1

common/auth.go

@@ -0,0 +1,34 @@
+package common
+
+import (
+	"errors"
+
+	"github.com/golang-jwt/jwt/v5"
+)
+
+const jwtSecret = "video-factory-jwt-secret-2024"
+
+type JwtClaims struct {
+	UserId  int64  `json:"user_id"`
+	Role    string `json:"role"`
+	AgentId int64  `json:"agent_id,omitempty"`
+	jwt.RegisteredClaims
+}
+
+func GetJwtSecret() string {
+	return jwtSecret
+}
+
+func ParseToken(tokenStr string) (*JwtClaims, error) {
+	token, err := jwt.ParseWithClaims(tokenStr, &JwtClaims{}, func(token *jwt.Token) (interface{}, error) {
+		return []byte(jwtSecret), nil
+	})
+	if err != nil {
+		return nil, err
+	}
+	claims, ok := token.Claims.(*JwtClaims)
+	if !ok || !token.Valid {
+		return nil, errors.New("invalid token")
+	}
+	return claims, nil
+}

common/auth_middleware.go

@@ -1,47 +1,16 @@
 package common
 
 import (
-	"errors"
 	"net/http"
 	"strings"
 
 	"github.com/gogf/gf/v2/net/ghttp"
-	"github.com/golang-jwt/jwt/v5"
 )
 
-const jwtSecret = "video-factory-jwt-secret-2024"
-
-func GetJwtSecret() string {
-	return jwtSecret
+var publicPaths = []string{
+	"/user/login",
 }
 
-type JwtClaims struct {
-	UserId  int64  `json:"user_id"`
-	Role    string `json:"role"`
-	AgentId int64  `json:"agent_id,omitempty"`
-	jwt.RegisteredClaims
-}
-
-func ParseToken(tokenStr string) (*JwtClaims, error) {
-	token, err := jwt.ParseWithClaims(tokenStr, &JwtClaims{}, func(token *jwt.Token) (interface{}, error) {
-		return []byte(jwtSecret), nil
-	})
-	if err != nil {
-		return nil, err
-	}
-	claims, ok := token.Claims.(*JwtClaims)
-	if !ok || !token.Valid {
-		return nil, errors.New("invalid token")
-	}
-	return claims, nil
-}
-
-var (
-	publicPaths = []string{
-		"/user/login",
-	}
-)
-
 func Auth(r *ghttp.Request) {
 	path := r.URL.Path
 	for _, p := range publicPaths {

