Merge pull request #21: 统一公开 API 为 /api/v1 前缀

合并公开 API V1 前缀统一、兼容路由、OpenAPI、部署健康检查与验证测试。

README.md

@@ -70,8 +70,10 @@ scripts/deploy-compose.sh
 部署成功后默认访问地址：
 
 - Web: `http://127.0.0.1:5178`
-- API: `http://127.0.0.1:8088/healthz`
-- Web 反代 API: `http://127.0.0.1:5178/gateway-api/healthz`
+- API: `http://127.0.0.1:8088/api/v1/healthz`
+- Web 反代公开 API: `http://127.0.0.1:5178/api/v1/healthz`
+
+公开接口统一使用 `/api/v1` 前缀，完整分组清单见 [公开 API V1 清单](docs/public-api-v1.md)。
 
 常用覆盖项：
 
@@ -99,7 +101,7 @@ scripts/deploy-compose.sh clean
 docker login --username=<your-aliyun-account> registry.cn-shanghai.aliyuncs.com
 ```
 
-Web 容器的 Nginx 配置通过 bind mount 挂载自仓库文件 [docker/nginx.conf](docker/nginx.conf)，可直接修改该文件调整静态资源和 `/gateway-api` 反向代理配置。修改后执行以下命令使配置生效：
+Web 容器的 Nginx 配置通过 bind mount 挂载自仓库文件 [docker/nginx.conf](docker/nginx.conf)，可直接修改该文件调整静态资源、规范 `/api/v1` 公开入口和旧 `/gateway-api` 兼容反向代理。修改后执行以下命令使配置生效：
 
 ```bash
 docker compose -f docker-compose.yml restart web

apps/api/go.mod

@@ -36,7 +36,7 @@ require (
 	github.com/tidwall/pretty v1.2.1 // indirect
 	github.com/tidwall/sjson v1.2.5 // indirect
 	go.uber.org/goleak v1.3.0 // indirect
-	golang.org/x/sync v0.20.0 // indirect
-	golang.org/x/text v0.37.0 // indirect
+	golang.org/x/sync v0.21.0 // indirect
+	golang.org/x/text v0.39.0 // indirect
 	gopkg.in/yaml.v3 v3.0.1 // indirect
 )

apps/api/go.sum

@@ -73,10 +73,10 @@ golang.org/x/crypto v0.52.0 h1:RMs7fP2rXdep0CftQlK8Uf+kibLm7qkCcradZWYz988=
 golang.org/x/crypto v0.52.0/go.mod h1:1QgfPxDqh0T2M/elOJtp9RvuR95kVjir0e6/BvEmGbc=
 golang.org/x/oauth2 v0.36.0 h1:peZ/1z27fi9hUOFCAZaHyrpWG5lwe0RJEEEeH0ThlIs=
 golang.org/x/oauth2 v0.36.0/go.mod h1:YDBUJMTkDnJS+A4BP4eZBjCqtokkg1hODuPjwiGPO7Q=
-golang.org/x/sync v0.20.0 h1:e0PTpb7pjO8GAtTs2dQ6jYa5BWYlMuX047Dco/pItO4=
-golang.org/x/sync v0.20.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
-golang.org/x/text v0.37.0 h1:Cqjiwd9eSg8e0QAkyCaQTNHFIIzWtidPahFWR83rTrc=
-golang.org/x/text v0.37.0/go.mod h1:a5sjxXGs9hsn/AJVwuElvCAo9v8QYLzvavO5z2PiM38=
+golang.org/x/sync v0.21.0 h1:HLII4xRRTtCRkxYp4HNFF0Js/Og6q2i++KXbg0gHCwM=
+golang.org/x/sync v0.21.0/go.mod h1:9xrNwdLfx4jkKbNva9FpL6vEN7evnE43NNNJQ2LF3+0=
+golang.org/x/text v0.39.0 h1:UbZz4pLOvn600D6Oh6GGEI6VAmndrEBLv8/6BEXzyus=
+golang.org/x/text v0.39.0/go.mod h1:3UwRclnC2g0TU9x8PZiyfOajCd1zaUNHF9cvqcQZ+ZM=
 gopkg.in/check.v1 v0.0.0-20161208181325-20d25e280405/go.mod h1:Co6ibVJAznAaIkqp8huTwlJQCZ016jof/cbN4VW5Yz0=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c h1:Hei/4ADfdWqJk1ZMxUNpqntNwaWcugrBjAiHlqqRiVk=
 gopkg.in/check.v1 v1.0.0-20201130134442-10cb98267c6c/go.mod h1:JHkPIbrfpd72SG/EVd6muEfDQjcINNoR0C8j2r3qZ4Q=

apps/api/internal/clients/clients_test.go

@@ -1306,6 +1306,25 @@ func TestGeminiClientImageGenerateBuildsNativeImageBody(t *testing.T) {
 	}
 }
 
+func TestGeminiGenerationConfigInitializesMissingImageConfig(t *testing.T) {
+	config := geminiGenerationConfig(map[string]any{
+		"aspect_ratio": "16:9",
+		"resolution":   "4K",
+	}, true)
+
+	imageConfig, ok := config["imageConfig"].(map[string]any)
+	if !ok {
+		t.Fatalf("imageConfig should be initialized, got %+v", config)
+	}
+	if imageConfig["aspectRatio"] != "16:9" || imageConfig["imageSize"] != "4K" {
+		t.Fatalf("unexpected imageConfig: %+v", imageConfig)
+	}
+	modalities, ok := config["responseModalities"].([]any)
+	if !ok || len(modalities) != 1 || modalities[0] != "IMAGE" {
+		t.Fatalf("image response modality should be initialized, got %+v", config)
+	}
+}
+
 func TestGeminiClientImageEditPreservesNativeContentsAndFileData(t *testing.T) {
 	var captured map[string]any
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
@@ -1539,10 +1558,39 @@ func TestGeminiClientChatConvertsFunctionCallResponse(t *testing.T) {
 }
 
 func TestGeminiURLAcceptsVersionedBaseURL(t *testing.T) {
-	got := geminiURL("https://generativelanguage.googleapis.com/v1beta", "gemini-2.5-flash", "test-key")
-	want := "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=test-key"
-	if got != want {
-		t.Fatalf("unexpected gemini url: %s", got)
+	tests := []struct {
+		name    string
+		baseURL string
+		want    string
+	}{
+		{
+			name:    "default version",
+			baseURL: "https://generativelanguage.googleapis.com",
+			want:    "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=test-key",
+		},
+		{
+			name:    "google beta version",
+			baseURL: "https://generativelanguage.googleapis.com/v1beta",
+			want:    "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=test-key",
+		},
+		{
+			name:    "compatible v1 endpoint",
+			baseURL: "https://cloud.dataeyes.ai/v1",
+			want:    "https://cloud.dataeyes.ai/v1/models/gemini-2.5-flash:generateContent?key=test-key",
+		},
+		{
+			name:    "openai suffix after version",
+			baseURL: "https://generativelanguage.googleapis.com/v1beta/openai",
+			want:    "https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key=test-key",
+		},
+	}
+	for _, test := range tests {
+		t.Run(test.name, func(t *testing.T) {
+			got := geminiURL(test.baseURL, "gemini-2.5-flash", "test-key")
+			if got != test.want {
+				t.Fatalf("unexpected gemini url: %s", got)
+			}
+		})
 	}
 }
 

apps/api/internal/clients/gemini.go

@@ -61,11 +61,13 @@ func geminiURL(baseURL string, model string, apiKey string) string {
 		base = "https://generativelanguage.googleapis.com"
 	}
 	base = strings.TrimSuffix(base, "/openai")
-	if strings.HasSuffix(base, "/v1beta") {
-		base = strings.TrimSuffix(base, "/v1beta")
+	if !strings.HasSuffix(base, "/v1") &&
+		!strings.HasSuffix(base, "/v1beta") &&
+		!strings.HasSuffix(base, "/v1alpha") {
+		base += "/v1beta"
 	}
 	escapedModel := url.PathEscape(model)
-	return fmt.Sprintf("%s/v1beta/models/%s:generateContent?key=%s", base, escapedModel, url.QueryEscape(apiKey))
+	return fmt.Sprintf("%s/models/%s:generateContent?key=%s", base, escapedModel, url.QueryEscape(apiKey))
 }
 
 func geminiBody(request Request) map[string]any {
@@ -195,14 +197,23 @@ func geminiApplyRequestOptions(body map[string]any, request Request, imageRespon
 func geminiGenerationConfig(body map[string]any, imageResponse bool) map[string]any {
 	source := mapFromAny(firstPresent(body["generationConfig"], body["generation_config"]))
 	out := cloneMapAny(source)
+	if out == nil {
+		out = map[string]any{}
+	}
 	if aspectRatio := firstNonEmptyString(body["aspect_ratio"], body["aspectRatio"]); aspectRatio != "" {
 		imageConfig := cloneMapAny(mapFromAny(firstPresent(out["imageConfig"], out["image_config"])))
+		if imageConfig == nil {
+			imageConfig = map[string]any{}
+		}
 		imageConfig["aspectRatio"] = aspectRatio
 		out["imageConfig"] = imageConfig
 		delete(out, "image_config")
 	}
 	if imageSize := firstNonEmptyString(body["resolution"], body["imageSize"], body["image_size"], body["size"]); imageSize != "" {
 		imageConfig := cloneMapAny(mapFromAny(firstPresent(out["imageConfig"], out["image_config"])))
+		if imageConfig == nil {
+			imageConfig = map[string]any{}
+		}
 		imageConfig["imageSize"] = imageSize
 		out["imageConfig"] = imageConfig
 		delete(out, "image_config")

apps/api/internal/clients/universal.go

@@ -44,6 +44,8 @@ func (c UniversalClient) Run(ctx context.Context, request Request) (Response, er
 		if err != nil {
 			return Response{}, annotateResponseError(err, submitRequestID, startedAt, time.Now())
 		}
+		submitResult = universalEffectiveResult(submitResult)
+		submitRequestID = firstNonEmptyString(submitRequestID, requestIDFromResult(submitResult))
 		if isUniversalSuccess(submitResult) && submitResult["data"] != nil {
 			return Response{
 				Result:             normalizeUniversalResult(request, submitResult, ""),
@@ -157,6 +159,7 @@ func (c UniversalClient) universalPollUntilDone(ctx context.Context, executor *s
 		if err != nil {
 			return nil, "", annotateResponseError(err, firstNonEmptyString(pollRequestID, requestID, upstreamTaskID), pollStarted, pollFinished)
 		}
+		result = universalEffectiveResult(result)
 		lastResult = result
 		requestID = firstNonEmptyString(pollRequestID, requestID, requestIDFromResult(result), upstreamTaskID)
 		if isUniversalSuccess(result) {
@@ -239,6 +242,14 @@ func universalScriptContext(request Request, modelType string, payload map[strin
 		return selectedBase + "/" + strings.TrimLeft(path, "/")
 	}
 	context["creatRequestURL"] = context["createRequestURL"]
+	processedParams := cloneBody(request.Body)
+	processedParams["model"] = upstreamModelName(request.Candidate)
+	context["processedParams"] = processedParams
+	context["preProcessParams"] = func(params map[string]any, _ ...string) map[string]any {
+		processed := cloneMapAny(params)
+		processed["model"] = upstreamModelName(request.Candidate)
+		return processed
+	}
 	context["resolveGetTaskURL"] = func(taskID string) string {
 		return resolveUniversalTaskURL(request.Candidate.PlatformConfig, taskID)
 	}
@@ -389,6 +400,20 @@ func universalStatus(result map[string]any) string {
 	return strings.ToLower(strings.TrimSpace(firstNonEmptyString(result["status"], result["state"], result["task_status"])))
 }
 
+func universalEffectiveResult(result map[string]any) map[string]any {
+	nested, ok := result["result"].(map[string]any)
+	if !ok || nested == nil || isUniversalFailure(result) {
+		return result
+	}
+	out := cloneMapAny(nested)
+	for _, key := range []string{"status", "request_id", "requestId", "upstream_task_id", "task_id", "taskId", "id"} {
+		if out[key] == nil && result[key] != nil {
+			out[key] = result[key]
+		}
+	}
+	return out
+}
+
 func universalTaskID(result map[string]any) string {
 	return firstNonEmptyString(result["upstream_task_id"], result["task_id"], result["taskId"], result["id"])
 }

apps/api/internal/clients/universal_test.go

@@ -92,6 +92,47 @@ func TestUniversalClientDefaultSubmitAndPoll(t *testing.T) {
 	}
 }
 
+func TestUniversalClientSupportsServerMainScriptContextAndNestedResult(t *testing.T) {
+	request := Request{
+		Kind:      "images.generations",
+		ModelType: "image_generate",
+		Model:     "custom-image",
+		Body:      map[string]any{"model": "custom-image", "prompt": "hello"},
+		Candidate: testUniversalCandidate(map[string]any{
+			"customGetParamsScript": map[string]any{
+				"image_generate": `async function getParams(params, context) {
+					const processed = await context.preProcessParams(params, context.type);
+					return { prompt: processed.prompt + "-" + processed.model + "-" + context.processedParams.model };
+				}`,
+			},
+			"customSubmitScript": map[string]any{
+				"image_generate": `async function submitTask(payload) {
+					return {
+						status: "success",
+						result: {
+							status: "success",
+							data: [{ url: "https://cdn.example/" + payload.prompt + ".png" }]
+						}
+					};
+				}`,
+			},
+		}),
+	}
+
+	response, err := (UniversalClient{}).Run(context.Background(), request)
+	if err != nil {
+		t.Fatalf("run failed: %v", err)
+	}
+	data, ok := response.Result["data"].([]any)
+	if !ok || len(data) != 1 {
+		t.Fatalf("unexpected nested result: %#v", response.Result)
+	}
+	image, ok := data[0].(map[string]any)
+	if !ok || image["url"] != "https://cdn.example/hello-provider-model-provider-model.png" {
+		t.Fatalf("unexpected image result: %#v", response.Result)
+	}
+}
+
 func TestUniversalClientResumeSkipsSubmit(t *testing.T) {
 	request := Request{
 		Kind:              "videos.generations",

apps/api/internal/httpapi/agent_resources_handlers.go

@@ -9,8 +9,8 @@ import (
 
 const (
 	opsManagementSkillDownloadPath = "/api/v1/public/skills/ai-gateway-ops-management/download"
-	apiDocsJSONPath                = "/api-docs-json"
-	apiDocsYAMLPath                = "/api-docs-yaml"
+	apiDocsJSONPath                = "/api/v1/openapi.json"
+	apiDocsYAMLPath                = "/api/v1/openapi.yaml"
 )
 
