch/model/entity"
+
+	"gitea.com/red-future/common/db/gfdb"
+)
+
+type opLogDao struct{}
+
+var OpLog = &opLogDao{}
+
+func (d *opLogDao) Insert(ctx context.Context, log *entity.AsynchOpLog) (id int64, err error) {
+	r, err := gfdb.DB(ctx).Model(ctx, public.TableNameOpLog).Data(log).Insert()
+	if err != nil {
+		return 0, err
+	}
+	return r.LastInsertId()
+}

dao/task_dao.go

@@ -11,6 +11,7 @@ import (
 	"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{}
@@ -136,6 +137,31 @@ func (d *taskDao) CountActiveByModel(ctx context.Context, modelName string) (int
 	return int64(n), 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
+}
+
 // ClaimPending 抢占 pending 任务（state=0），并在同一事务中更新为 running（state=1）
 // 使用 PostgreSQL: FOR UPDATE SKIP LOCKED 避免多 worker 重复消费
 func (d *taskDao) ClaimPending(ctx context.Context, batchSize int) (tasks []*entity.AsynchTask, err error) {

dao/task_dao_bg.go

@@ -20,7 +20,7 @@ func (d *taskDao) ClaimPendingGlobal(ctx context.Context, batchSize int) (tasks
 	}
 	err = gfdb.DB(ctx).Transaction(ctx, func(ctx context.Context, tx gdb.TX) error {
 		sql := fmt.Sprintf(
-			`SELECT id, tenant_id, model_name, task_id, input_ref, request_payload
+			`SELECT id, tenant_id, model_name, task_id, input_ref, request_payload, phase, tmp_file
 			   FROM %s
 			  WHERE deleted_at IS NULL AND state = 0
 			  ORDER BY enqueue_at ASC
@@ -58,7 +58,7 @@ func (d *taskDao) ClaimPendingGlobal(ctx context.Context, batchSize int) (tasks
 func (d *taskDao) UpdateSuccessGlobal(ctx context.Context, id int64, ossFile, fileType string, fileSize int64, expireAt *gtime.Time) error {
 	now := gtime.Now()
 	_, err := gfdb.DB(ctx).Exec(ctx,
-		fmt.Sprintf(`UPDATE %s SET state=2, oss_file=?, file_type=?, file_size=?, error_msg='', finished_at=?, expire_at=NULL, updated_at=? WHERE id=?`, public.TableNameTask),
+		fmt.Sprintf(`UPDATE %s SET state=2, oss_file=?, file_type=?, file_size=?, error_msg='', finished_at=?, expire_at=NULL, phase=0, tmp_file='', updated_at=? WHERE id=?`, public.TableNameTask),
 		ossFile, fileType, fileSize, now, now, id,
 	)
 	return err
@@ -67,12 +67,31 @@ func (d *taskDao) UpdateSuccessGlobal(ctx context.Context, id int64, ossFile, fi
 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=?, updated_at=? WHERE id=?`, public.TableNameTask),
+		fmt.Sprintf(`UPDATE %s SET state=3, error_msg=?, finished_at=?, phase=0, tmp_file='', updated_at=? WHERE id=?`, public.TableNameTask),
 		errorMsg, 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) SoftDeleteByTaskIDGlobal(ctx context.Context, taskID string) error {
 	_, err := gfdb.DB(ctx).Exec(ctx,
 		fmt.Sprintf(`UPDATE %s SET deleted_at=NOW(), updated_at=NOW() WHERE task_id=? AND deleted_at IS NULL`, public.TableNameTask),
@@ -113,7 +132,8 @@ func (d *taskDao) ListFailedRetryableGlobal(ctx context.Context, limit int) (lis
 	}
 	r, err := gfdb.DB(ctx).GetAll(ctx,
 		fmt.Sprintf(`
-SELECT t.*
+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
@@ -132,11 +152,13 @@ SELECT t.*
 	return
 }
 
-// RequeueForRetryGlobal 将任务重新入队（state=0，enqueue_at=now），并将 retry_count +1
-func (d *taskDao) RequeueForRetryGlobal(ctx context.Context, id int64) error {
+// 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=NOW(), updated_at=NOW() WHERE id=? AND state=3 AND deleted_at IS NULL`, public.TableNameTask),
-		id,
+		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
 }