common/util.go

@@ -2,6 +2,8 @@ package common
 
 import (
 	"encoding/base64"
+	"encoding/json"
+	"fmt"
 	"os"
 	"strings"
 )
@@ -34,3 +36,209 @@ func pathExt(path string) string {
 	}
 	return ""
 }
+
+// BuildSchemaRequest validates input values against a JSON schema definition,
+// fills in default values for missing optional fields,
+// and returns the result matching the schema's nested structure.
+//
+// The schema format follows test.json convention:
+//
+//	{
+//	  "section": {
+//	    "field_name": {
+//	      "type": "string|integer|number|boolean|array|object",
+//	      "required": true|false,
+//	      "default": value,
+//	      "enum": [...],
+//	      "min": number,
+//	      "max": number,
+//	      "max_chars": number,
+//	      "min_items": number,
+//	      "max_items": number
+//	    }
+//	  }
+//	}
+//
+// input is a flat map like {"prompt": "hello", "duration": 5}.
+// Fields not present in input but with a "default" in the schema are filled automatically.
+// Nodes without "type" are treated as grouping sections and recursed into.
+func BuildSchemaRequest(schema map[string]any, input map[string]any) (map[string]any, error) {
+	result := make(map[string]any)
+	for key, val := range schema {
+		fieldDef, ok := val.(map[string]any)
+		if !ok {
+			result[key] = val
+			continue
+		}
+		if _, hasType := fieldDef["type"]; hasType {
+			processed, err := processField(key, fieldDef, input)
+			if err != nil {
+				return nil, err
+			}
+			if processed != nil {
+				result[key] = processed
+			}
+			continue
+		}
+		nested, err := BuildSchemaRequest(fieldDef, input)
+		if err != nil {
+			return nil, err
+		}
+		if len(nested) > 0 {
+			result[key] = nested
+		}
+	}
+	return result, nil
+}
+
+func processField(name string, def map[string]any, input map[string]any) (any, error) {
+	fieldType, _ := def["type"].(string)
+	required, _ := def["required"].(bool)
+
+	rawVal, exists := input[name]
+	if !exists {
+		if required {
+			return nil, fmt.Errorf("'%s' is required", name)
+		}
+		if dflt, ok := def["default"]; ok {
+			return convertDefault(dflt, fieldType), nil
+		}
+		return nil, nil
+	}
+
+	switch fieldType {
+	case "string":
+		s, ok := rawVal.(string)
+		if !ok {
+			return nil, fmt.Errorf("'%s' must be a string", name)
+		}
+		if maxChars, ok := def["max_chars"].(float64); ok && len(s) > int(maxChars) {
+			return nil, fmt.Errorf("'%s' exceeds max length of %d", name, int(maxChars))
+		}
+		if enum, ok := def["enum"].([]any); ok && len(enum) > 0 {
+			if !containsValue(enum, s) {
+				return nil, fmt.Errorf("'%s' must be one of %v", name, enum)
+			}
+		}
+		return s, nil
+
+	case "integer":
+		v, err := toInt(rawVal)
+		if err != nil {
+			return nil, fmt.Errorf("'%s' must be an integer", name)
+		}
+		if minVal, ok := def["min"].(float64); ok && v < int(minVal) {
+			return nil, fmt.Errorf("'%s' must be >= %d", name, int(minVal))
+		}
+		if maxVal, ok := def["max"].(float64); ok && v > int(maxVal) {
+			return nil, fmt.Errorf("'%s' must be <= %d", name, int(maxVal))
+		}
+		return v, nil
+
+	case "number":
+		v, ok := rawVal.(float64)
+		if !ok {
+			if iv, err := toInt(rawVal); err == nil {
+				v = float64(iv)
+			} else {
+				return nil, fmt.Errorf("'%s' must be a number", name)
+			}
+		}
+		if minVal, ok := def["min"].(float64); ok && v < minVal {
+			return nil, fmt.Errorf("'%s' must be >= %v", name, minVal)
+		}
+		if maxVal, ok := def["max"].(float64); ok && v > maxVal {
+			return nil, fmt.Errorf("'%s' must be <= %v", name, maxVal)
+		}
+		return v, nil
+
+	case "boolean":
+		_, ok := rawVal.(bool)
+		if !ok {
+			return nil, fmt.Errorf("'%s' must be a boolean", name)
+		}
+		return rawVal, nil
+
+	case "array":
+		arr, ok := rawVal.([]any)
+		if !ok {
+			return nil, fmt.Errorf("'%s' must be an array", name)
+		}
+		if minItems, ok := def["min_items"].(float64); ok && len(arr) < int(minItems) {
+			return nil, fmt.Errorf("'%s' must have at least %d items", name, int(minItems))
+		}
+		if maxItems, ok := def["max_items"].(float64); ok && len(arr) > int(maxItems) {
+			return nil, fmt.Errorf("'%s' must have at most %d items", name, int(maxItems))
+		}
+		if itemsDef, ok := def["items"].(map[string]any); ok {
+			items, err := processArrayItems(arr, itemsDef)
+			if err != nil {
+				return nil, fmt.Errorf("'%s': %w", name, err)
+			}
+			return items, nil
+		}
+		return arr, nil
+	}
+
+	return rawVal, nil
+}
+
+func processArrayItems(arr []any, itemsDef map[string]any) ([]any, error) {
+	itemType, _ := itemsDef["type"].(string)
+	if itemType != "object" {
+		return arr, nil
+	}
+	props, _ := itemsDef["properties"].(map[string]any)
+	if props == nil {
+		return arr, nil
+	}
+
+	result := make([]any, len(arr))
+	for i, item := range arr {
+		itemMap, ok := item.(map[string]any)
+		if !ok {
+			result[i] = item
+			continue
+		}
+		processed, err := BuildSchemaRequest(props, itemMap)
+		if err != nil {
+			return nil, fmt.Errorf("item[%d]: %w", i, err)
+		}
+		result[i] = processed
+	}
+	return result, nil
+}
+
+func toInt(v any) (int, error) {
+	switch val := v.(type) {
+	case float64:
+		return int(val), nil
+	case int:
+		return val, nil
+	case int64:
+		return int(val), nil
+	case json.Number:
+		n, err := val.Int64()
+		return int(n), err
+	default:
+		return 0, fmt.Errorf("cannot convert %T to int", v)
+	}
+}
+
+func convertDefault(dflt any, fieldType string) any {
+	if fieldType == "integer" {
+		if f, ok := dflt.(float64); ok {
+			return int(f)
+		}
+	}
+	return dflt
+}
+
+func containsValue(arr []any, val any) bool {
+	for _, v := range arr {
+		if v == val {
+			return true
+		}
+	}
+	return false
+}

