fix(resource): 签发本地公开资源直链时追加文件扩展名，并对齐 AutoDL 工作流参数 (#341)

Co-authored-by: Danborad <danborad@github.com>
2026-08-29 19:31:48 +08:00
parent 80d2aa6882
commit aa804d43b1
9 changed files with 574 additions and 72 deletions
--- a/backend/internal/handler/user_data.go
+++ b/backend/internal/handler/user_data.go
@@ -312,7 +312,7 @@ func RegisterUserDataRoutes(r *gin.RouterGroup, svc *service.Service) {
 		}
 		c.DataFromReader(stream.StatusCode, stream.ContentLength, resource.MimeType, stream.Body, nil)
 	})
-	r.GET("/public/resources/:id/file", func(c *gin.Context) {
+	publicResourceHandler := func(c *gin.Context) {
 		stream, err := svc.OpenPublicResourceRange(c.Param("id"), c.Query("expires"), c.Query("signature"), c.GetHeader("Range"))
 		if err != nil {
 			failService(c, err)
@@ -336,7 +336,9 @@ func RegisterUserDataRoutes(r *gin.RouterGroup, svc *service.Service) {
 			return
 		}
 		c.DataFromReader(stream.StatusCode, stream.ContentLength, resource.MimeType, stream.Body, nil)
-	})
+	}
+	r.GET("/public/resources/:id/file", publicResourceHandler)
+	r.GET("/public/resources/:id/file/:filename", publicResourceHandler)
 	r.GET("/assets", func(c *gin.Context) {
 		user, err := currentUser(c, svc)
 		if err != nil {
--- a/backend/internal/protocol/autodl_plugin_artifact_test.go
+++ b/backend/internal/protocol/autodl_plugin_artifact_test.go
@@ -21,11 +21,20 @@ func TestAutoDLPluginArtifact(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	if len(adapters) != 1 {
+	if len(adapters) == 0 {
 		t.Fatalf("loaded adapters = %d", len(adapters))
 	}
-	adapter := adapters[0]
-	if adapter.Metadata().ID != "autodl-comfyui" || adapter.Metadata().Version != "1.0.2" || adapter.Metadata().Execution != "declarative" {
+	var adapter Adapter
+	for _, a := range adapters {
+		if a.Metadata().ID == "autodl-comfyui" {
+			adapter = a
+			break
+		}
+	}
+	if adapter == nil {
+		t.Fatal("autodl-comfyui provider adapter not found")
+	}
+	if adapter.Metadata().ID != "autodl-comfyui" || adapter.Metadata().Execution != "declarative" {
 		t.Fatalf("metadata = %#v", adapter.Metadata())
 	}
 	if !adapter.Metadata().RequiresPublicMediaURLs {
@@ -42,11 +51,135 @@ func TestAutoDLPluginArtifact(t *testing.T) {
 		t.Fatalf("create spec = %#v", create)
 	}
 	body, ok := create.Body.(map[string]any)
-	if !ok || body["prompt"] != "a cinematic test" || body["duration"] != 3 || body["resolution"] != "768p竖" || body["ref_image_0"] != "https://cdn.example/reference.png" {
+	if !ok || body["prompt"] != "a cinematic test" || body["duration"] != 3 || body["resolution"] != "768p竖" {
 		t.Fatalf("create body = %#v", create.Body)
 	}
-	if _, exists := body["ref_image_1"]; exists {
-		t.Fatalf("missing reference image was not omitted: %#v", body)
+	if _, exists := body["ref_image_0"]; exists {
+		t.Fatalf("no_pic workflow must not send ref_image_0: %#v", body)
+	}
+	if _, exists := body["first_frame"]; exists {
+		t.Fatalf("no_pic workflow must not send first_frame: %#v", body)
+	}
+
+	createV5, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_lightx2v_v5_15s", Prompt: "a cinematic test", Duration: 15, Resolution: "768p竖", Images: []MediaReference{{URL: "https://cdn.example/ref0.png"}, {URL: "https://cdn.example/ref1.png"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyV5 := createV5.Body.(map[string]any)
+	if bodyV5["ref_image_0"] != "https://cdn.example/ref0.png" || bodyV5["ref_image_1"] != "https://cdn.example/ref1.png" {
+		t.Fatalf("v5 create body missing ref images = %#v", bodyV5)
+	}
+	if _, exists := bodyV5["first_frame"]; exists {
+		t.Fatalf("v5 workflow must not send first_frame: %#v", bodyV5)
+	}
+
+	createLightX2V, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_lightx2v", Prompt: "a cinematic test", Duration: 5, Resolution: "768p竖", Images: []MediaReference{{URL: "https://cdn.example/first.png"}, {URL: "https://cdn.example/last.png"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyLightX2V := createLightX2V.Body.(map[string]any)
+	if bodyLightX2V["first_frame"] != "https://cdn.example/first.png" || bodyLightX2V["last_frame"] != "https://cdn.example/last.png" {
+		t.Fatalf("lightx2v body = %#v", bodyLightX2V)
+	}
+	if _, exists := bodyLightX2V["ref_image_0"]; exists {
+		t.Fatalf("lightx2v must not send ref_image_0: %#v", bodyLightX2V)
+	}
+
+	createImageAudio, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_image_audio_to_video", Prompt: "should be omitted", Duration: 5, Resolution: "768p横",