 // getOpsManagementSkillMetadata godoc
@@ -64,7 +64,7 @@ func (s *Server) downloadOpsManagementSkill(w http.ResponseWriter, _ *http.Reque
 // @Tags agent-resources
 // @Produce json
 // @Success 200 {object} map[string]interface{}
-// @Router /api-docs-json [get]
+// @Router /api/v1/openapi.json [get]
 func (s *Server) apiDocsJSON(w http.ResponseWriter, _ *http.Request) {
 	w.Header().Set("Content-Type", "application/json; charset=utf-8")
 	w.WriteHeader(http.StatusOK)
@@ -77,7 +77,7 @@ func (s *Server) apiDocsJSON(w http.ResponseWriter, _ *http.Request) {
 // @Tags agent-resources
 // @Produce application/yaml
 // @Success 200 {string} string
-// @Router /api-docs-yaml [get]
+// @Router /api/v1/openapi.yaml [get]
 func (s *Server) apiDocsYAML(w http.ResponseWriter, _ *http.Request) {
 	w.Header().Set("Content-Type", "application/yaml; charset=utf-8")
 	w.WriteHeader(http.StatusOK)

apps/api/internal/httpapi/agent_resources_handlers_test.go

@@ -32,7 +32,7 @@ func TestGetOpsManagementSkillMetadata(t *testing.T) {
 	if len(metadata.Modules) != 1 || metadata.Modules[0] != "model-runtime" {
 		t.Fatalf("unexpected metadata modules: %+v", metadata.Modules)
 	}
-	if metadata.APIDocsJSONPath != "/api-docs-json" || metadata.APIDocsYAMLPath != "/api-docs-yaml" {
+	if metadata.APIDocsJSONPath != "/api/v1/openapi.json" || metadata.APIDocsYAMLPath != "/api/v1/openapi.yaml" {
 		t.Fatalf("unexpected API docs paths: %+v", metadata)
 	}
 }
@@ -74,7 +74,7 @@ func TestEmbeddedAPIDocs(t *testing.T) {
 	server := &Server{}
 
 	jsonResponse := httptest.NewRecorder()
-	server.apiDocsJSON(jsonResponse, httptest.NewRequest(http.MethodGet, "/api-docs-json", nil))
+	server.apiDocsJSON(jsonResponse, httptest.NewRequest(http.MethodGet, "/api/v1/openapi.json", nil))
 	if jsonResponse.Code != http.StatusOK {
 		t.Fatalf("expected JSON docs status 200, got %d", jsonResponse.Code)
 	}
@@ -88,7 +88,7 @@ func TestEmbeddedAPIDocs(t *testing.T) {
 		t.Fatalf("decode embedded Swagger JSON: %v", err)
 	}
 	for _, path := range []string{
-		"/api-docs-json",
+		"/api/v1/openapi.json",
 		"/api/v1/public/skills/ai-gateway-ops-management/download",
 		"/api/admin/catalog/providers",
 		"/api/admin/catalog/base-models",
@@ -102,7 +102,7 @@ func TestEmbeddedAPIDocs(t *testing.T) {
 	}
 
 	yamlResponse := httptest.NewRecorder()
-	server.apiDocsYAML(yamlResponse, httptest.NewRequest(http.MethodGet, "/api-docs-yaml", nil))
+	server.apiDocsYAML(yamlResponse, httptest.NewRequest(http.MethodGet, "/api/v1/openapi.yaml", nil))
 	if yamlResponse.Code != http.StatusOK {
 		t.Fatalf("expected YAML docs status 200, got %d", yamlResponse.Code)
 	}

apps/api/internal/httpapi/chat_completions_mode_test.go

@@ -58,17 +58,18 @@ func TestPlanTaskResponseTreatsAPIV1EmbeddingAndRerankAsSynchronousCompatibleRes
 	}
 }
 
-func TestPlanTaskResponseKeepsAsyncTaskModeForOtherAPIV1Tasks(t *testing.T) {
-	req := httptest.NewRequest(http.MethodPost, "/api/v1/images/generations", nil)
-	req.Header.Set("X-Async", "true")
-
-	plan := planTaskResponse("images.generations", false, map[string]any{"stream": true}, req)
-
-	if !plan.asyncMode {
-		t.Fatal("non-chat /api/v1 task endpoints should keep X-Async task mode")
+func TestPlanTaskResponseUsesCompatibleAPIV1MediaResponsesAndKeepsAsyncOptIn(t *testing.T) {
+	defaultRequest := httptest.NewRequest(http.MethodPost, "/api/v1/images/generations", nil)
+	defaultPlan := planTaskResponse("images.generations", true, map[string]any{}, defaultRequest)
+	if defaultPlan.asyncMode || !defaultPlan.compatibleMode {
+		t.Fatalf("canonical /api/v1 media endpoints should default to synchronous compatible responses, got %+v", defaultPlan)
 	}
-	if plan.compatibleMode {
-		t.Fatal("non-compatible /api/v1 task endpoints should not return OpenAI-compatible payloads")
+
+	asyncRequest := httptest.NewRequest(http.MethodPost, "/api/v1/images/generations", nil)
+	asyncRequest.Header.Set("X-Async", "true")
+	asyncPlan := planTaskResponse("images.generations", true, map[string]any{}, asyncRequest)
+	if !asyncPlan.asyncMode || !asyncPlan.compatibleMode {
+		t.Fatalf("canonical /api/v1 media endpoints should keep compatible mode when X-Async is enabled, got %+v", asyncPlan)
 	}
 }
 

apps/api/internal/httpapi/core_flow_integration_test.go

@@ -504,7 +504,7 @@ VALUES ($1, 5, '{"purpose":"core-flow"}'::jsonb)`, inviteCode); err != nil {
 			Result map[string]any `json:"result"`
 		} `json:"task"`
 	}
-	doJSON(t, server.URL, http.MethodPost, "/api/v1/images/generations", apiKeyResponse.Secret, map[string]any{
+	doJSONWithHeaders(t, server.URL, http.MethodPost, "/api/v1/images/generations", apiKeyResponse.Secret, map[string]any{
 		"model":                defaultImageModel,
 		"runMode":              "simulation",
 		"prompt":               "a tiny gateway console",
@@ -512,7 +512,9 @@ VALUES ($1, 5, '{"purpose":"core-flow"}'::jsonb)`, inviteCode); err != nil {
 		"quality":              "medium",
 		"simulation":           true,
 		"simulationDurationMs": 5,
-	}, http.StatusAccepted, &imageResponse)
+	}, map[string]string{"X-Async": "true"}, http.StatusAccepted, &imageResponse)
+	waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, imageResponse.Task.ID, []string{"succeeded"}, 10*time.Second)
+	doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+imageResponse.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &imageResponse.Task)
 	if imageResponse.Task.Status != "succeeded" || imageResponse.Task.Result["id"] == "" {
 		t.Fatalf("unexpected image generation task: %+v", imageResponse.Task)
 	}
@@ -524,7 +526,7 @@ VALUES ($1, 5, '{"purpose":"core-flow"}'::jsonb)`, inviteCode); err != nil {
 			Result map[string]any `json:"result"`
 		} `json:"task"`
 	}
-	doJSON(t, server.URL, http.MethodPost, "/api/v1/images/edits", apiKeyResponse.Secret, map[string]any{
+	doJSONWithHeaders(t, server.URL, http.MethodPost, "/api/v1/images/edits", apiKeyResponse.Secret, map[string]any{
 		"model":                defaultImageModel,
 		"runMode":              "simulation",
 		"prompt":               "replace background with clean studio light",
@@ -532,7 +534,9 @@ VALUES ($1, 5, '{"purpose":"core-flow"}'::jsonb)`, inviteCode); err != nil {
 		"mask":                 "https://example.com/mask.png",
 		"simulation":           true,
 		"simulationDurationMs": 5,
-	}, http.StatusAccepted, &imageEditResponse)
+	}, map[string]string{"X-Async": "true"}, http.StatusAccepted, &imageEditResponse)
+	waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, imageEditResponse.Task.ID, []string{"succeeded"}, 10*time.Second)
+	doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+imageEditResponse.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &imageEditResponse.Task)
 	if imageEditResponse.Task.Status != "succeeded" || imageEditResponse.Task.Result["id"] == "" {
 		t.Fatalf("unexpected image edit task: %+v", imageEditResponse.Task)
 	}
@@ -1196,17 +1200,20 @@ WHERE reference_type = 'gateway_task'
 	}, http.StatusCreated, &videoRoutePlatformModel)
 	var textToVideoTask struct {
 		Task struct {
+			ID        string `json:"id"`
 			Status    string `json:"status"`
 			ModelType string `json:"modelType"`
 		} `json:"task"`
 	}
-	doJSON(t, server.URL, http.MethodPost, "/api/v1/videos/generations", apiKeyResponse.Secret, map[string]any{
+	doJSONWithHeaders(t, server.URL, http.MethodPost, "/api/v1/videos/generations", apiKeyResponse.Secret, map[string]any{
 		"model":                videoRouteModel,
 		"runMode":              "simulation",
 		"simulation":           true,
 		"simulationDurationMs": 5,
 		"prompt":               "text to video route",
-	}, http.StatusAccepted, &textToVideoTask)
+	}, map[string]string{"X-Async": "true"}, http.StatusAccepted, &textToVideoTask)
+	waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, textToVideoTask.Task.ID, []string{"succeeded"}, 10*time.Second)
+	doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+textToVideoTask.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &textToVideoTask.Task)
 	if textToVideoTask.Task.Status != "succeeded" || textToVideoTask.Task.ModelType != "video_generate" {
 		t.Fatalf("text-to-video request should use video_generate model_type: %+v", textToVideoTask.Task)
 	}
@@ -1218,14 +1225,16 @@ WHERE reference_type = 'gateway_task'
 			Metrics   map[string]any `json:"metrics"`
 		} `json:"task"`
 	}
-	doJSON(t, server.URL, http.MethodPost, "/api/v1/videos/generations", apiKeyResponse.Secret, map[string]any{
+	doJSONWithHeaders(t, server.URL, http.MethodPost, "/api/v1/videos/generations", apiKeyResponse.Secret, map[string]any{
 		"model":                videoRouteModel,
 		"runMode":              "simulation",
 		"simulation":           true,
 		"simulationDurationMs": 5,
 		"prompt":               "image to video route",
 		"image":                "https://example.com/source.png",
-	}, http.StatusAccepted, &imageToVideoTask)
+	}, map[string]string{"X-Async": "true"}, http.StatusAccepted, &imageToVideoTask)
+	waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, imageToVideoTask.Task.ID, []string{"succeeded"}, 10*time.Second)
+	doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+imageToVideoTask.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &imageToVideoTask.Task)
 	if imageToVideoTask.Task.Status != "succeeded" || imageToVideoTask.Task.ModelType != "image_to_video" {
 		t.Fatalf("image-to-video request should use image_to_video model_type: %+v", imageToVideoTask.Task)
 	}

apps/api/internal/httpapi/file_upload_handlers.go

@@ -26,7 +26,6 @@ const maxGatewayUploadBytes = 256 << 20
 // @Failure 502 {object} ErrorEnvelope
 // @Failure 503 {object} ErrorEnvelope
 // @Router /api/v1/files/upload [post]
-// @Router /v1/files/upload [post]
 func (s *Server) uploadFile(w http.ResponseWriter, r *http.Request) {
 	r.Body = http.MaxBytesReader(w, r.Body, maxGatewayUploadBytes)
 	if err := r.ParseMultipartForm(32 << 20); err != nil {

apps/api/internal/httpapi/gemini_compat.go

@@ -75,6 +75,21 @@ func geminiGenerateContentModelFromPath(prefix string, requestPath string) (stri
 	return model, true
 }
 
+// geminiGenerateContent godoc
+// @Summary Gemini generateContent 兼容接口
+// @Description 使用统一 /api/v1 前缀接收 Gemini generateContent 请求；旧 /v1 和 /v1beta 路径保留兼容。
+// @Tags gemini-compatible
+// @Accept json
+// @Produce json
+// @Security BearerAuth
+// @Param model path string true "模型名称"
+// @Param input body map[string]interface{} true "Gemini generateContent 请求"
+// @Success 200 {object} map[string]interface{}
+// @Failure 400 {object} ErrorEnvelope
+// @Failure 401 {object} ErrorEnvelope
+// @Failure 403 {object} ErrorEnvelope
+// @Failure 502 {object} ErrorEnvelope
+// @Router /api/v1/models/{model}:generateContent [post]
 func (s *Server) geminiGenerateContent(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok {
@@ -395,6 +410,18 @@ func geminiUsageMetadataFromOutput(output map[string]any) map[string]any {
 	return meta
 }
 
+// geminiFilesUpload godoc
+// @Summary Gemini Files 上传接口
+// @Description 使用统一 /api/v1 前缀启动或直接完成 Gemini Files 上传。
+// @Tags gemini-compatible
+// @Accept octet-stream
+// @Produce json
+// @Security BearerAuth
+// @Param version path string true "Gemini 版本（v1 或 v1beta）"
+// @Success 200 {object} map[string]interface{}
+// @Failure 400 {object} ErrorEnvelope
+// @Failure 401 {object} ErrorEnvelope
+// @Router /api/v1/gemini/upload/{version}/files [post]
 func (s *Server) geminiFilesUpload(w http.ResponseWriter, r *http.Request) {
 	if err := validateGeminiFilesVersion(r.PathValue("version")); err != nil {
 		writeError(w, http.StatusBadRequest, err.Error(), clients.ErrorCode(err))
@@ -412,6 +439,18 @@ func (s *Server) geminiFilesUpload(w http.ResponseWriter, r *http.Request) {
 	})
 }
 
+// geminiFilesUploadFinalize godoc
+// @Summary 完成 Gemini Files 分段上传
+// @Tags gemini-compatible
+// @Accept octet-stream
+// @Produce json
+// @Security BearerAuth
+// @Param version path string true "Gemini 版本（v1 或 v1beta）"
+// @Param uploadID path string true "上传会话 ID"
+// @Success 200 {object} map[string]interface{}
+// @Failure 400 {object} ErrorEnvelope
+// @Failure 401 {object} ErrorEnvelope
+// @Router /api/v1/gemini/upload/{version}/files/{uploadID} [post]
 func (s *Server) geminiFilesUploadFinalize(w http.ResponseWriter, r *http.Request) {
 	if err := validateGeminiFilesVersion(r.PathValue("version")); err != nil {
 		writeError(w, http.StatusBadRequest, err.Error(), clients.ErrorCode(err))
@@ -440,11 +479,18 @@ func (s *Server) startGeminiFilesUpload(w http.ResponseWriter, r *http.Request)
 		CreatedAt:   time.Now(),
 	}
 	s.geminiUploadSessions.Store(uploadID, session)
-	w.Header().Set("X-Goog-Upload-URL", absoluteRequestURL(r, "/upload/"+session.Version+"/files/"+uploadID))
+	w.Header().Set("X-Goog-Upload-URL", absoluteRequestURL(r, geminiUploadPath(r, session.Version, uploadID)))
 	w.Header().Set("X-Goog-Upload-Status", "active")
 	writeJSON(w, http.StatusOK, map[string]any{})
 }
 
