fix(acceptance): 校准 Seedance 图片转换验收

按 Volces 官方输入边界补齐 Seedance 2.0 候选能力，将错误的合法 4K 转换样本替换为真实越界图片，并让协议模拟器校验物化后的 Base64 data URL。\n\n验证：Go 全量测试、迁移安全检查、gofmt。
2026-07-31 20:07:28 +08:00
parent 2c7ed905a9
commit bfdabd3853
6 changed files with 233 additions and 31 deletions
@@ -848,7 +848,7 @@ func (o options) emulatorFixtureURLs() []string {
 		out = append(out, fmt.Sprintf("%s/fixtures/image-%02d.webp", o.emulatorURL, index))
 	}
 	for index := 12; index < 16; index++ {
-		out = append(out, fmt.Sprintf("%s/fixtures/image-%02d-4k.jpg", o.emulatorURL, index))
+		out = append(out, fmt.Sprintf("%s/fixtures/image-%02d-oversized.jpg", o.emulatorURL, index))
 	}
 	return out
 }
@@ -52,6 +52,28 @@ func TestSeedanceInputImageConstraintMigrationKeepsCatalogSnapshotsInSync(t *tes
 	}
 }

+func TestVolcesSeedanceInputImageConstraintMigrationUsesDocumentedBounds(t *testing.T) {
+	payload, err := os.ReadFile("../../migrations/0098_volces_seedance20_input_image_constraints.sql")
+	if err != nil {
+		t.Fatal(err)
+	}
+	content := string(payload)
+	for _, required := range []string{
+		"volces:doubao-seedance-2-0-260128",
+		"volces:doubao-seedance-2-0-fast-260128",
+		"volces:doubao-seedance-2-0-mini-260615",
+		`"long_edge":300`,
+		`"long_edge":6000`,
+		`'[0.4,2.5]'::jsonb`,
+		"'{metadata,rawModel,capabilities}'",
+		"UPDATE platform_models",
+	} {
+		if !strings.Contains(content, required) {
+			t.Fatalf("Volces Seedance input image migration is missing %q", required)
+		}
+	}
+}
+
 func TestSecurityEventSchemaMigrationsDefineCurrentLifecycle(t *testing.T) {
 	streamPayload, err := os.ReadFile("../../migrations/0063_oidc_security_events.sql")
 	if err != nil {
@@ -28,7 +28,10 @@ import (
 	_ "golang.org/x/image/webp"
 )

-const maxProtocolBodyBytes = 128 << 20
+const (
+	maxProtocolBodyBytes   = 128 << 20
+	maxImageReferenceBytes = 32 << 20
+)

 type Config struct {
 	Now        func() time.Time
@@ -434,37 +437,30 @@ func (s *Server) validateVideoRequest(ctx context.Context, body map[string]any)
 	}
 	if forceConversion {
 		for _, ref := range refs {
-			if ref.Width >= 3840 || ref.Height >= 2160 {
-				return nil, false, false, errors.New("forced 4K image did not pass through automatic normalization")
+			if !seedanceImageWithinOfficialInputRange(ref.Width, ref.Height) {
+				return nil, false, false, errors.New("forced oversized image did not pass through automatic normalization")
 			}
 		}
 	}
 	return refs, longRun, forceConversion, nil
 }

+func seedanceImageWithinOfficialInputRange(width int, height int) bool {
+	if width < 300 || width > 6000 || height < 300 || height > 6000 {
+		return false
+	}
+	ratio := float64(width) / float64(height)
+	return ratio >= 0.4 && ratio <= 2.5
+}
+
 func (s *Server) fetchImageReference(ctx context.Context, role string, rawURL string) (imageReference, error) {
 	if rawURL == "" {
 		return imageReference{}, errors.New("image reference URL is required")
 	}
-	request, err := http.NewRequestWithContext(ctx, http.MethodGet, rawURL, nil)
+	payload, err := s.readImageReference(ctx, rawURL)
 	if err != nil {
 		return imageReference{}, err
 	}
-	response, err := s.httpClient.Do(request)
-	if err != nil {
-		return imageReference{}, fmt.Errorf("fetch image reference: %w", err)
-	}
-	defer response.Body.Close()
-	if response.StatusCode < 200 || response.StatusCode >= 300 {
-		return imageReference{}, fmt.Errorf("fetch image reference: status %d", response.StatusCode)
-	}
-	payload, err := io.ReadAll(io.LimitReader(response.Body, 32<<20))
-	if err != nil {
-		return imageReference{}, err
-	}
-	if len(payload) == 0 {
-		return imageReference{}, errors.New("image reference is empty")
-	}
 	config, format, err := image.DecodeConfig(bytes.NewReader(payload))
 	if err != nil {
 		return imageReference{}, fmt.Errorf("decode image reference: %w", err)
@@ -476,6 +472,67 @@ func (s *Server) fetchImageReference(ctx context.Context, role string, rawURL st
 	}, nil
 }