shortdrama/service/episode_service.go

@@ -10,6 +10,7 @@ import (
 	"strings"
 	"time"
 
+	"video-factory/common"
 	"video-factory/shortdrama/agent"
 	consts "video-factory/shortdrama/consts"
 	"video-factory/shortdrama/dao"
@@ -403,59 +404,6 @@ func createPendingTasks(ctx context.Context, dramaId, epId int64, script string,
 		minSingle = effectiveMax
 	}
 
-	// 从 video_schema 构建完整请求体结构（所有默认值）
-	bodyBase := map[string]any{}
-	if vs != nil {
-		if bodyDef, ok := nested(vs, "body").(map[string]any); ok {
-			for sectionKey, sectionVal := range bodyDef {
-				sectionFields, ok := sectionVal.(map[string]any)
-				if !ok {
-					continue
-				}
-				sectionOut := map[string]any{}
-				for fieldKey, fieldDef := range sectionFields {
-					fd, ok := fieldDef.(map[string]any)
-					if !ok {
-						continue
-					}
-					// parameters 支持 supported_models 过滤
-					if sectionKey == "parameters" {
-						if sm, ok := fd["supported_models"]; ok {
-							if models, ok := sm.([]any); ok {
-								found := false
-								for _, m := range models {
-									if ms, ok := m.(string); ok && ms == modelCfg.ModelName {
-										found = true
-										break
-									}
-								}
-								if !found {
-									continue
-								}
-							}
-						}
-					}
-					// 始终包含 schema 中定义的所有字段（无 default 的字段留 nil，后续覆盖）
-					if def, ok := fd["default"]; ok {
-						sectionOut[fieldKey] = def
-					} else {
-						sectionOut[fieldKey] = nil
-					}
-				}
-				bodyBase[sectionKey] = sectionOut
-			}
-		}
-		// 尺寸映射：从 drama 的 Resolution/AspectRatio 查 video_schema
-		if sizes := nested(vs, "body", "parameters", "size", "sizes"); sizes != nil {
-			if sm, ok := sizes.(map[string]any); ok {
-				if v := resolveSizeFromSchema(sm, d.Resolution, d.AspectRatio); v != "" {
-					if params, ok := bodyBase["parameters"].(map[string]any); ok {
-						params["size"] = v
-					}
-				}
-			}
-		}
-	}
 	// ============ 加载参考素材 ============
 	type _namedRef struct {
 		path string
@@ -596,38 +544,39 @@ func createPendingTasks(ctx context.Context, dramaId, epId int64, script string,
 	}
 