+func geminiUploadPath(r *http.Request, version string, uploadID string) string {
+	if strings.HasPrefix(r.URL.Path, "/api/v1/gemini/upload/") {
+		return "/api/v1/gemini/upload/" + version + "/files/" + uploadID
+	}
+	return "/upload/" + version + "/files/" + uploadID
+}
+
 func (s *Server) finalizeGeminiFilesUpload(w http.ResponseWriter, r *http.Request, uploadID string, session geminiUploadSession) {
 	if uploadID == "" {
 		uploadID = newGeminiUploadID()

apps/api/internal/httpapi/gemini_compat_test.go

@@ -102,6 +102,18 @@ func TestRegisterGeminiGenerateContentRoutes(t *testing.T) {
 	}
 }
 
+func TestGeminiUploadPathKeepsCanonicalV1Prefix(t *testing.T) {
+	canonical := httptest.NewRequest(http.MethodPost, "/api/v1/gemini/upload/v1beta/files", nil)
+	if got := geminiUploadPath(canonical, "v1beta", "upload-1"); got != "/api/v1/gemini/upload/v1beta/files/upload-1" {
+		t.Fatalf("canonical upload path = %q", got)
+	}
+
+	legacy := httptest.NewRequest(http.MethodPost, "/upload/v1beta/files", nil)
+	if got := geminiUploadPath(legacy, "v1beta", "upload-1"); got != "/upload/v1beta/files/upload-1" {
+		t.Fatalf("legacy upload path = %q", got)
+	}
+}
+
 func TestGeminiImageTaskBodyMapsTextOnlyToImageGenerate(t *testing.T) {
 	mapping, err := geminiImageTaskBody("gemini-image", map[string]any{
 		"contents": []any{

apps/api/internal/httpapi/handlers.go

@@ -31,7 +31,7 @@ const (
 // @Tags system
 // @Produce json
 // @Success 200 {object} HealthResponse
-// @Router /healthz [get]
+// @Router /api/v1/healthz [get]
 func (s *Server) health(w http.ResponseWriter, r *http.Request) {
 	writeJSON(w, http.StatusOK, map[string]any{
 		"ok":           true,
@@ -48,7 +48,7 @@ func (s *Server) health(w http.ResponseWriter, r *http.Request) {
 // @Produce json
 // @Success 200 {object} ReadyResponse
 // @Failure 503 {object} ErrorEnvelope
-// @Router /readyz [get]
+// @Router /api/v1/readyz [get]
 func (s *Server) ready(w http.ResponseWriter, r *http.Request) {
 	ctx, cancel := context.WithTimeout(r.Context(), postgresReadinessTimeout)
 	defer cancel()
@@ -1021,7 +1021,7 @@ func (s *Server) listModelRateLimitStatuses(w http.ResponseWriter, r *http.Reque
 
 // createTask godoc
 // @Summary 创建或执行 AI 任务
-// @Description 网关任务接口按 model 选择平台模型；除 /api/v1/chat/completions 以外的 /api/v1 任务路径返回任务受理结果，OpenAI-compatible 路径同步返回兼容响应或 SSE 流。
+// @Description 统一公开入口按 model 选择平台模型并默认同步返回兼容响应；设置 X-Async=true 时异步创建任务并返回 202。
 // @Tags tasks
 // @Accept json
 // @Produce json
@@ -1047,22 +1047,6 @@ func (s *Server) listModelRateLimitStatuses(w http.ResponseWriter, r *http.Reque
 // @Router /api/v1/music/generations [post]
 // @Router /api/v1/speech/generations [post]
 // @Router /api/v1/voice_clone [post]
-// @Router /embeddings [post]
-// @Router /v1/embeddings [post]
-// @Router /reranks [post]
-// @Router /v1/reranks [post]
-// @Router /images/generations [post]
-// @Router /v1/images/generations [post]
-// @Router /images/edits [post]
-// @Router /v1/images/edits [post]
-// @Router /song/generations [post]
-// @Router /v1/song/generations [post]
-// @Router /music/generations [post]
-// @Router /v1/music/generations [post]
-// @Router /speech/generations [post]
-// @Router /v1/speech/generations [post]
-// @Router /voice_clone [post]
-// @Router /v1/voice_clone [post]
 func (s *Server) createTask(kind string, compatible bool) http.Handler {
 	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		user, ok := auth.UserFromContext(r.Context())
@@ -1214,8 +1198,6 @@ func (s *Server) createAPIV1ChatCompletions() http.Handler {
 // @Failure 403 {object} ErrorEnvelope
 // @Failure 429 {object} ErrorEnvelope
 // @Failure 502 {object} ErrorEnvelope
-// @Router /chat/completions [post]
-// @Router /v1/chat/completions [post]
 func openAIChatCompletionsDoc() {}
 
 // openAIResponsesDoc godoc
@@ -1234,8 +1216,6 @@ func openAIChatCompletionsDoc() {}
 // @Failure 402 {object} ErrorEnvelope
 // @Failure 403 {object} ErrorEnvelope
 // @Failure 503 {object} ErrorEnvelope "response_chain_unavailable"
-// @Router /responses [post]
-// @Router /v1/responses [post]
 // @Router /api/v1/responses [post]
 func openAIResponsesDoc() {}
 
@@ -1660,7 +1640,6 @@ func matchedRateLimitRule(policy map[string]any, metric string) map[string]any {
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/workspace/tasks [get]
 // @Router /api/v1/tasks [get]
-// @Router /tasks [get]
 func (s *Server) listTasks(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok {
@@ -1769,7 +1748,6 @@ func boolValue(body map[string]any, key string) bool {
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/workspace/tasks/{taskID} [get]
 // @Router /api/v1/tasks/{taskID} [get]
-// @Router /tasks/{taskID} [get]
 func (s *Server) getTask(w http.ResponseWriter, r *http.Request) {
 	task, err := s.store.GetTask(r.Context(), r.PathValue("taskID"))
 	if err == nil {
@@ -1802,8 +1780,6 @@ func (s *Server) getTask(w http.ResponseWriter, r *http.Request) {
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/workspace/tasks/{taskID}/cancel [post]
 // @Router /api/v1/tasks/{taskID}/cancel [post]
-// @Router /v1/tasks/{taskID}/cancel [post]
-// @Router /tasks/{taskID}/cancel [post]
 func (s *Server) cancelTask(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok {
@@ -1840,7 +1816,6 @@ func (s *Server) cancelTask(w http.ResponseWriter, r *http.Request) {
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/workspace/tasks/{taskID}/param-preprocessing [get]
 // @Router /api/v1/tasks/{taskID}/param-preprocessing [get]
-// @Router /tasks/{taskID}/param-preprocessing [get]
 func (s *Server) taskParamPreprocessing(w http.ResponseWriter, r *http.Request) {
 	task, err := s.store.GetTask(r.Context(), r.PathValue("taskID"))
 	if err != nil {
@@ -1874,7 +1849,6 @@ func (s *Server) taskParamPreprocessing(w http.ResponseWriter, r *http.Request)
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/workspace/tasks/{taskID}/events [get]
 // @Router /api/v1/tasks/{taskID}/events [get]
-// @Router /tasks/{taskID}/events [get]
 func (s *Server) taskEvents(w http.ResponseWriter, r *http.Request) {
 	task, err := s.store.GetTask(r.Context(), r.PathValue("taskID"))
 	if err != nil {

apps/api/internal/httpapi/keling_compat_handlers.go

@@ -132,7 +132,7 @@ func (s *Server) requireKelingAPIKey(next http.Handler) http.Handler {
 // @Failure 429 {object} KelingCompatibleEnvelope
 // @Failure 500 {object} KelingCompatibleEnvelope
 // @Failure 503 {object} KelingCompatibleEnvelope
-// @Router /v1/videos/omni-video [post]
+// @Router /api/v1/videos/omni-video [post]
 func (s *Server) createKelingOmniVideo(w http.ResponseWriter, r *http.Request) {
 	requestID := kelingCompatRequestID(r)
 	user, ok := auth.UserFromContext(r.Context())
@@ -206,7 +206,7 @@ func (s *Server) createKelingOmniVideo(w http.ResponseWriter, r *http.Request) {
 // @Failure 403 {object} KelingCompatibleEnvelope
 // @Failure 404 {object} KelingCompatibleEnvelope
 // @Failure 500 {object} KelingCompatibleEnvelope
-// @Router /v1/videos/omni-video/{taskID} [get]
+// @Router /api/v1/videos/omni-video/{taskID} [get]
 func (s *Server) getKelingOmniVideo(w http.ResponseWriter, r *http.Request) {
 	requestID := kelingCompatRequestID(r)
 	user, ok := auth.UserFromContext(r.Context())

apps/api/internal/httpapi/keling_compat_integration_test.go

@@ -145,7 +145,7 @@ func TestKelingOmniCompatibleHTTPFlow(t *testing.T) {
 	_, secondAPIKey := createKelingCompatIntegrationUser(t, ctx, db, gateway.URL, "second", suffix, false)
 
 	var created KelingCompatibleEnvelope
-	doJSON(t, gateway.URL, http.MethodPost, "/v1/videos/omni-video", firstAPIKey, map[string]any{
+	doJSON(t, gateway.URL, http.MethodPost, "/api/v1/videos/omni-video", firstAPIKey, map[string]any{
 		"model_name":       "kling-video-o1",
 		"prompt":           "A clean product reveal",
 		"mode":             "std",
@@ -162,7 +162,7 @@ func TestKelingOmniCompatibleHTTPFlow(t *testing.T) {
 	taskID := stringFromKelingCompat(createdData["task_id"])
 
 	var hidden KelingCompatibleEnvelope
-	doJSON(t, gateway.URL, http.MethodGet, "/v1/videos/omni-video/"+taskID, secondAPIKey, nil, http.StatusNotFound, &hidden)
+	doJSON(t, gateway.URL, http.MethodGet, "/api/v1/videos/omni-video/"+taskID, secondAPIKey, nil, http.StatusNotFound, &hidden)
 	if hidden.Code != 1203 {
 		t.Fatalf("cross-user task must be hidden: %+v", hidden)
 	}
@@ -278,7 +278,7 @@ func waitForKelingCompatTask(t *testing.T, baseURL string, apiKey string, taskID
 	deadline := time.Now().Add(timeout)
 	for time.Now().Before(deadline) {
 		var response KelingCompatibleEnvelope
-		doJSON(t, baseURL, http.MethodGet, "/v1/videos/omni-video/"+taskID, apiKey, nil, http.StatusOK, &response)
+		doJSON(t, baseURL, http.MethodGet, "/api/v1/videos/omni-video/"+taskID, apiKey, nil, http.StatusOK, &response)
 		data, _ := response.Data.(map[string]any)
 		switch data["task_status"] {
 		case "succeed":

apps/api/internal/httpapi/keling_simulation_integration_test.go

@@ -134,16 +134,22 @@ WHERE username = $1`, username); err != nil {
 					ResponseDurationMS int64          `json:"responseDurationMs"`
 				} `json:"task"`
 			}
-			doJSON(
+			doJSONWithHeaders(
 				t,
 				server.URL,
 				http.MethodPost,
 				"/api/v1/videos/generations",
 				apiKeyResponse.Secret,
 				request,
+				map[string]string{"X-Async": "true"},
 				http.StatusAccepted,
 				&response,
 			)
+			if response.Task.ID == "" {
+				t.Fatal("async Kling simulation response did not return a task id")
+			}
+			waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, response.Task.ID, []string{"succeeded"}, 10*time.Second)
+			doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+response.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &response.Task)
 
 			task := response.Task
 			if task.ID == "" ||

apps/api/internal/httpapi/kling_compat.go

@@ -27,6 +27,15 @@ func (s *Server) registerKlingCompatibilityRoutes(mux *http.ServeMux) {
 	handler := func(next http.HandlerFunc) http.Handler {
 		return s.requireUser(auth.PermissionBasic, http.HandlerFunc(next))
 	}
+	// /api/v1 is the canonical public prefix. The historical /kling paths
+	// remain registered below so existing clients can migrate without downtime.
+	mux.Handle("POST /api/v1/kling/v1/videos/omni-video", handler(s.klingV1CreateOmniVideo))
+	mux.Handle("GET /api/v1/kling/v1/videos/omni-video", handler(s.klingV1ListOmniVideos))
+	mux.Handle("GET /api/v1/kling/v1/videos/omni-video/{taskID}", handler(s.klingV1GetOmniVideo))
+	mux.Handle("POST /api/v1/kling/v2/omni-video/{model}", handler(s.klingV2CreateOmniVideo))
+	mux.Handle("GET /api/v1/kling/v2/tasks", handler(s.klingV2GetTasks))
+	mux.Handle("POST /api/v1/kling/v2/tasks", handler(s.klingV2ListTasks))
+
 	mux.Handle("POST /kling/v1/videos/omni-video", handler(s.klingV1CreateOmniVideo))
 	mux.Handle("GET /kling/v1/videos/omni-video", handler(s.klingV1ListOmniVideos))
 	mux.Handle("GET /kling/v1/videos/omni-video/{taskID}", handler(s.klingV1GetOmniVideo))
@@ -52,7 +61,7 @@ func (s *Server) registerKlingCompatibilityRoutes(mux *http.ServeMux) {
 // @Success 200 {object} map[string]interface{}
 // @Failure 400 {object} map[string]interface{}
 // @Failure 401 {object} ErrorEnvelope
-// @Router /kling/v1/videos/omni-video [post]
+// @Router /api/v1/kling/v1/videos/omni-video [post]
 func (s *Server) klingV1CreateOmniVideo(w http.ResponseWriter, r *http.Request) {
 	var native map[string]any
 	if err := decodeKlingJSON(r, &native); err != nil {
@@ -78,7 +87,7 @@ func (s *Server) klingV1CreateOmniVideo(w http.ResponseWriter, r *http.Request)
 // @Success 200 {object} map[string]interface{}
 // @Failure 400 {object} map[string]interface{}
 // @Failure 401 {object} ErrorEnvelope
-// @Router /kling/omni-video/{model} [post]
+// @Router /api/v1/kling/v2/omni-video/{model} [post]
 func (s *Server) klingV2CreateOmniVideo(w http.ResponseWriter, r *http.Request) {
 	model, ok := klingV2ProviderModel(r.PathValue("model"))
 	if !ok {
@@ -433,7 +442,7 @@ func validateKlingCompatBody(model string, body map[string]any) error {
 // @Param taskID path string true "任务 ID"
 // @Success 200 {object} map[string]interface{}
 // @Failure 404 {object} map[string]interface{}
-// @Router /kling/v1/videos/omni-video/{taskID} [get]
+// @Router /api/v1/kling/v1/videos/omni-video/{taskID} [get]
 func (s *Server) klingV1GetOmniVideo(w http.ResponseWriter, r *http.Request) {
 	user, _ := auth.UserFromContext(r.Context())
 	task, err := s.store.GetCompatTask(r.Context(), user, klingCompatProvider, "v1", r.PathValue("taskID"))
@@ -456,7 +465,7 @@ func (s *Server) klingV1GetOmniVideo(w http.ResponseWriter, r *http.Request) {
 // @Param pageNum query int false "页码" default(1)
 // @Param pageSize query int false "每页数量" default(30)
 // @Success 200 {object} map[string]interface{}
-// @Router /kling/v1/videos/omni-video [get]
+// @Router /api/v1/kling/v1/videos/omni-video [get]
 func (s *Server) klingV1ListOmniVideos(w http.ResponseWriter, r *http.Request) {
 	user, _ := auth.UserFromContext(r.Context())
 	page, err := positiveQueryInt(r.URL.Query().Get("pageNum"), 1)
@@ -489,7 +498,7 @@ func (s *Server) klingV1ListOmniVideos(w http.ResponseWriter, r *http.Request) {
 // @Param task_ids query string false "逗号分隔的任务 ID"
 // @Param external_task_ids query string false "逗号分隔的外部任务 ID"
 // @Success 200 {object} map[string]interface{}
-// @Router /kling/tasks [get]
+// @Router /api/v1/kling/v2/tasks [get]
 func (s *Server) klingV2GetTasks(w http.ResponseWriter, r *http.Request) {
 	user, _ := auth.UserFromContext(r.Context())
 	taskIDs := splitKlingIDs(r.URL.Query().Get("task_ids"))
@@ -525,7 +534,7 @@ func (s *Server) klingV2GetTasks(w http.ResponseWriter, r *http.Request) {
 // @Security BearerAuth
 // @Param input body map[string]interface{} true "游标、数量、时间范围和筛选条件"
 // @Success 200 {object} map[string]interface{}
-// @Router /kling/tasks [post]
+// @Router /api/v1/kling/v2/tasks [post]
 func (s *Server) klingV2ListTasks(w http.ResponseWriter, r *http.Request) {
 	user, _ := auth.UserFromContext(r.Context())
 	var body map[string]any