+func (s *Server) readImageReference(ctx context.Context, rawURL string) ([]byte, error) {
+	if strings.HasPrefix(strings.ToLower(rawURL), "data:") {
+		return decodeImageDataURL(rawURL)
+	}
+	request, err := http.NewRequestWithContext(ctx, http.MethodGet, rawURL, nil)
+	if err != nil {
+		return nil, err
+	}
+	response, err := s.httpClient.Do(request)
+	if err != nil {
+		return nil, fmt.Errorf("fetch image reference: %w", err)
+	}
+	defer response.Body.Close()
+	if response.StatusCode < 200 || response.StatusCode >= 300 {
+		return nil, fmt.Errorf("fetch image reference: status %d", response.StatusCode)
+	}
+	payload, err := io.ReadAll(io.LimitReader(response.Body, maxImageReferenceBytes+1))
+	if err != nil {
+		return nil, err
+	}
+	if len(payload) == 0 {
+		return nil, errors.New("image reference is empty")
+	}
+	if len(payload) > maxImageReferenceBytes {
+		return nil, errors.New("image reference exceeds 32 MiB")
+	}
+	return payload, nil
+}
+
+func decodeImageDataURL(raw string) ([]byte, error) {
+	header, encoded, ok := strings.Cut(raw, ",")
+	if !ok || len(header) <= len("data:") || encoded == "" {
+		return nil, errors.New("image data URL is malformed")
+	}
+	metadata := strings.Split(header[len("data:"):], ";")
+	if !strings.HasPrefix(strings.ToLower(strings.TrimSpace(metadata[0])), "image/") {
+		return nil, errors.New("image data URL must declare an image media type")
+	}
+	base64Encoded := false
+	for _, item := range metadata[1:] {
+		if strings.EqualFold(strings.TrimSpace(item), "base64") {
+			base64Encoded = true
+			break
+		}
+	}
+	if !base64Encoded {
+		return nil, errors.New("image data URL must use base64 encoding")
+	}
+	if base64.StdEncoding.DecodedLen(len(encoded)) > maxImageReferenceBytes {
+		return nil, errors.New("image reference exceeds 32 MiB")
+	}
+	payload, err := base64.StdEncoding.DecodeString(encoded)
+	if err != nil {
+		return nil, errors.New("image data URL contains invalid base64")
+	}
+	if len(payload) == 0 {
+		return nil, errors.New("image reference is empty")
+	}
+	return payload, nil
+}
+
 func validateGeminiImageRequest(body map[string]any) (int, error) {
 	generationConfig, _ := body["generationConfig"].(map[string]any)
 	modalities, _ := generationConfig["responseModalities"].([]any)
@@ -662,10 +719,10 @@ func buildFixtures() map[string]fixture {
 		fixtures[fmt.Sprintf("image-%02d.webp", index+8)] = fixture{ContentType: "image/webp", Payload: payload}
 	}
 	for index := 0; index < 4; index++ {
-		imageValue := patternedImage(4096, 2160, index+21)
+		imageValue := patternedImage(6144, 2160, index+21)
 		var encoded bytes.Buffer
 		_ = jpeg.Encode(&encoded, imageValue, &jpeg.Options{Quality: 90})
-		fixtures[fmt.Sprintf("image-%02d-4k.jpg", index+12)] = fixture{ContentType: "image/jpeg", Payload: encoded.Bytes()}
+		fixtures[fmt.Sprintf("image-%02d-oversized.jpg", index+12)] = fixture{ContentType: "image/jpeg", Payload: encoded.Bytes()}
 	}
 	return fixtures
 }
@@ -58,14 +58,14 @@ func TestGeminiAndVolcesProtocolEmulation(t *testing.T) {
 	}
 	_ = response.Body.Close()