+		Images: []MediaReference{{URL: "https://cdn.example/face.png"}},
+		Audios: []MediaReference{{URL: "https://cdn.example/voice.mp3"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyImageAudio := createImageAudio.Body.(map[string]any)
+	if bodyImageAudio["ref_image_0"] != "https://cdn.example/face.png" || bodyImageAudio["ref_audio_0"] != "https://cdn.example/voice.mp3" || bodyImageAudio["audio_duration"] != 5 || bodyImageAudio["resolution"] != "768p横" {
+		t.Fatalf("image_audio body = %#v", bodyImageAudio)
+	}
+	if _, exists := bodyImageAudio["prompt"]; exists {
+		t.Fatalf("image_audio workflow must not send prompt: %#v", bodyImageAudio)
+	}
+	if _, exists := bodyImageAudio["duration"]; exists {
+		t.Fatalf("image_audio workflow must not send duration: %#v", bodyImageAudio)
+	}
+	if _, exists := bodyImageAudio["ref_image_1"]; exists {
+		t.Fatalf("image_audio workflow must not send ref_image_1: %#v", bodyImageAudio)
+	}
+	if _, exists := bodyImageAudio["ref_audio_1"]; exists {
+		t.Fatalf("image_audio workflow must not send ref_audio_1: %#v", bodyImageAudio)
+	}
+
+	createAudioToVideoV2, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_image_audio_to_video_v2", Prompt: "cinematic sync", Duration: 8, Resolution: "1080p横",
+		Images: []MediaReference{{URL: "https://cdn.example/img0.png"}, {URL: "https://cdn.example/img1.png"}},
+		Audios: []MediaReference{{URL: "https://cdn.example/aud0.mp3"}, {URL: "https://cdn.example/aud1.mp3"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyV2 := createAudioToVideoV2.Body.(map[string]any)
+	if bodyV2["prompt"] != "cinematic sync" || bodyV2["duration"] != 8 || bodyV2["resolution"] != "1080p横" ||
+		bodyV2["ref_image_0"] != "https://cdn.example/img0.png" || bodyV2["ref_image_1"] != "https://cdn.example/img1.png" ||
+		bodyV2["ref_audio_0"] != "https://cdn.example/aud0.mp3" || bodyV2["ref_audio_1"] != "https://cdn.example/aud1.mp3" {
+		t.Fatalf("v2 body = %#v", bodyV2)
+	}
+	if _, exists := bodyV2["audio_duration"]; exists {
+		t.Fatalf("v2 must not send audio_duration: %#v", bodyV2)
+	}
+
+	createB99001, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_b99_001", Prompt: "a prompt test", Duration: 5, Resolution: "736p竖",
+		Images: []MediaReference{{URL: "https://cdn.example/should-be-omitted.png"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyB99001 := createB99001.Body.(map[string]any)
+	if bodyB99001["prompt"] != "a prompt test" || bodyB99001["duration"] != 5 || bodyB99001["resolution"] != "736p竖" {
+		t.Fatalf("b99_001 body = %#v", bodyB99001)
+	}
+	if _, exists := bodyB99001["ref_image_0"]; exists {
+		t.Fatalf("b99_001 must not send ref_image_0: %#v", bodyB99001)
+	}
+	if _, exists := bodyB99001["first_frame"]; exists {
+		t.Fatalf("b99_001 must not send first_frame: %#v", bodyB99001)
+	}
+
+	createB99002, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_b99_002", Prompt: "first to last", Duration: 6, Resolution: "736p横",
+		Images: []MediaReference{{URL: "https://cdn.example/f0.png"}, {URL: "https://cdn.example/f1.png"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyB99002 := createB99002.Body.(map[string]any)
+	if bodyB99002["prompt"] != "first to last" || bodyB99002["duration"] != 6 || bodyB99002["resolution"] != "736p横" ||
+		bodyB99002["first_frame"] != "https://cdn.example/f0.png" || bodyB99002["last_frame"] != "https://cdn.example/f1.png" {
+		t.Fatalf("b99_002 body = %#v", bodyB99002)
+	}
+	if _, exists := bodyB99002["ref_image_0"]; exists {
+		t.Fatalf("b99_002 must not send ref_image_0: %#v", bodyB99002)
+	}
+
+	createB99003, err := adapter.BuildCreate(context.Background(), RequestContext{BaseURL: "https://autodl.art", Request: GenerationRequest{
+		Model: "minimax_h3_b99_003_12s", Prompt: "multi image 12s", Duration: 12, Resolution: "736p(1:1)",
+		Images: []MediaReference{{URL: "https://cdn.example/img0.png"}, {URL: "https://cdn.example/img1.png"}},
+	}})
+	if err != nil {
+		t.Fatal(err)
+	}
+	bodyB99003 := createB99003.Body.(map[string]any)
+	if bodyB99003["prompt"] != "multi image 12s" || bodyB99003["duration"] != 12 || bodyB99003["resolution"] != "736p(1:1)" ||
+		bodyB99003["ref_image_0"] != "https://cdn.example/img0.png" || bodyB99003["ref_image_1"] != "https://cdn.example/img1.png" {
+		t.Fatalf("b99_003 body = %#v", bodyB99003)
+	}