apps/api/internal/httpapi/kling_compat_integration_test.go

@@ -29,6 +29,23 @@ func TestKlingCompatibilitySimulationFlow(t *testing.T) {
 	}
 	defer db.Close()
 
+	var upgradedBaseModels int
+	if err := db.Pool().QueryRow(ctx, `
+SELECT count(*)
+FROM base_model_catalog
+WHERE provider_key = 'keling'
+  AND provider_model_name IN ('kling-video-o1', 'kling-v3-omni')
+  AND model_type @> '["video_generate","image_to_video","omni_video"]'::jsonb
+  AND capabilities ? 'video_generate'
+  AND capabilities ? 'image_to_video'
+  AND capabilities ? 'omni_video'
+  AND metadata->'rawModel'->'types' @> '["video_generate","image_to_video","omni_video"]'::jsonb`).Scan(&upgradedBaseModels); err != nil {
+		t.Fatalf("read upgraded Kling Omni base model capabilities: %v", err)
+	}
+	if upgradedBaseModels != 2 {
+		t.Fatalf("expected both Kling Omni base models to expose base video capabilities, got %d", upgradedBaseModels)
+	}
+
 	serverCtx, cancelServer := context.WithCancel(ctx)
 	defer cancelServer()
 	server := httptest.NewServer(NewServerWithContext(serverCtx, config.Config{
@@ -89,11 +106,66 @@ func TestKlingCompatibilitySimulationFlow(t *testing.T) {
 			"displayName":       model,
 		}, http.StatusCreated, nil)
 	}
+	var upgradedPlatformModels int
+	if err := db.Pool().QueryRow(ctx, `
+SELECT count(*)
+FROM platform_models
+WHERE platform_id = $1::uuid
+  AND model_type @> '["video_generate","image_to_video","omni_video"]'::jsonb
+  AND capabilities ? 'video_generate'
+  AND capabilities ? 'image_to_video'
+  AND capabilities ? 'omni_video'`, platform.ID).Scan(&upgradedPlatformModels); err != nil {
+		t.Fatalf("read upgraded Kling Omni platform model capabilities: %v", err)
+	}
+	if upgradedPlatformModels != 2 {
+		t.Fatalf("expected both Kling Omni platform models to expose base video capabilities, got %d", upgradedPlatformModels)
+	}
+
+	assertGenericVideoGeneration := func(name string, model string, image string, expectedModelType string) {
+		t.Helper()
+		t.Run(name, func(t *testing.T) {
+			request := map[string]any{
+				"model":                model,
+				"prompt":               "通用视频接口模拟任务",
+				"duration":             5,
+				"resolution":           "720p",
+				"runMode":              "simulation",
+				"simulation":           true,
+				"simulationDurationMs": 5,
+			}
+			if image != "" {
+				request["image"] = image
+			}
+			var response struct {
+				Task struct {
+					ID            string `json:"id"`
+					Status        string `json:"status"`
+					ModelType     string `json:"modelType"`
+					ResolvedModel string `json:"resolvedModel"`
+				} `json:"task"`
+			}
+			doJSONWithHeaders(t, server.URL, http.MethodPost, "/api/v1/videos/generations", apiKeyResponse.Secret, request, map[string]string{"X-Async": "true"}, http.StatusAccepted, &response)
+			if response.Task.ID == "" {
+				t.Fatal("async generic video response did not return a task id")
+			}
+			waitForTaskStatus(t, server.URL, apiKeyResponse.Secret, response.Task.ID, []string{"succeeded"}, 10*time.Second)
+			doJSON(t, server.URL, http.MethodGet, "/api/v1/tasks/"+response.Task.ID, apiKeyResponse.Secret, nil, http.StatusOK, &response.Task)
+			resolvedModel, resolved := klingV2ProviderModel(response.Task.ResolvedModel)
+			if response.Task.Status != "succeeded" || response.Task.ModelType != expectedModelType || !resolved || resolvedModel != model {
+				t.Fatalf("generic video request without modelType should use inferred capability: %+v", response.Task)
+			}
+		})
+	}
+
+	for _, model := range []string{klingO1Model, klingV3OmniModel} {
+		assertGenericVideoGeneration(model+"-text-to-video", model, "", "video_generate")
+		assertGenericVideoGeneration(model+"-image-to-video", model, "https://example.com/first.png", "image_to_video")
+	}
 
 	createV1 := func(model string, duration int, externalID string) string {
 		t.Helper()
 		var response map[string]any
-		doJSON(t, server.URL, http.MethodPost, "/kling/v1/videos/omni-video", apiKeyResponse.Secret, map[string]any{
+		doJSON(t, server.URL, http.MethodPost, "/api/v1/kling/v1/videos/omni-video", apiKeyResponse.Secret, map[string]any{
 			"model_name":           model,
 			"prompt":               "兼容接口模拟任务",
 			"duration":             duration,
@@ -120,14 +192,14 @@ func TestKlingCompatibilitySimulationFlow(t *testing.T) {
 		waitKlingV1SimulationTask(t, server.URL, apiKeyResponse.Secret, taskID)
 	}
 	var listResponse map[string]any
-	doJSON(t, server.URL, http.MethodGet, "/kling/v1/videos/omni-video?pageNum=1&pageSize=10", apiKeyResponse.Secret, nil, http.StatusOK, &listResponse)
+	doJSON(t, server.URL, http.MethodGet, "/api/v1/kling/v1/videos/omni-video?pageNum=1&pageSize=10", apiKeyResponse.Secret, nil, http.StatusOK, &listResponse)
 	items, _ := listResponse["data"].([]any)
 	if len(items) < 2 {
 		t.Fatalf("V1 task list did not return compatibility tasks: %#v", listResponse)
 	}
 
 	var duplicateResponse map[string]any
-	doJSON(t, server.URL, http.MethodPost, "/kling/v1/videos/omni-video", apiKeyResponse.Secret, map[string]any{
+	doJSON(t, server.URL, http.MethodPost, "/api/v1/kling/v1/videos/omni-video", apiKeyResponse.Secret, map[string]any{
 		"model_name": klingO1Model, "prompt": "duplicate", "duration": 5,
 		"external_task_id": "compat-o1-" + suffix,
 		"runMode":          "simulation", "simulation": true,
@@ -137,7 +209,7 @@ func TestKlingCompatibilitySimulationFlow(t *testing.T) {
 	}
 
 	var v2Response map[string]any
-	doJSON(t, server.URL, http.MethodPost, "/kling/v2/omni-video/kling-v3-omni", apiKeyResponse.Secret, map[string]any{
+	doJSON(t, server.URL, http.MethodPost, "/api/v1/kling/v2/omni-video/kling-v3-omni", apiKeyResponse.Secret, map[string]any{
 		"contents":             []any{map[string]any{"type": "prompt", "text": "API 2.0 模拟任务"}},
 		"settings":             map[string]any{"duration": 3, "resolution": "720p", "aspect_ratio": "16:9", "audio": "off"},
 		"options":              map[string]any{"external_task_id": "compat-v2-" + suffix},
@@ -158,7 +230,7 @@ func waitKlingV1SimulationTask(t *testing.T, baseURL string, apiKey string, task
 	deadline := time.Now().Add(5 * time.Second)
 	for time.Now().Before(deadline) {
 		var response map[string]any
-		doJSON(t, baseURL, http.MethodGet, "/kling/v1/videos/omni-video/"+taskID, apiKey, nil, http.StatusOK, &response)
+		doJSON(t, baseURL, http.MethodGet, "/api/v1/kling/v1/videos/omni-video/"+taskID, apiKey, nil, http.StatusOK, &response)
 		data, _ := response["data"].(map[string]any)
 		switch data["task_status"] {
 		case "succeed":
@@ -176,7 +248,7 @@ func waitKlingV2SimulationTask(t *testing.T, baseURL string, apiKey string, task
 	deadline := time.Now().Add(5 * time.Second)
 	for time.Now().Before(deadline) {
 		var response map[string]any
-		doJSON(t, baseURL, http.MethodGet, "/kling/v2/tasks?task_ids="+taskID, apiKey, nil, http.StatusOK, &response)
+		doJSON(t, baseURL, http.MethodGet, "/api/v1/kling/v2/tasks?task_ids="+taskID, apiKey, nil, http.StatusOK, &response)
 		items, _ := response["data"].([]any)
 		if len(items) == 1 {
 			data, _ := items[0].(map[string]any)

apps/api/internal/httpapi/openapi_models.go

@@ -23,8 +23,8 @@ type SkillBundleMetadataResponse struct {
 	Modules         []string `json:"modules" example:"model-runtime"`
 	FileName        string   `json:"fileName" example:"ai-gateway-ops-management-v1.0.2.zip"`
 	DownloadPath    string   `json:"downloadPath" example:"/api/v1/public/skills/ai-gateway-ops-management/download"`
-	APIDocsJSONPath string   `json:"apiDocsJsonPath" example:"/api-docs-json"`
-	APIDocsYAMLPath string   `json:"apiDocsYamlPath" example:"/api-docs-yaml"`
+	APIDocsJSONPath string   `json:"apiDocsJsonPath" example:"/api/v1/openapi.json"`
+	APIDocsYAMLPath string   `json:"apiDocsYamlPath" example:"/api/v1/openapi.yaml"`
 }
 
 type ErrorEnvelope struct {

apps/api/internal/httpapi/public_api_prefix_test.go

@@ -0,0 +1,72 @@
+package httpapi
+
+import (
+	"encoding/json"
+	"strings"
+	"testing"
+
+	gatewaydocs "github.com/easyai/easyai-ai-gateway/apps/api/docs"
+)
+
+func TestOpenAPIPublicRoutesUseCanonicalV1Prefix(t *testing.T) {
+	var document struct {
+		Paths map[string]any `json:"paths"`
+	}
+	if err := json.Unmarshal(gatewaydocs.SwaggerJSON, &document); err != nil {
+		t.Fatalf("decode embedded OpenAPI document: %v", err)
+	}
+
+	legacyPrefixes := []string{
+		"/v1/",
+		"/v1beta/",
+		"/kling/",
+		"/upload/",
+		"/api/v3/",
+		"/chat/",
+		"/images/",
+		"/song/",
+		"/music/",
+		"/speech/",
+		"/voice_clone",
+		"/tasks",
+	}
+	legacyExact := map[string]bool{
+		"/healthz":       true,
+		"/readyz":        true,
+		"/api-docs-json": true,
+		"/api-docs-yaml": true,
+		"/responses":     true,
+		"/embeddings":    true,
+		"/reranks":       true,
+	}
+	for route := range document.Paths {
+		if legacyExact[route] {
+			t.Errorf("legacy public route must not be advertised in OpenAPI: %s", route)
+		}
+		for _, prefix := range legacyPrefixes {
+			if strings.HasPrefix(route, prefix) {
+				t.Errorf("legacy public route must not be advertised in OpenAPI: %s", route)
+			}
+		}
+	}
+
+	required := []string{
+		"/api/v1/healthz",
+		"/api/v1/readyz",
+		"/api/v1/openapi.json",
+		"/api/v1/chat/completions",
+		"/api/v1/responses",
+		"/api/v1/images/generations",
+		"/api/v1/videos/generations",
+		"/api/v1/models/{model}:generateContent",
+		"/api/v1/videos/omni-video",
+		"/api/v1/kling/v1/videos/omni-video",
+		"/api/v1/kling/v2/omni-video/{model}",
+		"/api/v1/contents/generations/tasks",
+	}
+	for _, route := range required {
+		if _, ok := document.Paths[route]; !ok {
+			t.Errorf("canonical public route missing from OpenAPI: %s", route)
+		}
+	}
+}

apps/api/internal/httpapi/security_events_test.go

@@ -40,6 +40,18 @@ func TestReceiveSecurityEventUsesPreparedReceiverBeforeFirstActivation(t *testin
 	}
 }
 
