feat: 支持视频模型首帧参数及顺序执行逻辑

添加 FirstFrame 字段到模型信息结构体，重构视频节点数据转换处理，支持从上游节点获取视频配置并区分输出字段。实现视频结果重新上传至 OSS，修正字段匹配逻辑为精确匹配，并基于 return_last_frame 参数优化顺序执行中的首帧与内容传递。

workflow/model/dto/flow/flow_execution_dto.go

@@ -49,6 +49,7 @@ type GetModelInfoReq struct {
 
 type GetModelInfoRes struct {
 	Model struct {
+		FirstFrame         string         `json:"firstFrame"`
 		LastFrame          string         `json:"lastFrame"`
 		ResponseTokenField string         `json:"responseTokenField"`
 		ResponseMapping    map[string]any `json:"responseMapping"`

workflow/service/flow/lambda_node.go

@@ -11,6 +11,7 @@ import (
 	flowDto "ai-agent/workflow/model/dto/flow"
 	"context"
 	"fmt"
+	"path/filepath"
 	"strconv"
 	"strings"
 	"sync"
@@ -292,7 +293,23 @@ func VideoModelLambda(ctx context.Context, input any) (any, error) {
 		}
 		videoUrl = newS + msg.FileURL
 	} else {
-		videoUrl = videoURL[0]
+		var bytes []byte
+		bytes, err = GetFileBytesFromURL(ctx, videoURL[0])
+		if err != nil {
+			return nil, fmt.Errorf("下载图片失败: %w", err)
+		}
+		// 构造文件名
+		fileName := fmt.Sprintf("ai_video_%d%s", time.Now().UnixMilli(), strings.ToLower(filepath.Ext(videoURL[0])))
+		// 上传到你的OSS（你项目已有的Upload方法）
+		var upResp *dto.UploadFileBytesRes
+		upResp, err = Upload(ctx, &dto.UploadFileBytesReq{
+			FileName:  fileName,
+			FileBytes: bytes,
+		})
+		if err != nil {
+			return nil, fmt.Errorf("上传OSS失败: %w", err)
+		}
+		videoUrl = upResp.FileURL
 	}
 
 	outputRes := make([]node.NodeFormField, 0)

workflow/service/flow/lambda_node_imp.go

@@ -788,18 +788,41 @@ func VideoOptimizeNode(ctx context.Context, nodeInput *flowDto.NodeExecutionInpu
 }
 
 func DataConversionNode(ctx context.Context, nodeInput *flowDto.NodeExecutionInput, skillName string, form []map[string]any, userForm []map[string]any) ([]node.NodeFormField, error) {
-	if strings.Contains(nodeInput.Config.Name, "字幕") {
+	if strings.Contains(nodeInput.Config.Name, "字幕") || strings.Contains(nodeInput.Config.Name, "视频") {
 		jsonStr := ``
-		for _, field := range nodeInput.Config.OutputConfig {
-			jsonStr, _ = sjson.Set(jsonStr, field.Field, field.Value)
-		}
 		outputRes := make([]node.NodeFormField, 0)
-		outputRes = append(outputRes, node.NodeFormField{
-			Field: fmt.Sprintf("data_conversion"),
-			Value: gconv.Map(jsonStr),
-			Label: fmt.Sprintf("data_conversion"),
-			Type:  "string",
-		})
+		for _, field := range nodeInput.Config.OutputConfig {
+			if strings.Contains(nodeInput.Config.Name, "视频") {
+				for _, item := range nodeInput.Global.ExecutedNodes {
+					refNode, ok := nodeInput.Global.ConfigMap[item.NodeId]
+					if !ok {
+						continue
+					}
+					for _, v := range refNode.FormConfig {
+						if v.Field == field.Value {
+							jsonStr, _ = sjson.Set(jsonStr, field.Field, v.Value)
+						}
+					}
+				}
+			} else {
+				jsonStr, _ = sjson.Set(jsonStr, field.Field, field.Value)
+			}
+		}
+		if strings.Contains(nodeInput.Config.Name, "视频") {
+			outputRes = append(outputRes, node.NodeFormField{
+				Field: fmt.Sprintf("data_content"),
+				Value: gconv.Map(jsonStr),
+				Label: fmt.Sprintf("data_content"),
+				Type:  "string",
+			})
+		} else {
+			outputRes = append(outputRes, node.NodeFormField{
+				Field: fmt.Sprintf("data_conversion"),
+				Value: gconv.Map(jsonStr),
+				Label: fmt.Sprintf("data_conversion"),
+				Type:  "string",
+			})
+		}
 		return outputRes, nil
 	}
 
@@ -1035,7 +1058,7 @@ func GetNodeContextContent(execInput *flowDto.FlowExecutionInput, nodeEntity *en
 				// 取指定字段
 				for _, f := range source.Field {
 					for _, v := range refNode.FormConfig {
-						if strings.Contains(v.Label, f) {
+						if v.Label == f {
 							input = append(input, v)
 						}
 					}

workflow/service/flow/lambda_node_util.go

@@ -299,7 +299,22 @@ func GetModelResult(ctx context.Context, sessionId string, nodeInput *flowDto.No
 	if !nodeInput.Global.IsDialogue {
 		sessionId = ""
 	}
-
+	needSequential := false
+	for _, item := range userForm {
+		if g.NewVar(item).IsMap() {
+			valMap := gconv.Map(item)
+			for _, v := range valMap {
+				if g.NewVar(v).IsMap() {
+					vv := gconv.Map(v)
+					for kk, vvv := range vv {
+						if kk == "return_last_frame" {
+							needSequential = vvv.(bool)
+						}
+					}
+				}
+			}
+		}
+	}
 	composeResult, err := GetComposeResult(ctx, buildType, nodeInput.Config.ModelConfig.ModelName, nodeInput.Config.PromptContent, skillName, form, userForm, nodeInput.Global.FileUrl, sessionId, nodeInput.Config.Id, nodeInput.Config.Name)
 	if err != nil {
 		return nil, err
@@ -316,13 +331,12 @@ func GetModelResult(ctx context.Context, sessionId string, nodeInput *flowDto.No
 	mapTaskResult = make([]map[string]any, len(composeResult.Messages.Rounds))
 	var taskResultMap map[string]any
 
-	needSequential := false
 	if buildType == 1 {
 		if needSequential {
 			for idx, item := range composeResult.Messages.Rounds {
 				if !g.IsEmpty(taskResultMap) {
 					var set string
-					set, err = sjson.Set(gconv.String(item), modelInfo.Model.LastFrame, gconv.String(taskResultMap[modelInfo.Model.ResponseBody]))
+					set, err = sjson.Set(gconv.String(item), modelInfo.Model.FirstFrame, gconv.String(taskResultMap["content"]))
 					if err != nil {
 						return nil, err
 					}
@@ -339,7 +353,7 @@ func GetModelResult(ctx context.Context, sessionId string, nodeInput *flowDto.No
 				}
 
 				if nodeInput.Config.NodeCode == node.NodeTypeVideoModel {
-					ext := GetFileTypeByPath(gconv.String(taskResult[modelInfo.Model.ResponseBody]))
+					ext := GetFileTypeByPath(gconv.String(taskResult["content"]))
 					if ext == "image" {
 						taskResultMap = taskResult
 					} else {