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>

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"}`))

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)
 		}
 	}

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.