+func TestReceiveSecurityEventReturnsNotFoundWithoutConfiguredReceiver(t *testing.T) {
+	server := &Server{identityRuntime: identityruntime.NewManager(nil, &preparedReceiverBuilder{})}
+	request := httptest.NewRequest(http.MethodPost, "/api/v1/security-events/ssf", nil)
+	response := httptest.NewRecorder()
+
+	server.receiveSecurityEvent(response, request)
+
+	if response.Code != http.StatusNotFound {
+		t.Fatalf("unconfigured SSF status=%d, want %d", response.Code, http.StatusNotFound)
+	}
+}
+
 func TestSecurityEventWriteAuditFailsClosedBeforeMutation(t *testing.T) {
 	server := &Server{}
 	request := httptest.NewRequest(http.MethodPost, "/api/admin/system/security-events/connection/verify", nil)

apps/api/internal/httpapi/server.go

@@ -127,12 +127,16 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
 	mux := http.NewServeMux()
 	mux.HandleFunc("GET /healthz", server.health)
 	mux.HandleFunc("GET /readyz", server.ready)
+	mux.HandleFunc("GET /api/v1/healthz", server.health)
+	mux.HandleFunc("GET /api/v1/readyz", server.ready)
 	mux.Handle("GET /metrics", securityEventMetrics.DynamicHandler(db))
 	mux.HandleFunc("GET /static/simulation/{asset}", serveSimulationAsset)
 	mux.HandleFunc("GET /static/generated/{asset}", server.serveGeneratedStaticAsset)
 	mux.HandleFunc("GET /static/uploaded/{asset}", server.serveUploadedStaticAsset)
 	mux.HandleFunc("GET /api-docs-json", server.apiDocsJSON)
 	mux.HandleFunc("GET /api-docs-yaml", server.apiDocsYAML)
+	mux.HandleFunc("GET /api/v1/openapi.json", server.apiDocsJSON)
+	mux.HandleFunc("GET /api/v1/openapi.yaml", server.apiDocsYAML)
 	mux.HandleFunc("GET /api/v1/public/skills/ai-gateway-ops-management/metadata", server.getOpsManagementSkillMetadata)
 	mux.HandleFunc("GET /api/v1/public/skills/ai-gateway-ops-management/download", server.downloadOpsManagementSkill)
 
@@ -252,12 +256,12 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
 	mux.Handle("GET /api/admin/runtime/rate-limit-windows", server.requireAdmin(auth.PermissionPower, http.HandlerFunc(server.listRateLimitWindows)))
 	mux.Handle("GET /api/admin/runtime/model-rate-limits", server.requireAdmin(auth.PermissionPower, http.HandlerFunc(server.listModelRateLimitStatuses)))
 	mux.Handle("POST /api/v1/chat/completions", server.requireUser(auth.PermissionBasic, server.createAPIV1ChatCompletions()))
-	mux.Handle("POST /api/v1/responses", server.requireUser(auth.PermissionBasic, server.createTask("responses", false)))
-	mux.Handle("POST /api/v1/embeddings", server.requireUser(auth.PermissionBasic, server.createTask("embeddings", false)))
-	mux.Handle("POST /api/v1/reranks", server.requireUser(auth.PermissionBasic, server.createTask("reranks", false)))
-	mux.Handle("POST /api/v1/images/generations", server.requireUser(auth.PermissionBasic, server.createTask("images.generations", false)))
-	mux.Handle("POST /api/v1/images/edits", server.requireUser(auth.PermissionBasic, server.createTask("images.edits", false)))
-	mux.Handle("POST /api/v1/videos/generations", server.requireUser(auth.PermissionBasic, server.createTask("videos.generations", false)))
+	mux.Handle("POST /api/v1/responses", server.requireUser(auth.PermissionBasic, server.createTask("responses", true)))
+	mux.Handle("POST /api/v1/embeddings", server.requireUser(auth.PermissionBasic, server.createTask("embeddings", true)))
+	mux.Handle("POST /api/v1/reranks", server.requireUser(auth.PermissionBasic, server.createTask("reranks", true)))
+	mux.Handle("POST /api/v1/images/generations", server.requireUser(auth.PermissionBasic, server.createTask("images.generations", true)))
+	mux.Handle("POST /api/v1/images/edits", server.requireUser(auth.PermissionBasic, server.createTask("images.edits", true)))
+	mux.Handle("POST /api/v1/videos/generations", server.requireUser(auth.PermissionBasic, server.createTask("videos.generations", true)))
 	mux.Handle("POST /api/v1/video/generations", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.createLegacyVolcesVideoGeneration)))
 	mux.Handle("GET /api/v1/ai/result/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.getLegacyVolcesVideoResult)))
 	mux.Handle("POST /api/v1/song/generations", server.requireUser(auth.PermissionBasic, server.createTask("song.generations", true)))
@@ -275,10 +279,16 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
 	mux.Handle("GET /api/v3/contents/generations/tasks", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.listVolcesContentsGenerationTasks)))
 	mux.Handle("GET /api/v3/contents/generations/tasks/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.getVolcesContentsGenerationTask)))
 	mux.Handle("DELETE /api/v3/contents/generations/tasks/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.deleteVolcesContentsGenerationTask)))
+	mux.Handle("POST /api/v1/contents/generations/tasks", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.createVolcesContentsGenerationTask)))
+	mux.Handle("GET /api/v1/contents/generations/tasks", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.listVolcesContentsGenerationTasks)))
+	mux.Handle("GET /api/v1/contents/generations/tasks/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.getVolcesContentsGenerationTask)))
+	mux.Handle("DELETE /api/v1/contents/generations/tasks/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.deleteVolcesContentsGenerationTask)))
 	server.registerGeminiGenerateContentRoutes(mux)
 	server.registerKlingCompatibilityRoutes(mux)
 	mux.Handle("POST /upload/{version}/files", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.geminiFilesUpload)))
 	mux.Handle("POST /upload/{version}/files/{uploadID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.geminiFilesUploadFinalize)))
+	mux.Handle("POST /api/v1/gemini/upload/{version}/files", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.geminiFilesUpload)))
+	mux.Handle("POST /api/v1/gemini/upload/{version}/files/{uploadID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.geminiFilesUploadFinalize)))
 	mux.Handle("GET /api/v1/tasks", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.listTasks)))
 	mux.Handle("GET /api/v1/tasks/{taskID}", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.getTask)))
 	mux.Handle("POST /api/v1/tasks/{taskID}/cancel", server.requireUser(auth.PermissionBasic, http.HandlerFunc(server.cancelTask)))
@@ -303,6 +313,8 @@ func NewServerWithContext(ctx context.Context, cfg config.Config, db *store.Stor
 	mux.Handle("POST /v1/images/edits", server.requireUser(auth.PermissionBasic, server.createTask("images.edits", true)))
 	mux.Handle("POST /v1/videos/omni-video", server.requireKelingAPIKey(http.HandlerFunc(server.createKelingOmniVideo)))
 	mux.Handle("GET /v1/videos/omni-video/{taskID}", server.requireKelingAPIKey(http.HandlerFunc(server.getKelingOmniVideo)))
+	mux.Handle("POST /api/v1/videos/omni-video", server.requireKelingAPIKey(http.HandlerFunc(server.createKelingOmniVideo)))
+	mux.Handle("GET /api/v1/videos/omni-video/{taskID}", server.requireKelingAPIKey(http.HandlerFunc(server.getKelingOmniVideo)))
 	mux.Handle("POST /song/generations", server.requireUser(auth.PermissionBasic, server.createTask("song.generations", true)))
 	mux.Handle("POST /v1/song/generations", server.requireUser(auth.PermissionBasic, server.createTask("song.generations", true)))
 	mux.Handle("POST /music/generations", server.requireUser(auth.PermissionBasic, server.createTask("music.generations", true)))

apps/api/internal/httpapi/voice_clone_handlers.go

@@ -16,8 +16,6 @@ import (
 // @Failure 401 {object} ErrorEnvelope
 // @Failure 500 {object} ErrorEnvelope
 // @Router /api/v1/voice_clone/voices [get]
-// @Router /v1/voice_clone/voices [get]
-// @Router /voice_clone/voices [get]
 func (s *Server) listClonedVoices(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok {
@@ -51,8 +49,6 @@ func (s *Server) listClonedVoices(w http.ResponseWriter, r *http.Request) {
 // @Failure 404 {object} ErrorEnvelope
 // @Failure 502 {object} ErrorEnvelope
 // @Router /api/v1/voice_clone/voices/{voiceID} [delete]
-// @Router /v1/voice_clone/voices/{voiceID} [delete]
-// @Router /voice_clone/voices/{voiceID} [delete]
 func (s *Server) deleteClonedVoice(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok {

apps/api/internal/httpapi/volces_compat_handlers.go

@@ -17,13 +17,13 @@ const volcesContentsCompatibilityMarker = "volces_contents_generations_v3"
 
 // createVolcesContentsGenerationTask godoc
 // @Summary 创建火山内容生成任务
-// @Description 兼容火山方舟 POST /api/v3/contents/generations/tasks。网关 task id 是查询与取消用的公开 id；上游 id 另以 upstream_task_id 保留。
+// @Description 统一公开入口兼容火山方舟内容生成任务。网关 task id 是查询与取消用的公开 id；上游 id 另以 upstream_task_id 保留。
 // @Tags volces-compatible
 // @Accept json
 // @Produce json
 // @Security BearerAuth
 // @Success 200 {object} map[string]any
-// @Router /api/v3/contents/generations/tasks [post]
+// @Router /api/v1/contents/generations/tasks [post]
 func (s *Server) createVolcesContentsGenerationTask(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok || user == nil {
@@ -49,7 +49,7 @@ func (s *Server) createVolcesContentsGenerationTask(w http.ResponseWriter, r *ht
 // @Produce json
 // @Security BearerAuth
 // @Success 200 {object} map[string]any
-// @Router /api/v3/contents/generations/tasks/{taskID} [get]
+// @Router /api/v1/contents/generations/tasks/{taskID} [get]
 func (s *Server) getVolcesContentsGenerationTask(w http.ResponseWriter, r *http.Request) {
 	task, ok := s.volcesCompatibleTaskForUser(w, r)
 	if !ok {
@@ -64,7 +64,7 @@ func (s *Server) getVolcesContentsGenerationTask(w http.ResponseWriter, r *http.
 // @Produce json
 // @Security BearerAuth
 // @Success 200 {object} map[string]any
-// @Router /api/v3/contents/generations/tasks [get]
+// @Router /api/v1/contents/generations/tasks [get]
 func (s *Server) listVolcesContentsGenerationTasks(w http.ResponseWriter, r *http.Request) {
 	user, ok := auth.UserFromContext(r.Context())
 	if !ok || user == nil {
@@ -105,7 +105,7 @@ func (s *Server) listVolcesContentsGenerationTasks(w http.ResponseWriter, r *htt
 // @Produce json
 // @Security BearerAuth
 // @Success 200 {object} map[string]any
-// @Router /api/v3/contents/generations/tasks/{taskID} [delete]
+// @Router /api/v1/contents/generations/tasks/{taskID} [delete]
 func (s *Server) deleteVolcesContentsGenerationTask(w http.ResponseWriter, r *http.Request) {
 	task, ok := s.volcesCompatibleTaskForUser(w, r)
 	if !ok {

apps/api/internal/identityruntime/manager.go

@@ -105,7 +105,10 @@ func (manager *Manager) SecurityEventReceiver() http.Handler {
 	if runtime := manager.Current(); runtime != nil && runtime.SecurityEvents != nil {
 		return runtime.SecurityEvents
 	}
-	return manager.SecurityEventManager()
+	if securityEventManager := manager.SecurityEventManager(); securityEventManager != nil {
+		return securityEventManager
+	}
+	return nil
 }
 
 // SecurityEventManager resolves the manager used by administrative recovery

apps/api/internal/identityruntime/manager_test.go

@@ -630,6 +630,14 @@ func TestSecurityEventManagerExposesPreparedRecoveryManagerWithoutActiveRuntime(
 	}
 }
 
+func TestSecurityEventReceiverReturnsNilWithoutConfiguredManager(t *testing.T) {
+	manager := NewManager(&runtimeRepositoryFake{}, &runtimeBuilderFake{})
+
+	if manager.SecurityEventReceiver() != nil {
+		t.Fatal("unconfigured security event receiver should be nil")
+	}
+}
+
 func TestSecurityEventManagerPrefersActiveRuntime(t *testing.T) {
 	active := &securityevents.ConnectionManager{}
 	prepared := &securityevents.ConnectionManager{}

apps/api/internal/skillbundle/content/skills/ai-gateway-ops-management/SKILL.md

@@ -18,7 +18,7 @@ Use this skill to operate AI Gateway administration APIs through documented, evi
 - Reuse existing pricing rules, runtime policy sets, providers, protocol clients, base models, and platforms whenever their effective behavior satisfies the target. Do not create a near-duplicate resource merely because the upstream account, base URL, or provider-side model name differs.
 - Prefer a supported standard client before using `universal` scripts. Use custom scripts only when the upstream contract cannot be represented by the existing OpenAI, Gemini, or provider-specific clients.
 - Do not invent platform config fields or assume an arbitrary config key is enforced. For `universal`, use only the recognized keys documented in `references/model-universal-platforms.md`; treat any extra key as script-owned data available through `context.env`.
-- Use the module references as the primary API source. Only when the required API is absent, inspect `<gateway-api-base-url>/api-docs-json`; continue only when path, method, schema, authentication, permission, and side effects are unambiguous.
+- Use the module references as the primary API source. Only when the required API is absent, inspect `<gateway-origin>/api/v1/openapi.json`; continue only when path, method, schema, authentication, permission, and side effects are unambiguous.
 