+	if _, exists := bodyB99003["first_frame"]; exists {
+		t.Fatalf("b99_003 must not send first_frame: %#v", bodyB99003)
 	}

 	failed, err := adapter.ParseCreate(context.Background(), []byte(`{"code":"RequestParameterIsWrong","data":null,"msg":"参数: resolution 的值: 768P 不在 options 列表中"}`))
--- a/backend/internal/protocol/builtin_test.go
+++ b/backend/internal/protocol/builtin_test.go
@@ -372,7 +372,7 @@ func TestDeclarativeManifestKeepsLegacyVideoResolutionTransformCompatible(t *tes
 		"768P横": "768p横",
 	}
 	for input, expected := range tests {
-		if actual := applyManifestTransform(input, "video-resolution"); actual != expected {
+		if actual := applyManifestTransform(input, "video-resolution", GenerationRequest{}); actual != expected {
 			t.Errorf("legacy video resolution transform %q = %#v, want %q", input, actual, expected)
 		}
 	}
--- a/backend/internal/protocol/manifest.go
+++ b/backend/internal/protocol/manifest.go
@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net/http"
 	"net/url"
+	"regexp"
 	"strconv"
 	"strings"
 )
@@ -440,10 +441,38 @@ func (a manifestAdapter) parse(payload map[string]any, c PollContext) CreateResu
 	return CreateResult{TaskID: id, Status: status, Result: result, Message: message}
 }

+var (
+	manifestMDImageRegex   = regexp.MustCompile(`!\[[^\]]*\]\(([^)\s]+)\)`)
+	manifestHTMLVideoRegex = regexp.MustCompile(`(?i)<video[^>]+src=['"]([^'"]+)['"]`)
+)
+
 func mediaPathValues(payload map[string]any, path string) []string {
 	value := pathValue(payload, path)
 	if text, ok := value.(string); ok && strings.TrimSpace(text) != "" {
-		return []string{strings.TrimSpace(text)}
+		trimmed := strings.TrimSpace(text)
+		if matches := manifestMDImageRegex.FindAllStringSubmatch(trimmed, -1); len(matches) > 0 {
+			var res []string
+			for _, m := range matches {
+				if len(m) > 1 && m[1] != "" {
+					res = append(res, m[1])
+				}
+			}
+			if len(res) > 0 {
+				return res
+			}
+		}
+		if matches := manifestHTMLVideoRegex.FindAllStringSubmatch(trimmed, -1); len(matches) > 0 {
+			var res []string
+			for _, m := range matches {
+				if len(m) > 1 && m[1] != "" {
+					res = append(res, m[1])
+				}
+			}
+			if len(res) > 0 {
+				return res
+			}
+		}
+		return []string{trimmed}
 	}
 	items, ok := value.([]any)
 	if !ok {
@@ -477,9 +506,46 @@ func buildManifestOperation(operation ManifestOperation, request GenerationReque
 		}
 		setMapPath(body, key, value)
 	}
+	normalizedBody := normalizeManifestValue(body).(map[string]any)
 	path := strings.ReplaceAll(operation.Path, "{{taskId}}", url.PathEscape(taskID))
 	path = strings.ReplaceAll(path, "{{model}}", url.PathEscape(request.Model))
-	return RequestSpec{Method: strings.ToUpper(operation.Method), Path: path, ContentType: defaultValue(operation.ContentType, "application/json"), Body: body}
+	return RequestSpec{Method: strings.ToUpper(operation.Method), Path: path, ContentType: defaultValue(operation.ContentType, "application/json"), Body: normalizedBody}
+}
+
+func normalizeManifestValue(value any) any {
+	switch v := value.(type) {
+	case map[string]any:
+		if len(v) == 0 {
+			return v
+		}
+		isSequential := true
+		for i := 0; i < len(v); i++ {
+			if _, ok := v[strconv.Itoa(i)]; !ok {
+				isSequential = false
+				break
+			}
+		}
+		if isSequential {
+			arr := make([]any, len(v))
+			for i := 0; i < len(v); i++ {
+				arr[i] = normalizeManifestValue(v[strconv.Itoa(i)])
+			}
+			return arr
+		}
+		res := make(map[string]any, len(v))
+		for k, val := range v {
+			res[k] = normalizeManifestValue(val)
+		}
+		return res
+	case []any:
+		arr := make([]any, len(v))
+		for i, val := range v {
+			arr[i] = normalizeManifestValue(val)
+		}
+		return arr
+	default:
+		return value
+	}
 }

 func manifestExpressionValue(expression string, request GenerationRequest, taskID string) any {
@@ -501,7 +567,10 @@ func manifestExpressionValue(expression string, request GenerationRequest, taskI
 		value = source
 	}
 	for _, transform := range transforms {
-		value = applyManifestTransform(value, transform)
+		value = applyManifestTransform(value, transform, request)
+		if value == nil {
+			break
+		}
 	}
 	return value
 }
@@ -510,9 +579,40 @@ func manifestExpressionValue(expression string, request GenerationRequest, taskI
 // platform request. It lets uploaded manifests address media items by index
 // without exposing Go structs or adding provider-specific host code.