-	fixtureResponse, err := http.Get(httpServer.URL + "/fixtures/image-12-4k.jpg")
+	fixtureResponse, err := http.Get(httpServer.URL + "/fixtures/image-12-oversized.jpg")
 	if err != nil {
-		t.Fatalf("get 4K fixture: %v", err)
+		t.Fatalf("get oversized fixture: %v", err)
 	}
 	config, format, err := image.DecodeConfig(fixtureResponse.Body)
 	_ = fixtureResponse.Body.Close()
-	if err != nil || format != "jpeg" || config.Width != 4096 || config.Height != 2160 {
-		t.Fatalf("4K fixture format=%s config=%+v err=%v", format, config, err)
+	if err != nil || format != "jpeg" || config.Width != 6144 || config.Height != 2160 {
+		t.Fatalf("oversized fixture format=%s config=%+v err=%v", format, config, err)
 	}

 	content := []any{map[string]any{"type": "text", "text": "video"}}
@@ -125,6 +125,29 @@ func TestGeminiAndVolcesProtocolEmulation(t *testing.T) {
 	}
 }

+func TestForcedConversionRejectsImagesOutsideOfficialSeedanceRange(t *testing.T) {
+	server := New(Config{Wait: func(context.Context, time.Duration) error { return nil }})
+	httpServer := httptest.NewServer(server.Handler())
+	defer httpServer.Close()
+
+	content := []any{map[string]any{"type": "text", "text": "acceptance-force-conversion"}}
+	for _, name := range []string{"image-12-oversized.jpg", "image-00.png", "image-04.jpg"} {
+		content = append(content, map[string]any{
+			"type": "image_url", "role": "reference_image",
+			"image_url": map[string]any{"url": httpServer.URL + "/fixtures/" + name},
+		})
+	}
+	body, _ := json.Marshal(map[string]any{"model": "seedance-test", "content": content})
+	response, err := postIdempotent(httpServer.URL+"/contents/generations/tasks", body, "oversized")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer response.Body.Close()
+	if response.StatusCode != http.StatusBadRequest {
+		t.Fatalf("status=%d, want %d", response.StatusCode, http.StatusBadRequest)
+	}
+}
+
 func TestPaddedPNGIsValidAndExact(t *testing.T) {
 	for _, size := range []int{256 << 10, 4 << 20, 8 << 20} {
 		payload := paddedPNG(size)
@@ -148,6 +171,7 @@ func TestVolcesProtocolAcceptsThreeSixAndNineReferenceImages(t *testing.T) {
 	server := New(Config{Wait: func(context.Context, time.Duration) error { return nil }})
 	httpServer := httptest.NewServer(server.Handler())
 	defer httpServer.Close()
+	dataImage := "data:image/png;base64," + base64.StdEncoding.EncodeToString(paddedPNG(0))