 ## Module Routing
 

apps/api/internal/skillbundle/content/skills/ai-gateway-ops-management/references/api-discovery-and-safety.md

@@ -2,7 +2,7 @@
 
 ## Required Inputs
 
-- Gateway API base URL. When using the bundled Web deployment this commonly includes `/gateway-api`; direct API access commonly uses port `8088`.
+- Gateway origin and public API base URL. Public API access always ends with `/api/v1`; direct local access commonly uses `http://127.0.0.1:8088/api/v1`.
 - Administrator JWT with the `manager` or `admin` role.
 - Target provider documentation and authorization material.
 - Clear requested outcome and whether real upstream calls are allowed.
@@ -10,7 +10,8 @@
 Do not place credentials in files or reusable commands. Use shell environment variables:
 
 ```bash
-export GATEWAY_BASE_URL='https://gateway.example.com/gateway-api'
+export GATEWAY_ORIGIN='https://gateway.example.com'
+export GATEWAY_PUBLIC_API_BASE="$GATEWAY_ORIGIN/api/v1"
 export GATEWAY_ADMIN_TOKEN='<administrator-jwt>'
 ```
 
@@ -24,7 +25,7 @@ For standalone or hybrid deployments, local login can return a JWT:
 curl --fail-with-body \
   -H 'Content-Type: application/json' \
   -d '{"account":"<admin-account>","password":"<admin-password>"}' \
-  "$GATEWAY_BASE_URL/api/v1/auth/login"
+  "$GATEWAY_PUBLIC_API_BASE/auth/login"
 ```
 
 Do not use local login when the deployment requires OIDC or server-main identity. Obtain the deployment's administrator access token instead.
@@ -34,7 +35,7 @@ Verify identity and role before writes:
 ```bash
 curl --fail-with-body \
   -H "Authorization: Bearer $GATEWAY_ADMIN_TOKEN" \
-  "$GATEWAY_BASE_URL/api/v1/me"
+  "$GATEWAY_PUBLIC_API_BASE/me"
 ```
 
 ## Request Pattern
@@ -47,7 +48,7 @@ curl --fail-with-body \
   -H 'Content-Type: application/json' \
   -X POST \
   -d '<json-body>' \
-  "$GATEWAY_BASE_URL/api/admin/<resource>"
+  "$GATEWAY_ORIGIN/api/admin/<resource>"
 ```
 
 Always read current state before PATCH, DELETE, reset, disable, or full replacement. PATCH handlers for providers, base models, pricing rule sets, runtime policy sets, runner policy, and platforms write complete resource shapes rather than merging every omitted field.
@@ -67,7 +68,7 @@ Obtain explicit confirmation after showing the current snapshot and impact befor
 
 The live machine-readable documents are:
 
-- `<gateway-api-base-url>/api-docs-json`
-- `<gateway-api-base-url>/api-docs-yaml`
+- `<gateway-origin>/api/v1/openapi.json`
+- `<gateway-origin>/api/v1/openapi.yaml`
 
 Use them only when this Skill does not document the required operation. Before acting, confirm the exact path, method, body, authentication, permission, response, and side effect. Do not infer a write operation from a similarly named endpoint.

apps/api/internal/skillbundle/content/skills/ai-gateway-ops-management/references/model-acceptance-runbook.md

@@ -33,7 +33,7 @@ Use an authorized user JWT:
 ```bash
 curl --fail-with-body \
   -H "Authorization: Bearer <user-jwt>" \
-  "$GATEWAY_BASE_URL/api/v1/model-catalog"
+  "$GATEWAY_PUBLIC_API_BASE/model-catalog"
 ```
 
 Confirm model alias, model types, provider source, effective capabilities, pricing summary, rate limits, permissions, and enabled state.
@@ -57,7 +57,7 @@ curl --fail-with-body \
     "simulation": true,
     "stream": false
   }' \
-  "$GATEWAY_BASE_URL/v1/chat/completions"
+  "$GATEWAY_PUBLIC_API_BASE/chat/completions"
 ```
 
 Simulation verifies Gateway routing, permissions, parameter normalization, pricing, and task behavior, but it does not execute the real universal submit or poll scripts. Validate universal scripts separately against a local mock or approved provider test environment before enabling production traffic.
@@ -82,4 +82,4 @@ With explicit approval, run one real minimal request and verify upstream request
 
 ## Final Report
 
-Report resource IDs, before/after behavior, requests used for verification, simulation or real mode, billing evidence, remaining risks, rollback readiness, and whether `/api-docs-json` was used. Never include credentials or raw secret-bearing payloads.
+Report resource IDs, before/after behavior, requests used for verification, simulation or real mode, billing evidence, remaining risks, rollback readiness, and whether `/api/v1/openapi.json` was used. Never include credentials or raw secret-bearing payloads.

apps/api/internal/store/base_models.go

@@ -488,6 +488,8 @@ func modelTypeAliases(value string) []string {
 		return []string{"image_edit"}
 	case "video", "videos.generations":
 		return []string{"video_generate"}
+	case "omni_video":
+		return []string{"video_generate", "image_to_video", "omni_video"}
 	case "song", "music", "song.generations", "music.generations", "music_generate":
 		return []string{"audio_generate"}
 	case "speech", "speech.generations", "tts":

apps/api/internal/store/candidates_test.go

@@ -9,6 +9,19 @@ func TestNormalizeModelMatchKeyRemovesWhitespace(t *testing.T) {
 	}
 }
 
+func TestNormalizeModelTypeListExpandsOmniVideoBaseCapabilities(t *testing.T) {
+	got := normalizeModelTypeList([]string{"omni_video"})
+	want := StringList{"video_generate", "image_to_video", "omni_video"}
+	if len(got) != len(want) {
+		t.Fatalf("omni_video should include text-to-video and image-to-video capabilities: got=%v want=%v", got, want)
+	}
+	for index := range want {
+		if got[index] != want[index] {
+			t.Fatalf("omni_video capability mismatch at %d: got=%v want=%v", index, got, want)
+		}
+	}
+}
+
 func TestTaskBillingModelIdentityKeepsRequestedModelPrimary(t *testing.T) {
 	identity := taskBillingModelIdentity(GatewayTask{
 		Model:          "doubao-5.0 图像编辑",

apps/api/migrations/0074_keling_omni_base_video_capabilities.sql

@@ -0,0 +1,182 @@
+WITH keling_omni_models(provider_model_name, model_type, capabilities) AS (
+  VALUES
+    (
+      'kling-video-o1',
+      '["video_generate","image_to_video","omni_video"]'::jsonb,
+      '{
+        "video_generate": {
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "output_audio": false,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "image_to_video": {
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "input_first_frame": true,
+          "input_last_frame": false,
+          "input_first_last_frame": true,
+          "input_reference_generate_single": true,
+          "input_reference_generate_multiple": true,
+          "max_images": 7,
+          "max_images_for_last_frame": 2,
+          "support_video_effect_template": false,
+          "output_audio": false,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "omni_video": {
+          "supported_modes": ["text_to_video", "image_reference", "element_reference", "first_last_frame", "video_reference", "video_edit"],
+          "output_resolutions": ["720p", "1080p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10],
+          "output_audio": false,
+          "input_audio": false,
+          "max_videos": 1,
+          "max_audios": 0,
+          "max_images": 7,
+          "max_elements": 7,
+          "max_images_and_elements": 7,
+          "limits_with_video": {"max_images_and_elements": 4},
+          "max_images_for_last_frame": 2,
+          "support_instruction_edit": true,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "originalTypes": ["video_generate", "image_to_video", "omni_video"]
+      }'::jsonb
+    ),
+    (
+      'kling-v3-omni',
+      '["video_generate","image_to_video","omni_video"]'::jsonb,
+      '{
+        "video_generate": {
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "output_audio": true,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "image_to_video": {
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "input_first_frame": true,
+          "input_last_frame": false,
+          "input_first_last_frame": true,
+          "input_reference_generate_single": true,
+          "input_reference_generate_multiple": true,
+          "max_images": 7,
+          "max_images_for_last_frame": 2,
+          "support_video_effect_template": false,
+          "output_audio": true,
+          "input_audio": false,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "omni_video": {
+          "supported_modes": ["text_to_video", "image_reference", "element_reference", "first_last_frame", "video_reference", "video_edit", "multi_shot"],
+          "output_resolutions": ["720p", "1080p", "2160p"],
+          "aspect_ratio_allowed": ["16:9", "1:1", "9:16"],
+          "duration_options": [3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15],
+          "output_audio": true,
+          "input_audio": false,
+          "max_videos": 1,
+          "max_audios": 0,
+          "max_images": 7,
+          "max_elements": 7,
+          "max_images_and_elements": 7,
+          "limits_with_video": {
+            "max_images_and_elements": 4,
+            "duration_options": [3, 4, 5, 6, 7, 8, 9, 10]
+          },
+          "max_images_for_last_frame": 2,
+          "support_instruction_edit": true,
+          "prompt_length_limit": {
+            "max": 2500,
+            "count_mode": "non_ascii_weighted",
+            "label": "可灵口径"
+          }
+        },
+        "originalTypes": ["video_generate", "image_to_video", "omni_video"]
+      }'::jsonb
+    )
+)
+UPDATE base_model_catalog model
+SET model_type = defs.model_type,
+    capabilities = defs.capabilities,
+    metadata = COALESCE(model.metadata, '{}'::jsonb)
+      || jsonb_build_object(
+        'originalTypes', defs.model_type,
+        'rawModel', COALESCE(model.metadata->'rawModel', '{}'::jsonb)
+          || jsonb_build_object(
+            'types', defs.model_type,
+            'capabilities', defs.capabilities
+          )
+      ),
+    default_snapshot = CASE
+      WHEN COALESCE(model.default_snapshot, '{}'::jsonb) = '{}'::jsonb THEN model.default_snapshot
+      ELSE jsonb_set(
+        jsonb_set(
+          jsonb_set(model.default_snapshot, '{modelType}', defs.model_type, true),
+          '{capabilities}',
+          defs.capabilities,
+          true
+        ),
+        '{metadata}',
+        COALESCE(model.default_snapshot->'metadata', '{}'::jsonb)
+          || jsonb_build_object(
+            'originalTypes', defs.model_type,
+            'rawModel', COALESCE(model.default_snapshot->'metadata'->'rawModel', '{}'::jsonb)
+              || jsonb_build_object(
+                'types', defs.model_type,
+                'capabilities', defs.capabilities
+              )
+          ),
+        true
+      )
+    END,
+    updated_at = now()
+FROM keling_omni_models defs
+WHERE model.provider_model_name = defs.provider_model_name
+  AND model.model_type @> '["omni_video"]'::jsonb;
+
+WITH keling_omni_models(provider_model_name, model_type, capabilities) AS (
+  SELECT DISTINCT ON (model.provider_model_name)
+    model.provider_model_name,
+    model.model_type,
+    model.capabilities
+  FROM base_model_catalog model
+  WHERE model.provider_model_name IN ('kling-video-o1', 'kling-v3-omni')
+    AND model.model_type @> '["omni_video"]'::jsonb
+  ORDER BY model.provider_model_name, (model.provider_key = 'keling') DESC, model.created_at ASC
+)
+UPDATE platform_models model
+SET model_type = defs.model_type,
+    capabilities = defs.capabilities,
+    updated_at = now()
+FROM keling_omni_models defs
+WHERE COALESCE(NULLIF(model.provider_model_name, ''), model.model_name) = defs.provider_model_name
+  AND model.model_type @> '["omni_video"]'::jsonb;

apps/api/project.json

@@ -24,7 +24,7 @@
       "outputs": ["{projectRoot}/docs/swagger.json", "{projectRoot}/docs/swagger.yaml"],
       "options": {
         "cwd": "apps/api",
-        "command": "go run github.com/swaggo/swag/cmd/swag@v1.16.4 init --parseInternal -d ./cmd/gateway,./internal/httpapi,./internal/store,./internal/auth,./internal/identity -g main.go -o docs --outputTypes json,yaml"
+        "command": "go run github.com/swaggo/swag/cmd/swag@v1.16.4 init --parseInternal -d ./cmd/gateway,./internal/httpapi,./internal/store,./internal/auth,./internal/identity,./internal/runner -g main.go -o docs --outputTypes json,yaml"
       }
     },
     "test": {

apps/web/src/api.test.ts

@@ -266,8 +266,8 @@ describe('Public Agent resources', () => {
       modules: ['model-runtime'],
       fileName: 'ai-gateway-ops-management-v1.0.2.zip',
       downloadPath: '/api/v1/public/skills/ai-gateway-ops-management/download',
-      apiDocsJsonPath: '/api-docs-json',
-      apiDocsYamlPath: '/api-docs-yaml',
+      apiDocsJsonPath: '/api/v1/openapi.json',
+      apiDocsYamlPath: '/api/v1/openapi.yaml',
     };
     const fetchMock = vi.fn().mockResolvedValue(new Response(JSON.stringify(metadata), {
       status: 200,

apps/web/src/pages/ApiDocsPage.tsx

@@ -35,17 +35,17 @@ interface ApiGuideItem {
 }
 
 export const apiDocs: ApiDocItem[] = [
-  { key: 'chat', group: '文本', kind: 'chat.completions', method: 'POST', path: '/v1/chat/completions', title: 'Chat Completions', lead: 'OpenAI 兼容的对话接口，支持本地 API Key 授权、simulation 测试和非流式/流式响应。' },
-  { key: 'responses', group: '文本', kind: 'responses', method: 'POST', path: '/v1/responses', title: 'Responses', lead: 'OpenAI 兼容的 Responses 接口，原生支持 input、previous_response_id、工具调用和流式输出；不支持原生 Responses 的模型会由网关转换到 Chat Completions。' },
-  { key: 'embeddings', group: '文本', kind: 'embeddings', method: 'POST', path: '/v1/embeddings', title: '文本向量 Embeddings', lead: 'OpenAI 兼容的文本向量接口，可直接用 input 数组或字符串生成 embedding，API Key 需要 embedding 权限。' },
-  { key: 'reranks', group: '文本', kind: 'reranks', method: 'POST', path: '/v1/reranks', title: '文本重排序 Reranks', lead: 'OpenAI 风格的重排序接口，传入 query 和 documents 后返回 relevance_score，API Key 需要 rerank 权限。' },
-  { key: 'imageGeneration', group: '图片', kind: 'images.generations', method: 'POST', path: '/v1/images/generations', title: '创建图片', lead: 'OpenAI 兼容的图片生成接口，支持 prompt、size、quality 和 simulation 测试。' },
-  { key: 'imageEdit', group: '图片', kind: 'images.edits', method: 'POST', path: '/v1/images/edits', title: '编辑图片', lead: 'OpenAI 兼容的图片编辑接口，支持 image、mask、prompt 和 simulation 测试。' },
+  { key: 'chat', group: '文本', kind: 'chat.completions', method: 'POST', path: '/api/v1/chat/completions', title: 'Chat Completions', lead: 'OpenAI 兼容的对话接口，支持本地 API Key 授权、simulation 测试和非流式/流式响应。' },
+  { key: 'responses', group: '文本', kind: 'responses', method: 'POST', path: '/api/v1/responses', title: 'Responses', lead: 'OpenAI 兼容的 Responses 接口，原生支持 input、previous_response_id、工具调用和流式输出；不支持原生 Responses 的模型会由网关转换到 Chat Completions。' },
+  { key: 'embeddings', group: '文本', kind: 'embeddings', method: 'POST', path: '/api/v1/embeddings', title: '文本向量 Embeddings', lead: 'OpenAI 兼容的文本向量接口，可直接用 input 数组或字符串生成 embedding，API Key 需要 embedding 权限。' },
+  { key: 'reranks', group: '文本', kind: 'reranks', method: 'POST', path: '/api/v1/reranks', title: '文本重排序 Reranks', lead: 'OpenAI 风格的重排序接口，传入 query 和 documents 后返回 relevance_score，API Key 需要 rerank 权限。' },
+  { key: 'imageGeneration', group: '图片', kind: 'images.generations', method: 'POST', path: '/api/v1/images/generations', title: '创建图片', lead: 'OpenAI 兼容的图片生成接口，支持 prompt、size、quality 和 simulation 测试。' },
+  { key: 'imageEdit', group: '图片', kind: 'images.edits', method: 'POST', path: '/api/v1/images/edits', title: '编辑图片', lead: 'OpenAI 兼容的图片编辑接口，支持 image、mask、prompt 和 simulation 测试。' },
   { key: 'videoGeneration', group: '视频', kind: 'videos.generations', method: 'POST', path: '/api/v1/videos/generations', title: '生成视频', lead: '视频生成任务接口，支持文生视频、首尾帧、图片/视频/音频参考，以及时长、分辨率、画幅和声音等模型能力参数。' },
   { key: 'asyncMode', group: '异步任务', title: '异步模式', lead: '所有 AI 任务创建接口使用同一种异步开启方式：保留原接口和原请求 Body，只需增加 X-Async: true。' },
   { key: 'taskRetrieve', group: '异步任务', kind: 'tasks.retrieve', method: 'GET', path: '/api/v1/tasks/{taskID}', title: '取回任务', lead: '使用异步提交返回的 taskId 查询任务状态、结果、错误、用量、计费和执行尝试；queued、running、submitting 为进行中状态。' },
   { key: 'pricing', group: '计费', method: 'POST', path: '/api/v1/pricing/estimate', title: '价格预估', lead: '按请求体估算输入输出 token、模型倍率和折扣后的预估费用。' },
-  { key: 'files', group: '文件', method: 'POST', path: '/v1/files/upload', title: '上传文件', lead: '上传在线测试所需的图片、音频或视频资源，后续请求可复用返回的文件 URL。' },
+  { key: 'files', group: '文件', method: 'POST', path: '/api/v1/files/upload', title: '上传文件', lead: '上传在线测试所需的图片、音频或视频资源，后续请求可复用返回的文件 URL。' },
 ];
 
 const guideItems: ApiGuideItem[] = [
@@ -73,8 +73,8 @@ const defaultOpsSkillMetadata: GatewaySkillBundleMetadata = {
   modules: ['model-runtime'],
   fileName: 'ai-gateway-ops-management.zip',
   downloadPath: '/api/v1/public/skills/ai-gateway-ops-management/download',
-  apiDocsJsonPath: '/api-docs-json',
-  apiDocsYamlPath: '/api-docs-yaml',
+  apiDocsJsonPath: '/api/v1/openapi.json',
+  apiDocsYamlPath: '/api/v1/openapi.yaml',
 };
 
 export function ApiDocsPage(props: {
@@ -545,8 +545,8 @@ function GuideDetails(props: { onCreateApiKey: () => void; section: ApiGuideSect
     return (
       <>
         <GuideSection title="1. 确认 Base URL">
-          <p>Base URL 由当前 Gateway 部署环境提供。将接口文档中的路径拼接到 Base URL 后调用，不要重复添加末尾斜杠。</p>
-          <pre>{`export EASYAI_BASE_URL="https://your-gateway.example.com"\ncurl "$EASYAI_BASE_URL/healthz"`}</pre>
+          <p>公开 API Base URL 固定以 <code>/api/v1</code> 结尾。接口卡片展示的是完整路径；SDK 配置 Base URL 后只追加资源路径，不要再次添加 <code>/api/v1</code>。</p>
+          <pre>{`export EASYAI_BASE_URL="https://your-gateway.example.com/api/v1"\ncurl "$EASYAI_BASE_URL/healthz"`}</pre>
         </GuideSection>
         <GuideSection title="2. 创建并使用 API Key">
           <p>登录后在“用户工作台 → API Key”创建 Key，并为它分配实际需要的 chat、embedding、rerank、image 或 video 权限。密钥只在创建或重置时完整展示，请立即安全保存。</p>

deploy/ci/production-migration-base

@@ -1 +1 @@
-2d6c16fec0bec9c0288e5cb142b458af982fff8f
+d95cecd0ebac48bda7ff0a2dcf453bb02b85ca7e

deploy/nginx/ai.51easyai.com-api-v1.inc

@@ -0,0 +1,19 @@
+# Include inside the TLS server block for ai.51easyai.com.
+# The application owns routing below /api/v1; the edge proxy preserves the URI.
+
+location = /api/v1/metrics {
+  return 404;
+}
+
+location ^~ /api/v1/ {
+  proxy_http_version 1.1;
+  proxy_set_header Host $host;
+  proxy_set_header X-Real-IP $remote_addr;
+  proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+  proxy_set_header X-Forwarded-Proto $scheme;
+  proxy_set_header Connection "";
+  proxy_buffering off;
+  proxy_read_timeout 3600s;
+  proxy_send_timeout 3600s;
+  proxy_pass http://127.0.0.1:8088;
+}

docker-compose.yml

@@ -87,7 +87,7 @@ services:
       postgres:
         condition: service_healthy
     healthcheck:
-      test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:8088/readyz | grep -q '\"ok\":true'"]
+      test: ["CMD-SHELL", "wget -qO- http://127.0.0.1:8088/api/v1/readyz | grep -q '\"ok\":true'"]
       interval: 10s
       timeout: 5s
       retries: 20
@@ -117,7 +117,7 @@ services:
       api:
         condition: service_healthy
     healthcheck:
-      test: ["CMD-SHELL", "wget -qO- http://127.0.0.1/gateway-api/healthz | grep -q 'easyai-ai-gateway'"]
+      test: ["CMD-SHELL", "wget -qO- http://127.0.0.1/api/v1/healthz | grep -q 'easyai-ai-gateway'"]
       interval: 10s
       timeout: 5s
       retries: 20

docker/nginx.conf

@@ -33,6 +33,10 @@ server {
     return 404;
   }
 
+  location = /api/v1/metrics {
+    return 404;
+  }
+
   location = /gateway-api/api/v1/auth/login {
     proxy_http_version 1.1;
     proxy_set_header Host $host;
@@ -47,6 +51,35 @@ server {
     proxy_pass http://api:8088/api/v1/auth/login;
   }
 
+  location = /api/v1/auth/login {
+    proxy_http_version 1.1;
+    proxy_set_header Host $host;
+    proxy_set_header X-Real-IP $remote_addr;
+    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+    proxy_set_header X-Forwarded-Proto $scheme;
+    proxy_set_header Connection "";
+    proxy_connect_timeout 3s;
+    proxy_read_timeout 15s;
+    proxy_send_timeout 15s;
+    proxy_redirect off;
+    proxy_pass http://api:8088;
+  }
+
+  # Canonical public API. Keep the request URI so /api/v1 reaches the
+  # versioned handlers without another prefix rewrite.
+  location ^~ /api/v1/ {
+    proxy_http_version 1.1;
+    proxy_set_header Host $host;
+    proxy_set_header X-Real-IP $remote_addr;
+    proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
+    proxy_set_header X-Forwarded-Proto $scheme;
+    proxy_set_header Connection "";
+    proxy_buffering off;
+    proxy_read_timeout 3600s;
+    proxy_send_timeout 3600s;
+    proxy_pass http://api:8088;
+  }
+
   location /gateway-api/ {
     proxy_http_version 1.1;
     proxy_set_header Host $host;

docs/design.md

@@ -91,7 +91,7 @@ flowchart LR
   QUEUE --> PG
   QUEUE --> CALLBACK -->|POST task progress callback to server-main| WSGW
   QUEUE -->|settlement event| BILL
-  API -->|POST /v1/files/upload| FILES
+  API -->|POST /api/v1/files/upload| FILES
 ```
 