 func manifestRequestValues(request GenerationRequest) map[string]any {
+	userContent := make([]any, 0)
+	if strings.TrimSpace(request.Prompt) != "" {
+		userContent = append(userContent, map[string]any{"type": "text", "text": request.Prompt})
+	}
+	for _, img := range request.Images {
+		val := defaultValue(img.URL, img.DataURL)
+		if val != "" {
+			userContent = append(userContent, map[string]any{"type": "image_url", "image_url": map[string]any{"url": val}})
+		}
+	}
+	for _, vid := range request.Videos {
+		val := defaultValue(vid.URL, vid.DataURL)
+		if val != "" {
+			userContent = append(userContent, map[string]any{"type": "video_url", "video_url": map[string]any{"url": val}})
+		}
+	}
+	for _, aud := range request.Audios {
+		val := defaultValue(aud.URL, aud.DataURL)
+		if val != "" {
+			userContent = append(userContent, map[string]any{"type": "audio_url", "audio_url": map[string]any{"url": val}})
+		}
+	}
+	messages := make([]any, 0)
+	if len(userContent) > 0 {
+		messages = append(messages, map[string]any{
+			"role":    "user",
+			"content": userContent,
+		})
+	}
+
 	return map[string]any{
 		"model":         request.Model,
 		"prompt":        request.Prompt,
+		"messages":      messages,
 		"images":        manifestMediaValues(request.Images),
 		"videos":        manifestMediaValues(request.Videos),
 		"audios":        manifestMediaValues(request.Audios),
@@ -528,6 +628,14 @@ func manifestRequestValues(request GenerationRequest) map[string]any {
 	}
 }

+func manifestModelID(request GenerationRequest) string {
+	modelID := strings.TrimSpace(request.Model)
+	if separator := strings.LastIndex(modelID, "::"); separator >= 0 {
+		modelID = modelID[separator+2:]
+	}
+	return strings.ToLower(modelID)
+}
+
 func manifestMediaValues(values []MediaReference) []any {
 	result := make([]any, 0, len(values))
 	for _, value := range values {
@@ -541,9 +649,90 @@ func manifestMediaValues(values []MediaReference) []any {
 	return result
 }

-func applyManifestTransform(value any, transform string) any {
+func applyManifestTransform(value any, transform string, request GenerationRequest) any {
 	transform = strings.ToLower(strings.TrimSpace(transform))
+	if transform == "bool" || transform == "boolean" {
+		switch v := value.(type) {
+		case bool:
+			return v
+		case string:
+			s := strings.ToLower(strings.TrimSpace(v))
+			return s == "true" || s == "1" || s == "yes"
+		case int, int64, float64:
+			return v != 0
+		default:
+			return false
+		}
+	}
+	if transform == "int" || transform == "integer" {
+		switch v := value.(type) {
+		case int:
+			return v
+		case int64:
+			return int(v)
+		case float64:
+			return int(v)
+		case string:
+			if n, err := strconv.Atoi(strings.TrimSpace(v)); err == nil {
+				return n
+			}
+			return 0
+		default:
+			return 0
+		}
+	}
+	if transform == "omit_zero" {
+		switch v := value.(type) {
+		case int:
+			if v == 0 {
+				return nil
+			}
+		case int64:
+			if v == 0 {
+				return nil
+			}
+		case float64:
+			if v == 0 {
+				return nil
+			}
+		case string:
+			if strings.TrimSpace(v) == "0" || strings.TrimSpace(v) == "" {
+				return nil
+			}
+		}
+	}
+	cleanModel := manifestModelID(request)
+	if strings.HasPrefix(transform, "omit_unless_model_contains:") {
+		needle := strings.ToLower(strings.TrimSpace(strings.TrimPrefix(transform, "omit_unless_model_contains:")))
+		if needle == "" || !strings.Contains(cleanModel, needle) {
+			return nil
+		}
+	}
+	if strings.HasPrefix(transform, "omit_if_model_contains:") {
+		needle := strings.ToLower(strings.TrimSpace(strings.TrimPrefix(transform, "omit_if_model_contains:")))
+		if needle != "" && strings.Contains(cleanModel, needle) {
+			return nil
+		}
+	}
+	if strings.HasPrefix(transform, "omit_unless_model_equals:") {
+		target := strings.ToLower(strings.TrimSpace(strings.TrimPrefix(transform, "omit_unless_model_equals:")))
+		if target == "" || cleanModel != target {
+			return nil
+		}
+	}
+	if strings.HasPrefix(transform, "omit_if_model_equals:") {
+		target := strings.ToLower(strings.TrimSpace(strings.TrimPrefix(transform, "omit_if_model_equals:")))
+		if target != "" && cleanModel == target {
+			return nil
+		}
+	}
 	text, ok := value.(string)
+	if transform == "omit_empty" && (!ok || strings.TrimSpace(text) == "") {
+		return nil
+	}
+	if transform == "omit_auto" && (!ok || strings.TrimSpace(text) == "" || strings.EqualFold(strings.TrimSpace(text), "auto") || strings.EqualFold(strings.TrimSpace(text), "default")) {
+		return nil
+	}
 	if !ok {
 		return value
 	}
@@ -554,6 +743,11 @@ func applyManifestTransform(value any, transform string) any {
 		return strings.ToLower(text)
 	case "upper", "uppercase":
 		return strings.ToUpper(text)
+	case "resolution_p":
+		if regexp.MustCompile(`^\d+$`).MatchString(strings.TrimSpace(text)) {
+			return strings.TrimSpace(text) + "p"
+		}
+		return text
 	case "video-resolution", "video_resolution", "videoresolution":
 		// Compatibility for already-installed 1.0.1 manifests. New plugins
 		// should declare exact enum values and use generic string transforms.