 	// ============ 插入任务 ============
+	bodyDef, _ := nested(vs, "body").(map[string]any)
 	for i, tg := range taskGroups {
-		// 深拷贝 bodyBase，覆盖动态字段
-		body := make(map[string]any, len(bodyBase))
-		for bk, bv := range bodyBase {
-			if section, ok := bv.(map[string]any); ok {
-				nested := make(map[string]any, len(section))
-				for sk, sv := range section {
-					nested[sk] = sv
-				}
-				body[bk] = nested
-			} else {
-				body[bk] = bv
-			}
+		segInput := map[string]any{
+			"prompt":   buildFinalPrompt(tg.promptText),
+			"duration": tg.dur,
+			"seed":     seed,
 		}
-		body["model"] = modelCfg.ModelName
-		setSchemaField(body, "prompt", buildFinalPrompt(tg.promptText))
 		if negativePrompt != "" {
-			setSchemaField(body, "negative_prompt", negativePrompt)
+			segInput["negative_prompt"] = negativePrompt
 		}
-		setSchemaField(body, "duration", tg.dur)
-		// 参考素材
 		if len(refURLs) > 0 {
 			_, mediaTypeField, mediaURLField, _ := resolveMediaFields(modelCfg.Schema, modelCfg.FirstFrameMapping)
-			setSchemaField(body, "media", buildSchemaMedia(refURLs, mediaTypeField, mediaURLField))
-			setSchemaField(body, "reference_urls", refURLs)
-		} else if getSchemaField(body, "media") == nil {
-			setSchemaField(body, "media", []any{})
+			segInput["media"] = buildSchemaMedia(refURLs, mediaTypeField, mediaURLField)
+			refsAny := make([]any, len(refURLs))
+			for j, u := range refURLs {
+				refsAny[j] = u
+			}
+			segInput["reference_urls"] = refsAny
 		}
-		if seed > 0 {
-			setSchemaField(body, "seed", seed)
+		body, _ := common.BuildSchemaRequest(bodyDef, segInput)
+		if body == nil {
+			body = map[string]any{}
+		}
+		body["model"] = modelCfg.ModelName
+		if sizes := nested(vs, "body", "parameters", "size", "sizes"); sizes != nil {
+			if sm, ok := sizes.(map[string]any); ok {
+				if v := resolveSizeFromSchema(sm, d.Resolution, d.AspectRatio); v != "" {
+					if params, ok := body["parameters"].(map[string]any); ok {
+						params["size"] = v
+					}
+				}
+			}
 		}
-		// 非第0段添加首帧占位（URL 为空），第0段不需要首帧参考
 		if i > 0 {
 			injectFirstFrame(body, "", modelCfg.FirstFrameMapping, modelCfg.Schema)
 		}

shortdrama/service/generation_service.go

@@ -783,7 +783,7 @@ func (s *generationService) ListEpisodeTasks(ctx context.Context, epId int64) ([
 	return dao.GenerationTask.ListByEpisode(ctx, epId)
 }
 
-// createVideoTask 调用视频生成API提交任务，返回 (taskID, requestBodyJSON, error)
+// submitVideoGenRequest 构建请求体并调用视频生成API，返回 (taskID, requestBodyJSON, error)
 
 func (s *generationService) mergeEpisodeVideo(ctx context.Context, epId int64, lastTaskId int64) error {
 	tasks, err := dao.GenerationTask.ListByEpisode(ctx, epId)
@@ -2118,7 +2118,7 @@ func (s *generationService) submitVideoTask(ctx context.Context, d *entity.Drama
 