 ## 4. Monorepo 方案
@@ -1785,9 +1785,9 @@ SimulationClient 根据 `simulation_profile` 生成确定性行为：
 - `/chat/completions`
 - `/images/generations`
 - `/video/generations`
-- `/v1/chat/completions`
-- `/v1/images/generations`
-- `/v1/video/generations`
+- `/api/v1/chat/completions`
+- `/api/v1/images/generations`
+- `/api/v1/video/generations`
 
 内部 `OpenaiService` 变成薄门面：
 

docs/kling-compatible-api.md

@@ -5,11 +5,11 @@
 生产环境统一配置：
 
 ```text
-baseURL = https://ai.51easyai.com/gateway-api/kling
+baseURL = https://ai.51easyai.com/api/v1/kling
 Authorization = Bearer <EasyAI Gateway API Key>
 ```
 
-本地环境使用 `baseURL = http://localhost:8088/kling`。
+本地环境使用 `baseURL = http://localhost:8088/api/v1/kling`。旧 `/gateway-api/kling` 和 `/kling` 路径仅作为兼容别名保留。
 
 ## V1（AK/SK 旧版协议兼容）
 

docs/kling-omni-compatible.md

@@ -2,6 +2,12 @@
 
 EasyAI AI Gateway 提供 Kling 旧版 Omni 协议兼容接口。调用方继续使用 Gateway API Key，任务仍经过网关候选选择、异步队列、审计和计费；响应中的 `task_id` 是网关任务 UUID，不是上游任务 ID。
 
+```bash
+export GATEWAY_ORIGIN="https://ai.51easyai.com"
+export GATEWAY_PUBLIC_API_BASE="$GATEWAY_ORIGIN/api/v1"
+export GATEWAY_API_KEY="<EasyAI Gateway API Key>"
+```
+
 ## 模型与参数映射
 
 | 请求 `model_name` | 网关模型别名 | TranStreams 原生 `model_name` | 时长范围 |
@@ -21,10 +27,10 @@ EasyAI AI Gateway 提供 Kling 旧版 Omni 协议兼容接口。调用方继续
 
 ## 创建任务
 
-`POST /v1/videos/omni-video` 固定异步受理，不需要 `X-Async`，成功返回 HTTP 200。
+`POST /api/v1/videos/omni-video` 固定异步受理，不需要 `X-Async`，成功返回 HTTP 200。
 
 ```bash
-curl -sS -X POST "$GATEWAY_BASE_URL/v1/videos/omni-video" \
+curl -sS -X POST "$GATEWAY_ORIGIN/api/v1/videos/omni-video" \
   -H "Authorization: Bearer $GATEWAY_API_KEY" \
   -H "Content-Type: application/json" \
   -d '{
@@ -61,7 +67,7 @@ curl -sS -X POST "$GATEWAY_BASE_URL/v1/videos/omni-video" \
 ```bash
 curl -sS \
   -H "Authorization: Bearer $GATEWAY_API_KEY" \
-  "$GATEWAY_BASE_URL/v1/videos/omni-video/$TASK_ID"
+  "$GATEWAY_ORIGIN/api/v1/videos/omni-video/$TASK_ID"
 ```
 
 `task_status` 为 `submitted`、`processing`、`succeed` 或 `failed`。成功时结果位于 `data.task_result.videos`：
@@ -93,7 +99,7 @@ curl -sS \
 标准接口仍为 `POST /api/v1/videos/generations`。异步调用需要 `X-Async: true`，再通过 `GET /api/v1/tasks/{taskId}` 轮询。
 
 ```bash
-curl -sS -X POST "$GATEWAY_BASE_URL/api/v1/videos/generations" \
+curl -sS -X POST "$GATEWAY_PUBLIC_API_BASE/videos/generations" \
   -H "Authorization: Bearer $GATEWAY_API_KEY" \
   -H "Content-Type: application/json" \
   -H "X-Async: true" \
@@ -112,7 +118,7 @@ curl -sS -X POST "$GATEWAY_BASE_URL/api/v1/videos/generations" \
 ```bash
 curl -sS \
   -H "Authorization: Bearer $GATEWAY_API_KEY" \
-  "$GATEWAY_BASE_URL/api/v1/tasks/$TASK_ID"
+  "$GATEWAY_PUBLIC_API_BASE/tasks/$TASK_ID"
 ```
 
 ## 错误格式
@@ -129,4 +135,4 @@ curl -sS \
 
 业务码分类：`1001/1002` 为鉴权错误，`1101/1103` 为余额或权限错误，`1201/1203` 为参数或资源错误，`1302/1303` 为限流错误，`5000/5001` 为网关或上游服务错误。HTTP 状态码仍反映错误类型。
 
-OpenAPI 文档由服务的 `/openapi.json` 和 `/openapi.yaml` 提供。
+OpenAPI 文档由服务的 `/api/v1/openapi.json` 和 `/api/v1/openapi.yaml` 提供。