--- a/backend/internal/service/protocol_runtime_test.go
+++ b/backend/internal/service/protocol_runtime_test.go
@@ -176,7 +176,7 @@ func TestAutoDLPluginPackageInstallsThroughUploadRuntime(t *testing.T) {
 			t.Fatal("legacy AutoDL provider ID leaked into administrator catalog")
 		}
 	}
-	if autoDL == nil || len(autoDL.Workflows) != 1 {
+	if autoDL == nil || len(autoDL.Workflows) == 0 {
 		t.Fatalf("AutoDL administrator catalog = %#v", catalog)
 	}
 }
--- a/backend/internal/service/resource.go
+++ b/backend/internal/service/resource.go
@@ -94,7 +94,7 @@ func (s *Service) directResourceURL(resource *model.Resource, expiresAt time.Tim
 		return "", BadAuthRequest("资源尚未上传完成")
 	}
 	if resource.Provider == "local" {
-		return s.signedPublicResourceURL(resource.ID, expiresAt)
+		return s.signedPublicResourceURL(resource, expiresAt)
 	}
 	setting, err := s.ossSettingForResource(resource.UserID, resource)
 	if err != nil {
@@ -104,7 +104,7 @@ func (s *Service) directResourceURL(resource *model.Resource, expiresAt time.Tim
 	setting.Endpoint = firstNonEmpty(resource.Endpoint, setting.Endpoint)
 	setting.Bucket = firstNonEmpty(resource.Bucket, setting.Bucket)
 	if setting.Provider == s3Provider && !publicHTTPSStorageEndpoint(setting.Endpoint) {
-		return s.signedHTTPSPublicResourceURL(resource.ID, expiresAt)
+		return s.signedHTTPSPublicResourceURL(resource, expiresAt)
 	}
 	return signedOSSObjectURL(setting, resource.ObjectKey, expiresAt)
 }
@@ -156,7 +156,7 @@ func (s *Service) prepareResourceDelivery(userID string, resource *model.Resourc
 		}
 		if options.ForceDirect {
 			if setting.Provider == s3Provider && !publicHTTPSStorageEndpoint(setting.Endpoint) {
-				redirectURL, err := s.signedHTTPSPublicResourceURL(resource.ID, time.Now().Add(directResourceURLTTL))
+				redirectURL, err := s.signedHTTPSPublicResourceURL(resource, time.Now().Add(directResourceURLTTL))
 				if err != nil {
 					return nil, err
 				}
@@ -172,17 +172,28 @@ func (s *Service) prepareResourceDelivery(userID string, resource *model.Resourc
 	return &ResourceDelivery{Resource: resource}, nil
 }

-func (s *Service) signedPublicResourceURL(resourceID string, expiresAt time.Time) (string, error) {
+func (s *Service) signedPublicResourceURL(resource *model.Resource, expiresAt time.Time) (string, error) {
+	if resource == nil {
+		return "", errors.New("资源不存在")
+	}
 	baseURL, err := s.publicResourceBaseURL()
 	if err != nil {
 		return "", err
 	}
 	expires := strconv.FormatInt(expiresAt.UTC().Unix(), 10)
-	signature, err := s.signPublicResource(resourceID, expires)
+	signature, err := s.signPublicResource(resource.ID, expires)
 	if err != nil {
 		return "", err
 	}
-	baseURL.Path = strings.TrimRight(baseURL.Path, "/") + "/api/public/resources/" + url.PathEscape(resourceID) + "/file"
+	ext := resourceFileExtension(resource.ObjectKey, resource.MimeType, resource.Kind)
+	filename := resource.ID
+	if ext != "" {
+		if !strings.HasPrefix(ext, ".") {
+			ext = "." + ext
+		}
+		filename += ext
+	}
+	baseURL.Path = strings.TrimRight(baseURL.Path, "/") + "/api/public/resources/" + url.PathEscape(resource.ID) + "/file/" + url.PathEscape(filename)
 	query := baseURL.Query()
 	query.Set("expires", expires)
 	query.Set("signature", signature)
@@ -190,7 +201,7 @@ func (s *Service) signedPublicResourceURL(resourceID string, expiresAt time.Time
 	return baseURL.String(), nil
 }

-func (s *Service) signedHTTPSPublicResourceURL(resourceID string, expiresAt time.Time) (string, error) {
+func (s *Service) signedHTTPSPublicResourceURL(resource *model.Resource, expiresAt time.Time) (string, error) {
 	baseURL, err := s.publicResourceBaseURL()
 	if err != nil {
 		return "", err
@@ -198,7 +209,7 @@ func (s *Service) signedHTTPSPublicResourceURL(resourceID string, expiresAt time
 	if baseURL.Scheme != "https" {
 		return "", BadAuthRequest("私网或 HTTP S3 用于上游资源时，服务器公开访问地址必须使用 HTTPS")
 	}
-	return s.signedPublicResourceURL(resourceID, expiresAt)
+	return s.signedPublicResourceURL(resource, expiresAt)
 }

 func (s *Service) verifyPublicResourceSignature(resourceID string, expires string, signature string) error {
--- a/backend/internal/service/resource_test.go