 	// 固定 seed 确保各段画风/人物形象一致（基于短剧ID+剧集序号）
 	seed := int(d.Id)*1000 + ep.Index*10
-	taskId, requestJSON, err := createVideoTask(ctx, modelCfg.ApiKey, modelCfg.BaseUrl, modelCfg.ModelName,
+	taskId, requestJSON, err := submitVideoGenRequest(ctx, modelCfg.ApiKey, modelCfg.BaseUrl, modelCfg.ModelName,
 		prompt, negativePrompt, refURLs, effectiveDur, d.Resolution, d.AspectRatio, modelCfg.Schema, seed, segIdx, genTaskId, ep.Id, d.Title, ep.Title, modelCfg.FirstFrameMapping)
 	if err != nil {
 		return "", "", fmt.Errorf("video composition request failed: %w", err)
@@ -2178,78 +2178,53 @@ func callVideoAPI(ctx context.Context, apiKey, baseURL string, payload []byte, s
 //
 //	params: 额外参数（如 audio/shot_type/watermark 等）
 //	sizes:  分辨率→尺寸字符串映射表
-func createVideoTask(ctx context.Context, apiKey, baseURL, modelName, prompt, negativePrompt string, refURLs []string, duration int, resolution, aspectRatio, videoSchema string, seed int, segIdx int, genTaskId int64, episodeId int64, dramaTitle, epTitle, firstFrameMapping string) (string, string, error) {
-	// 从 schema body 构建请求体基础结构（vs 只解析一次）
-	body := map[string]any{}
+func submitVideoGenRequest(ctx context.Context, apiKey, baseURL, modelName, prompt, negativePrompt string, refURLs []string, duration int, resolution, aspectRatio, videoSchema string, seed int, segIdx int, genTaskId int64, episodeId int64, dramaTitle, epTitle, firstFrameMapping string) (string, string, error) {
 	var vs map[string]any
 	if videoSchema != "" {
 		json.Unmarshal([]byte(videoSchema), &vs)
 	}
-	if bodyDef, ok := nested(vs, "body").(map[string]any); ok {
-		for sectionKey, sectionVal := range bodyDef {
-			sectionFields, ok := sectionVal.(map[string]any)
-			if !ok {
-				continue
-			}
-			sectionOut := map[string]any{}
-			for fieldKey, fieldDef := range sectionFields {
-				fd, ok := fieldDef.(map[string]any)
-				if !ok {
-					continue
-				}
-				// parameters 支持 supported_models 过滤
-				if sectionKey == "parameters" {
-					if sm, ok := fd["supported_models"]; ok {
-						if models, ok := sm.([]any); ok {
-							found := false
-							for _, m := range models {
-								if ms, ok := m.(string); ok && ms == modelName {
-									found = true
-									break
-								}
-							}
-							if !found {
-								continue
-							}
-						}
-					}
-				}
-				// 始终包含 schema 中定义的所有字段（无 default 的字段留 nil，后续覆盖）
-				if def, ok := fd["default"]; ok {
-					sectionOut[fieldKey] = def
-				} else {
-					sectionOut[fieldKey] = nil
-				}
-			}
-			body[sectionKey] = sectionOut
-		}
-	}
-	// 动态字段覆盖 — setSchemaField 按字段名自动匹配 section
-	body["model"] = modelName
-	setSchemaField(body, "prompt", prompt)
-	if negativePrompt != "" {
-		setSchemaField(body, "negative_prompt", negativePrompt)
-	}
+
 	_, mediaTypeField, mediaURLField, _ := resolveMediaFields(videoSchema, firstFrameMapping)
 
-	if len(refURLs) > 0 {
-		setSchemaField(body, "media", buildSchemaMedia(refURLs, mediaTypeField, mediaURLField))
-		setSchemaField(body, "reference_urls", refURLs)
+	// 通过 BuildSchemaRequest 构建请求体：flat input → schema 校验/填充 → 嵌套结构
+	input := map[string]any{
+		"prompt": prompt,
+		"seed":   seed,
+	}
+	if negativePrompt != "" {
+		input["negative_prompt"] = negativePrompt
 	}
 	if duration > 0 {
-		setSchemaField(body, "duration", duration)
+		input["duration"] = duration
 	}
-	if seed > 0 {
-		setSchemaField(body, "seed", seed)
+	if len(refURLs) > 0 {
+		input["media"] = buildSchemaMedia(refURLs, mediaTypeField, mediaURLField)
+		refsAny := make([]any, len(refURLs))
+		for i, u := range refURLs {
+			refsAny[i] = u
+		}
+		input["reference_urls"] = refsAny
 	}
 
+	bodyDef, _ := nested(vs, "body").(map[string]any)
+	body, err := common.BuildSchemaRequest(bodyDef, input)
+	if err != nil {
+		return "", "", fmt.Errorf("build request body failed: %w", err)
+	}
+	if body == nil {
+		body = map[string]any{}
+	}
+	body["model"] = modelName
+
 	// 统一处理首帧（segIdx==0 删除，>0 解析URL）
 	prepareFirstFrame(ctx, body, segIdx, genTaskId, episodeId, dramaTitle, epTitle, firstFrameMapping, videoSchema)
 	// 尺寸映射：从 drama 的 Resolution/AspectRatio 查 video_schema
 	if sizes := nested(vs, "body", "parameters", "size", "sizes"); sizes != nil {
 		if sm, ok := sizes.(map[string]any); ok {
 			if v := resolveSizeFromSchema(sm, resolution, aspectRatio); v != "" {
-				setSchemaField(body, "size", v)
+				if params, ok := body["parameters"].(map[string]any); ok {
+					params["size"] = v
+				}
 			}
 		}
 	}