docs/public-api-v1.md

@@ -0,0 +1,137 @@
+# EasyAI Gateway 公开 API V1 清单
+
+生产公开 API 的统一 Base URL：
+
+```text
+https://ai.51easyai.com/api/v1
+```
+
+下表路径均以 `/api/v1` 开头。调用方使用 `Authorization: Bearer <API Key>`；标记为“公开”的接口不要求用户登录，OIDC 与 SSF 接口按各自协议鉴权。
+
+旧 `/gateway-api`、`/v1`、无版本路径、`/kling`、`/v1beta`、`/upload` 和 `/api/v3` 入口仅作为兼容别名保留，不再用于新接入文档。
+
+## 运行状态与接口发现
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| GET | `/api/v1/healthz` | 存活检查 |
+| GET | `/api/v1/readyz` | 就绪检查 |
+| GET | `/api/v1/openapi.json` | OpenAPI JSON |
+| GET | `/api/v1/openapi.yaml` | OpenAPI YAML |
+| GET | `/api/v1/public/identity` | 公开身份配置 |
+| GET | `/api/v1/public/client-customization` | 公开客户端配置 |
+| GET | `/api/v1/public/catalog/providers` | 公开供应商目录 |
+| GET | `/api/v1/public/catalog/base-models` | 公开基础模型目录 |
+| GET | `/api/v1/public/skills/ai-gateway-ops-management/metadata` | 运维 Skill 元数据 |
+| GET | `/api/v1/public/skills/ai-gateway-ops-management/download` | 下载运维 Skill |
+
+## 账号、授权与 API Key
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/auth/register` | 注册本地账号 |
+| POST | `/api/v1/auth/login` | 本地账号登录 |
+| GET | `/api/v1/auth/oidc/login` | 发起 OIDC 登录 |
+| GET | `/api/v1/auth/oidc/callback` | OIDC 回调 |
+| POST | `/api/v1/auth/oidc/logout` | OIDC 登出 |
+| DELETE | `/api/v1/auth/oidc/session` | 删除浏览器会话 |
+| GET | `/api/v1/me` | 当前用户 |
+| GET, POST | `/api/v1/api-keys` | 查询、创建 API Key |
+| GET | `/api/v1/api-keys/access-rules` | 查询 Key 访问规则 |
+| POST | `/api/v1/api-keys/access-rules/batch` | 批量设置 Key 访问规则 |
+| GET | `/api/v1/api-keys/assignable-models` | 查询可分配模型 |
+| PATCH | `/api/v1/api-keys/{apiKeyID}/scopes` | 更新 Key 权限范围 |
+| PATCH | `/api/v1/api-keys/{apiKeyID}/disable` | 禁用 Key |
+| DELETE | `/api/v1/api-keys/{apiKeyID}` | 删除 Key |
+
+## 模型、平台与计费查询
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| GET | `/api/v1/model-catalog` | 模型能力目录 |
+| GET | `/api/v1/platforms` | 当前用户可用平台 |
+| GET | `/api/v1/models` | 当前用户可用模型 |
+| GET | `/api/v1/playground/models` | Playground 可用模型 |
+| POST | `/api/v1/pricing/estimate` | 请求价格预估 |
+
+## 通用与 OpenAI 兼容生成接口
+
+这些接口默认同步返回兼容响应；需要异步执行时增加 `X-Async: true`，并使用任务接口取回结果。
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/chat/completions` | Chat Completions，支持 SSE |
+| POST | `/api/v1/responses` | Responses，支持 SSE |
+| POST | `/api/v1/embeddings` | 文本向量 |
+| POST | `/api/v1/reranks` | 文本重排序 |
+| POST | `/api/v1/images/generations` | 文生图 |
+| POST | `/api/v1/images/edits` | 图片编辑 |
+| POST | `/api/v1/videos/generations` | 文生视频、图生视频及多模态视频 |
+| POST | `/api/v1/song/generations` | 歌曲生成 |
+| POST | `/api/v1/music/generations` | 音乐生成 |
+| POST | `/api/v1/speech/generations` | 语音生成 |
+| POST | `/api/v1/voice_clone` | 声音克隆 |
+| GET | `/api/v1/voice_clone/voices` | 查询克隆声音 |
+| DELETE | `/api/v1/voice_clone/voices/{voiceID}` | 删除克隆声音 |
+| POST | `/api/v1/files/upload` | 上传生成任务输入文件 |
+
+## 异步任务
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| GET | `/api/v1/tasks` | 查询任务列表 |
+| GET | `/api/v1/tasks/{taskID}` | 查询任务详情和结果 |
+| POST | `/api/v1/tasks/{taskID}/cancel` | 取消任务 |
+| GET | `/api/v1/tasks/{taskID}/events` | 查询任务事件 |
+| GET | `/api/v1/tasks/{taskID}/param-preprocessing` | 查询参数预处理记录 |
+
+## Gemini 兼容接口
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/models/{model}:generateContent` | Gemini generateContent |
+| POST | `/api/v1/gemini/upload/{version}/files` | Gemini Files 启动或直接上传，`version` 为 `v1` 或 `v1beta` |
+| POST | `/api/v1/gemini/upload/{version}/files/{uploadID}` | 完成 Gemini 分段上传 |
+
+## 可灵兼容接口
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/videos/omni-video` | 可灵官方 V1 Omni 创建任务 |
+| GET | `/api/v1/videos/omni-video/{taskID}` | 可灵官方 V1 Omni 查询任务 |
+| POST | `/api/v1/kling/v1/videos/omni-video` | 网关可灵 V1 创建任务 |
+| GET | `/api/v1/kling/v1/videos/omni-video` | 网关可灵 V1 任务列表 |
+| GET | `/api/v1/kling/v1/videos/omni-video/{taskID}` | 网关可灵 V1 查询任务 |
+| POST | `/api/v1/kling/v2/omni-video/{model}` | 网关可灵 API 2.0 创建任务 |
+| GET | `/api/v1/kling/v2/tasks` | 网关可灵 API 2.0 查询任务 |
+| POST | `/api/v1/kling/v2/tasks` | 网关可灵 API 2.0 任务列表 |
+
+可灵客户端可使用 `https://ai.51easyai.com/api/v1/kling` 作为 Base URL，然后继续请求 `/v1/...` 或 `/v2/...`。
+
+## 火山兼容与真人资产
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/contents/generations/tasks` | 创建火山内容生成任务 |
+| GET | `/api/v1/contents/generations/tasks` | 查询火山内容生成任务列表 |
+| GET | `/api/v1/contents/generations/tasks/{taskID}` | 查询火山内容生成任务 |
+| DELETE | `/api/v1/contents/generations/tasks/{taskID}` | 删除或取消火山内容生成任务 |
+| POST | `/api/v1/video/generations` | server-main 兼容视频创建接口 |
+| GET | `/api/v1/ai/result/{taskID}` | server-main 兼容结果查询接口 |
+| GET | `/api/v1/resource/material/seedance-portrait-assets/capability` | 真人资产能力 |
+| GET | `/api/v1/resource/material/user/materials` | 查询用户真人资产 |
+| POST | `/api/v1/resource/material` | 上传真人资产 |
+| POST | `/api/v1/resource/material/seedance-portrait-assets/sync` | 同步真人资产到平台 |
+
+## 安全集成
+
+| 方法 | 路径 | 说明 |
+|---|---|---|
+| POST | `/api/v1/security-events/ssf` | RFC 8935 Security Event 接收端点 |
+
+## 不属于公开 API 的路径
+
+- `/api/admin/...`：管理后台接口。
+- `/api/workspace/...`、`/api/playground/...`：Web/BFF 内部接口。
+- `/metrics`：仅监控网络可访问。
+- `/static/...`：生成结果和上传文件的资源 URL，不是 API Base URL。

docs/runbooks/production-ci-cd.md

@@ -126,9 +126,11 @@ dispatcher 以完整 Git SHA 发布 Registry Tag，并把 Registry 返回的 dig
 
 ## 发布后验证
 
+宿主 Nginx 的 `ai.51easyai.com` TLS server 必须包含仓库中的 `deploy/nginx/ai.51easyai.com-api-v1.inc` 等价规则，保留完整 URI 转发到 `127.0.0.1:8088`。修改前先备份现有配置，执行 `nginx -t` 成功后才能 reload。
+
 ```bash
-curl -fsS https://ai.51easyai.com/gateway-api/healthz
-curl -fsS https://ai.51easyai.com/gateway-api/readyz
+curl -fsS https://ai.51easyai.com/api/v1/healthz
+curl -fsS https://ai.51easyai.com/api/v1/readyz
 ssh root@110.42.51.33 'cd /root/easyai-ai-gateway-deploy && ./gateway-ops.sh ps'
 ```
 

docs/seedance-portrait-assets.md

@@ -33,10 +33,12 @@
 
 ## 火山任务兼容路由
 
-- `POST /api/v3/contents/generations/tasks`
-- `GET /api/v3/contents/generations/tasks`
-- `GET /api/v3/contents/generations/tasks/{taskID}`
-- `DELETE /api/v3/contents/generations/tasks/{taskID}`
+- `POST /api/v1/contents/generations/tasks`
+- `GET /api/v1/contents/generations/tasks`
+- `GET /api/v1/contents/generations/tasks/{taskID}`
+- `DELETE /api/v1/contents/generations/tasks/{taskID}`
+
+旧 `/api/v3/contents/generations/tasks` 路径仅作为火山客户端兼容别名保留。
 
 列表接口兼容火山的 `page_num`、`page_size`、`filter.status`、`filter.task_ids`（可重复）和 `filter.model`，并返回官方 `items`、`total` 字段；`data`、`page` 是保留的网关附加字段。
 

docs/server-main-integration.md

@@ -36,7 +36,7 @@ Content-Type: application/json
 ### 1.2 文件上传
 
 ```http
-POST /v1/files/upload
+POST /api/v1/files/upload
 Authorization: Bearer ${USER_JWT_OR_SK}
 Content-Type: multipart/form-data
 

docs/test/loopback-test-checklist.md

@@ -53,7 +53,7 @@
 
 | ID | 任务 | 接口 / 方式 | 成功判定 | 状态 | 结果记录 |
 | --- | --- | --- | --- | --- | --- |
-| SETUP-01 | 确认服务可用 | `GET /healthz`、`GET /readyz` | `healthz.ok=true`，`readyz.ok=true` | 未执行 | 待填写 |
+| SETUP-01 | 确认服务可用 | `GET /api/v1/healthz`、`GET /api/v1/readyz` | `healthz.ok=true`，`readyz.ok=true` | 未执行 | 待填写 |
 | SETUP-02 | 准备管理员权限 | 本地注册 / 登录，必要时将测试用户提升为 `admin` 或 `manager` | `GET /api/v1/me` 返回 `role` 具备 `manager` 权限 | 未执行 | 待填写 |
 | SETUP-03 | 记录用户提供的真实平台、模型和 KEY | `GET /api/v1/platforms`、`GET /api/v1/models` | Chat 模型、`doubao-4.5图像编辑`、`豆包Seedance-1.5-pro` 均已启用，并能被管理员看到 | 未执行 | 待填写 |
 | SETUP-04 | 创建内部测试用户组 | `POST /api/v1/user-groups` | 创建 `loopback-allow-group`、`loopback-deny-group`、`loopback-limit-group` | 未执行 | 待填写 |
@@ -91,7 +91,7 @@
 | ID | 能力 | 请求 | 成功判定 | 状态 | 结果记录 |
 | --- | --- | --- | --- | --- | --- |
 | TASK-CHAT-01 | Chat 成功 | `POST /api/v1/chat/completions`，真实 Chat 模型 | `task.status=succeeded`，`result.object=chat.completion`，`choices[0].message.content` 非空 | 未执行 | taskId、requestId、content 摘要、charge 待填写 |
-| TASK-CHAT-02 | Chat 兼容路由成功 | `POST /v1/chat/completions`，真实 Chat 模型 | HTTP 200，返回 `object=chat.completion`，内容非空 | 未执行 | requestId、content 摘要待填写 |
+| TASK-CHAT-02 | Chat 同步兼容响应成功 | `POST /api/v1/chat/completions`，真实 Chat 模型 | HTTP 200，返回 `object=chat.completion`，内容非空 | 未执行 | requestId、content 摘要待填写 |
 | TASK-IMAGE-01 | 文生图成功 | `POST /api/v1/images/generations`，模型 `doubao-4.5图像编辑` 或用户补充的文生图模型 | `task.status=succeeded`，`result.id` 非空，`data[0].url` 或文件 URL 可访问 | 未执行 | taskId、image URL、charge 待填写 |
 | TASK-IMAGE-02 | 图生图成功 | `POST /api/v1/images/edits`，模型 `doubao-4.5图像编辑`，传入测试源图和 mask | `task.status=succeeded`，`result.id` 非空，`data[0].url` 或文件 URL 可访问 | 未执行 | taskId、source URL、mask URL、image URL、charge 待填写 |
 | TASK-VIDEO-01 | 文生视频成功 | `POST /api/v1/videos/generations`，模型 `豆包Seedance-1.5-pro`，仅传 prompt | `task.status=succeeded`，返回可下载或可播放的视频结果，任务事件完整 | 未执行 | taskId、video URL、duration、charge 待填写 |

pnpm-lock.yaml

@@ -6,17 +6,18 @@ settings:
 
 overrides:
   '@babel/core@<=7.29.0': 7.29.6
-  '@nx/js>picomatch': 4.0.5
-  '@nx/vite>picomatch': 4.0.5
-  '@nx/workspace>picomatch': 4.0.5
   axios@<1.18.0: 1.18.0
   brace-expansion@>=2.0.0 <2.1.2: 2.1.2
   dompurify@<=3.4.10: 3.4.11
   esbuild@>=0.27.3 <0.28.1: 0.28.1
+  fast-uri@>=3.0.0 <3.1.4: 3.1.4
   form-data@>=4.0.0 <4.0.6: 4.0.6
   js-yaml@<4.3.0: 4.3.0
   mermaid@>=11.0.0-alpha.1 <=11.14.0: 11.15.0
   minimatch@>=9.0.0 <9.0.7: 9.0.7
+  '@nx/js>picomatch': 4.0.5
+  '@nx/vite>picomatch': 4.0.5
+  '@nx/workspace>picomatch': 4.0.5
   tmp@<0.2.7: 0.2.7
 
 importers:
@@ -3034,8 +3035,8 @@ packages:
   fast-deep-equal@3.1.3:
     resolution: {integrity: sha512-f3qQ9oQy9j2AhBe/H9VC91wLmKBCCU/gDOnKNAYG5hswO7BLKj09Hc5HYNz9cGI++xlpDCIgDaitVs03ATR84Q==}
 
-  fast-uri@3.1.2:
-    resolution: {integrity: sha512-rVjf7ArG3LTk+FS6Yw81V1DLuZl1bRbNrev6Tmd/9RaroeeRRJhAt7jg/6YFxbvAQXUCavSoZhPPj6oOx+5KjQ==}
+  fast-uri@3.1.4:
+    resolution: {integrity: sha512-8JnbkQ4juDyvYs4mgFGQqg4yCYtFDtUtmp2QIQq11ZZe5CFQ5wcqm1rqDgAh/QdMySuBnPzMUiJUNZG5N/AiQw==}
 
   fdir@6.5.0:
     resolution: {integrity: sha512-tIbYtZbucOs0BRGqPJkshJUYdL+SDH7dVM8gjy+ERp3WAUjLEFJE+02kanyHtwjWOnwrKYBiwAmM0p4kLJAnXg==}
@@ -7181,7 +7182,7 @@ snapshots:
   ajv@8.20.0:
     dependencies:
       fast-deep-equal: 3.1.3
-      fast-uri: 3.1.2
+      fast-uri: 3.1.4
       json-schema-traverse: 1.0.0
       require-from-string: 2.0.2
 
@@ -7850,7 +7851,7 @@ snapshots:
 
   fast-deep-equal@3.1.3: {}
 
-  fast-uri@3.1.2: {}
+  fast-uri@3.1.4: {}
 
   fdir@6.5.0(picomatch@4.0.4):
     optionalDependencies:

pnpm-workspace.yaml

@@ -12,6 +12,7 @@ overrides:
   'brace-expansion@>=2.0.0 <2.1.2': 2.1.2
   'dompurify@<=3.4.10': 3.4.11
   'esbuild@>=0.27.3 <0.28.1': 0.28.1
+  'fast-uri@>=3.0.0 <3.1.4': 3.1.4
   'form-data@>=4.0.0 <4.0.6': 4.0.6
   'js-yaml@<4.3.0': 4.3.0
   'mermaid@>=11.0.0-alpha.1 <=11.14.0': 11.15.0

scripts/deploy-compose.sh

@@ -248,14 +248,14 @@ deploy() {
   api_port="$(published_port api 8088 "${AI_GATEWAY_API_PORT:-8088}")"
   web_port="$(published_port web 80 "${AI_GATEWAY_WEB_PORT:-5178}")"
 
-  wait_for_http "api health" "http://127.0.0.1:${api_port}/healthz" "easyai-ai-gateway"
-  wait_for_http "api readiness" "http://127.0.0.1:${api_port}/readyz" '"ok":true'
-  wait_for_http "web reverse proxy" "http://127.0.0.1:${web_port}/gateway-api/healthz" "easyai-ai-gateway"
+  wait_for_http "api health" "http://127.0.0.1:${api_port}/api/v1/healthz" "easyai-ai-gateway"
+  wait_for_http "api readiness" "http://127.0.0.1:${api_port}/api/v1/readyz" '"ok":true'
+  wait_for_http "web reverse proxy" "http://127.0.0.1:${web_port}/api/v1/healthz" "easyai-ai-gateway"
   wait_for_http "web app" "http://127.0.0.1:${web_port}/" "EasyAI AI Gateway"
 
   echo "[ai-gateway] deployment succeeded"
   echo "[ai-gateway] Web: http://127.0.0.1:${web_port}"
-  echo "[ai-gateway] API: http://127.0.0.1:${api_port}/healthz"
+  echo "[ai-gateway] API: http://127.0.0.1:${api_port}/api/v1/healthz"
 }
 
 case "$ACTION" in