+++ b/backend/internal/service/resource_test.go
@@ -718,7 +718,7 @@ func TestHydrateNewAPIChannel1ResourceUsesSignedLocalURL(t *testing.T) {
 	if err := svc.hydrateProviderMedia("user-1", &media, true); err != nil {
 		t.Fatalf("hydrateProviderMedia() error = %v", err)
 	}
-	if !strings.HasPrefix(media.URL, server.URL+"/api/public/resources/resource-local/file?") || !strings.Contains(media.URL, "signature=") || media.DataURL != "" {
+	if !strings.HasPrefix(media.URL, server.URL+"/api/public/resources/resource-local/file/resource-local.png?") || !strings.Contains(media.URL, "signature=") || media.DataURL != "" {
 		t.Fatalf("media = %#v", media)
 	}
 	stored, err := svc.repo.Resource("resource-local")
--- a/plugin-packages/autodl-comfyui.yingce-plugin
+++ b/plugin-packages/autodl-comfyui.yingce-plugin
--- a/plugin-packages/autodl-comfyui/manifest.json
+++ b/plugin-packages/autodl-comfyui/manifest.json
@@ -1,11 +1,11 @@
 {
  "apiVersion": "yingce.plugin/v1",
  "id": "autodl-comfyui",
-  "name": "AutoDL ComfyUI 视频",
-  "version": "1.0.2",
+  "name": "AutoDL ComfyUI 视频与音频",
+  "version": "1.0.3",
  "author": "影策社区",
-  "description": "通过 AutoDL.Art ComfyUI 工作流 API 接入异步视频生成。",
-  "documentation": "# AutoDL ComfyUI 视频\n\n## 接口\n\n创建任务：`POST /api/v1/comfyui/comfyui_workflow/{workflow_id}`。清单把渠道中的模型字段作为 `workflow_id`，因此每个 ComfyUI 工作流都可以通过工作流贡献加入，不需要修改宿主代码。\n\n查询任务：`GET /api/v1/comfyui/comfyui_workflow/result/{task_id}`。\n\n## 工作流\n\n示例 `minimax_h3_lightx2v_no_pic` 作为 MiniMax H3 文生视频入口，默认时长为 5 秒、分辨率为 `768p竖`。支持 `480p竖`、`768p竖`、`480p横`、`768p横`。上传后可以在渠道中直接选择该工作流；新增工作流只需扩展本清单。\n\n## 请求与鉴权\n\n创建请求为 JSON，发送 `prompt`、`duration`、`resolution`，并按存在的参考图片发送 `ref_image_0` 至 `ref_image_8`。分辨率选项由画布按工作流声明完整保留，插件在发送前仅统一转换为小写；`768p竖`、`768p横` 等值不会被宿主折叠为通用数字档位。参考图片由宿主转换为短期签名公网 URL。AutoDL Token 在渠道密钥字段中配置，宿主以后端出站请求发送 `Authorization: Bearer <token>`；清单不包含密钥。\n\n## 状态与结果\n\n`QUEUED` 和 `RUNNING` 映射为统一任务的排队与处理中，`SUCCESS` 和 `completed` 映射为成功，`FAILED` 映射为失败。错误响应的 `code` 和 `msg` 会保留为任务错误。成功结果从 `data.results[].url` 读取为视频地址。AutoDL 地址是短期地址，清单标记为临时结果，宿主会在完成时下载并保存到影策资源存储。\n\n官方文档：[AutoDL ComfyUI API](https://autodl.art/docs/comfyui_api/)。",
+  "description": "通过 AutoDL.Art ComfyUI 工作流 API 接入异步视频与音频生成，支持 H3 视频系列及 IndexTTS 语音工作流。",
+  "documentation": "# AutoDL ComfyUI 工作流\n\n## 接口\n\n- 创建任务：`POST /api/v1/comfyui/comfyui_workflow/{workflow_id}`\n- 查询结果：`GET /api/v1/comfyui/comfyui_workflow/result/{task_id}`\n\n## 鉴权\n\n在渠道密钥中配置 AutoDL ComfyUI 令牌（Authorization: <TOKEN>）。\n\n## 工作流清单（8 个工作流）\n\n1. `minimax_h3_image_audio_to_video_v2_15s` - H3 多图多音频生视频 15 秒\n2. `minimax_h3_lightx2v_v5_15s` - H3 多图生视频 15 秒\n3. `minimax_h3_image_audio_to_video_v2` - H3 多图多音频生视频\n4. `minimax_h3_image_audio_to_video` - H3 图生视频音频同步\n5. `minimax_h3_lightx2v_v5` - H3 多图参考生视频\n6. `minimax_h3_lightx2v_no_pic` - H3 文生视频\n7. `minimax_h3_lightx2v` - H3 首尾帧生成视频\n8. `indextts2-v1` - IndexTTS2 语音\n",
  "permissions": [
    "generation.run",
    "media.read"
@@ -17,7 +17,7 @@
        "type": "secret",
        "label": "AutoDL Token",
        "required": true,
-        "description": "填写 AutoDL Token；仅保存到渠道密钥配置。"
+        "description": "填写 AutoDL 大模型令牌管理中创建的 ComfyUI 分组 Token。"
      }
    ]
  },
@@ -38,7 +38,7 @@
        "baseUrl": "https://autodl.art",
        "requiresPublicMediaUrls": true,
        "auth": {
-          "type": "bearer",
+          "type": "header",
          "field": "apiKey",
          "header": "Authorization"
        },
@@ -54,13 +54,13 @@
            "name": "duration",
            "type": "integer",
            "mapping": "duration",
-            "description": "视频时长，单位由工作流决定。"
+            "description": "视频时长（秒）。"
          },
          {
            "name": "resolution",
            "type": "string",
            "mapping": "resolution",
-            "description": "工作流支持的分辨率档位。"
+            "description": "输出分辨率档位，如 480p竖、768p竖、1080p竖、480p横、768p横、1080p横、480p(1:1)、768p(1:1) 等。"
          }
        ],
        "create": {
@@ -68,18 +68,24 @@
          "path": "/api/v1/comfyui/comfyui_workflow/{{model}}",
          "contentType": "application/json",
          "fields": {
-            "prompt": "request.prompt",
-            "duration": "request.duration",
+            "prompt": "request.prompt|omit_if_model_equals:minimax_h3_image_audio_to_video",
+            "duration": "request.duration|omit_if_model_equals:minimax_h3_image_audio_to_video",
+            "audio_duration": "request.duration|omit_unless_model_equals:minimax_h3_image_audio_to_video",
            "resolution": "request.resolution|lower",
-            "ref_image_0": "request.images.0.url",
-            "ref_image_1": "request.images.1.url",
-            "ref_image_2": "request.images.2.url",
-            "ref_image_3": "request.images.3.url",
-            "ref_image_4": "request.images.4.url",
-            "ref_image_5": "request.images.5.url",
-            "ref_image_6": "request.images.6.url",
-            "ref_image_7": "request.images.7.url",
-            "ref_image_8": "request.images.8.url"
+            "ref_image_0": "request.images.0.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_1": "request.images.1.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_2": "request.images.2.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_3": "request.images.3.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_4": "request.images.4.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_5": "request.images.5.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_6": "request.images.6.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_7": "request.images.7.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "ref_image_8": "request.images.8.url|omit_if_model_equals:minimax_h3_lightx2v|omit_if_model_equals:minimax_h3_b99_002|omit_if_model_equals:minimax_h3_image_audio_to_video|omit_if_model_contains:no_pic|omit_if_model_contains:b99_001",
+            "first_frame": "request.images.0.url|omit_if_model_contains:v5|omit_if_model_contains:no_pic|omit_if_model_contains:image_audio_to_video|omit_if_model_contains:b99_001|omit_if_model_contains:b99_003",
+            "last_frame": "request.images.1.url|omit_if_model_contains:v5|omit_if_model_contains:no_pic|omit_if_model_contains:image_audio_to_video|omit_if_model_contains:b99_001|omit_if_model_contains:b99_003",
+            "ref_audio_0": "request.audios.0.url|omit_unless_model_contains:image_audio_to_video",
+            "ref_audio_1": "request.audios.1.url|omit_unless_model_contains:image_audio_to_video_v2",
+            "ref_audio_2": "request.audios.2.url|omit_unless_model_contains:image_audio_to_video_v2"
          }
        },
        "poll": {
@@ -106,44 +112,200 @@
          "resultKind": "video",
          "resultEphemeral": true
        }
+      },
+      {
+        "id": "autodl-comfyui-audio",
+        "label": "AutoDL ComfyUI 语音",
+        "capabilities": [
+          "audio"
+        ],
+        "scopes": [
+          "admin.system-channel",
+          "user.custom-channel",
+          "canvas",
+          "creation"
+        ],
+        "baseUrl": "https://autodl.art",
+        "requiresPublicMediaUrls": true,
+        "auth": {
+          "type": "header",
+          "field": "apiKey",
+          "header": "Authorization"
+        },
+        "parameters": [
+          {
+            "name": "prompt_text",
+            "type": "string",
+            "required": true,
+            "mapping": "prompt_text",
+            "description": "语音合成文本。"
+          }
+        ],
+        "create": {
+          "method": "POST",
+          "path": "/api/v1/comfyui/comfyui_workflow/{{model}}",
+          "contentType": "application/json",
+          "fields": {
+            "prompt_text": "request.prompt",
+            "prompt_simple": "request.audios.0.url",
+            "emo_control_method": "与音色参考音频相同"
+          }
+        },
+        "poll": {
+          "method": "GET",
+          "path": "/api/v1/comfyui/comfyui_workflow/result/{{taskId}}"
+        },
+        "response": {
+          "taskIdPaths": [
+            "data.task_id"
+          ],
+          "statusPaths": [
+            "data.status"
+          ],
+          "errorPaths": [
+            "code"
+          ],
+          "messagePaths": [
+            "data.message",
+            "msg"
+          ],
+          "resultPaths": [
+            "data.results"
+          ],
+          "resultKind": "audio",
+          "resultEphemeral": true
+        }
      }
    ],
    "workflows": [
      {
-        "id": "minimax_h3_lightx2v_no_pic",
-        "label": "MiniMax H3 文生视频",
+        "id": "minimax_h3_b99_002",
+        "label": "H3 首尾帧生成视频",
        "providerId": "autodl-comfyui",
        "capability": "video",
        "parameters": [
-          {
-            "name": "prompt",
-            "type": "string",
-            "required": true,
-            "mapping": "prompt"
-          },
-          {
-            "name": "duration",
-            "type": "integer",
-            "mapping": "duration",
-            "description": "视频时长。"
-          },
-          {
-            "name": "resolution",
-            "type": "string",
-            "mapping": "resolution",
-            "values": [
-              "480p竖",
-              "768p竖",
-              "480p横",
-              "768p横"
-            ],
-            "description": "输出分辨率档位。"
-          }
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["736p竖", "736p横", "736p(1:1)"] }
        ],
-        "defaults": {
-          "duration": 5,
-          "resolution": "768p竖"
-        }
+        "defaults": { "duration": 5, "resolution": "736p竖" }
+      },
+      {
+        "id": "minimax_h3_b99_001",
+        "label": "H3 文生视频",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["736p竖", "736p横", "736p(1:1)"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "736p竖" }
+      },
+      {
+        "id": "minimax_h3_b99_003_12s",
+        "label": "H3 多图生视频 12 秒",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~12 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["736p竖", "736p横", "736p(1:1)"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "736p竖" }
+      },
+      {
+        "id": "minimax_h3_image_audio_to_video_v2_15s",
+        "label": "H3 多图多音频生视频 15 秒",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "480p横", "768p横"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_lightx2v_v5_15s",
+        "label": "H3 多图生视频 15 秒",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "480p横", "768p横", "480p(1:1)", "768p(1:1)"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_image_audio_to_video_v2",
+        "label": "H3 多图多音频生视频",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~10 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "1080p竖", "480p横", "768p横", "1080p横"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_image_audio_to_video",
+        "label": "H3 图生视频音频同步",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "audio_duration", "type": "integer", "mapping": "duration", "description": "音频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "1080p竖", "480p横", "768p横", "1080p横"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_lightx2v_v5",
+        "label": "H3 多图参考生视频",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~10 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "1080p竖", "480p横", "768p横", "1080p横", "480p(1:1)", "768p(1:1)", "1080p(1:1)"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_lightx2v_no_pic",
+        "label": "H3 文生视频",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "480p横", "768p横"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "minimax_h3_lightx2v",
+        "label": "H3 首尾帧生成视频",
+        "providerId": "autodl-comfyui",
+        "capability": "video",
+        "parameters": [
+          { "name": "prompt", "type": "string", "required": true, "mapping": "prompt" },
+          { "name": "duration", "type": "integer", "mapping": "duration", "description": "视频时长 1~15 秒" },
+          { "name": "resolution", "type": "string", "mapping": "resolution", "values": ["480p竖", "768p竖", "480p横", "768p横"] }
+        ],
+        "defaults": { "duration": 5, "resolution": "768p竖" }
+      },
+      {
+        "id": "indextts2-v1",
+        "label": "IndexTTS2 语音",
+        "providerId": "autodl-comfyui-audio",
+        "capability": "audio",
+        "parameters": [
+          { "name": "prompt_text", "type": "string", "required": true, "mapping": "prompt" }
+        ],
+        "defaults": {}
      }
    ]
  }