 	for _, imageCount := range []int{3, 6, 9} {
 		content := []any{map[string]any{"type": "text", "text": "multi-reference video"}}
@@ -159,11 +183,13 @@ func TestVolcesProtocolAcceptsThreeSixAndNineReferenceImages(t *testing.T) {
 			if imageCount > 3 && index == imageCount-1 {
 				role = "last_frame"
 			}
+			imageURL := fmt.Sprintf("%s/fixtures/image-%02d.png", httpServer.URL, index%4)
+			if index%2 == 0 {
+				imageURL = dataImage
+			}
 			content = append(content, map[string]any{
 				"type": "image_url", "role": role,
-				"image_url": map[string]any{
-					"url": fmt.Sprintf("%s/fixtures/image-%02d.png", httpServer.URL, index%4),
-				},
+				"image_url": map[string]any{"url": imageURL},
 			})
 		}
 		body, _ := json.Marshal(map[string]any{"model": "seedance-test", "content": content, "seed": imageCount})
@@ -0,0 +1,95 @@
+-- Volces documents Seedance 2.0 image inputs as 300..6000 px on each edge
+-- with a width/height ratio of 0.4..2.5. The runtime consumes these
+-- capability fields to normalize only genuinely out-of-range references.
+WITH input_constraints AS (
+  SELECT jsonb_build_object(
+    'input_image_resolution_range',
+    '{"min":{"long_edge":300,"short_edge":300},"max":{"long_edge":6000,"short_edge":6000}}'::jsonb,
+    'input_image_aspect_ratio_range',
+    '[0.4,2.5]'::jsonb
+  ) AS value
+),
+target_base_models AS (
+  SELECT
+    base_model.id,
+    COALESCE(base_model.capabilities, '{}'::jsonb) || jsonb_build_object(
+      'image_to_video',
+      COALESCE(base_model.capabilities->'image_to_video', '{}'::jsonb) || input_constraints.value,
+      'omni_video',
+      COALESCE(base_model.capabilities->'omni_video', '{}'::jsonb) || input_constraints.value
+    ) AS image_capabilities
+  FROM base_model_catalog base_model
+  CROSS JOIN input_constraints
+  WHERE base_model.provider_key = 'volces'
+    AND (
+      base_model.canonical_model_key IN (
+        'volces:doubao-seedance-2-0-260128',
+        'volces:doubao-seedance-2-0-fast-260128',
+        'volces:doubao-seedance-2-0-mini-260615'
+      )
+      OR base_model.provider_model_name IN (
+        'doubao-seedance-2-0-260128',
+        'doubao-seedance-2-0-fast-260128',
+        'doubao-seedance-2-0-mini-260615'
+      )
+    )
+),
+updated_base_models AS (
+  UPDATE base_model_catalog base_model
+  SET capabilities = target.image_capabilities,
+      metadata = jsonb_set(
+        COALESCE(base_model.metadata, '{}'::jsonb) || jsonb_build_object(
+          'rawModel',
+          COALESCE(base_model.metadata->'rawModel', '{}'::jsonb)
+        ),
+        '{rawModel,capabilities}',
+        target.image_capabilities - 'originalTypes',
+        true
+      ),
+      default_snapshot = CASE
+        WHEN COALESCE(base_model.default_snapshot, '{}'::jsonb) = '{}'::jsonb
+          THEN base_model.default_snapshot
+        ELSE jsonb_set(
+          jsonb_set(
+            base_model.default_snapshot || jsonb_build_object(
+              'metadata',
+              COALESCE(base_model.default_snapshot->'metadata', '{}'::jsonb) || jsonb_build_object(
+                'rawModel',
+                COALESCE(base_model.default_snapshot#>'{metadata,rawModel}', '{}'::jsonb)
+              )
+            ),
+            '{capabilities}',
+            target.image_capabilities,
+            true
+          ),
+          '{metadata,rawModel,capabilities}',
+          target.image_capabilities - 'originalTypes',
+          true
+        )
+      END,
+      updated_at = now()
+  FROM target_base_models target
+  WHERE base_model.id = target.id
+  RETURNING base_model.id
+)
+UPDATE platform_models platform_model
+SET capabilities = COALESCE(platform_model.capabilities, '{}'::jsonb) || jsonb_build_object(
+      'image_to_video',
+      COALESCE(platform_model.capabilities->'image_to_video', '{}'::jsonb) || input_constraints.value,
+      'omni_video',
+      COALESCE(platform_model.capabilities->'omni_video', '{}'::jsonb) || input_constraints.value
+    ),
+    updated_at = now()
+FROM integration_platforms platform
+CROSS JOIN input_constraints
+WHERE platform_model.platform_id = platform.id
+  AND platform.deleted_at IS NULL
+  AND platform.provider = 'volces'
+  AND (
+    platform_model.base_model_id IN (SELECT id FROM updated_base_models)
+    OR COALESCE(NULLIF(platform_model.provider_model_name, ''), platform_model.model_name) IN (
+      'doubao-seedance-2-0-260128',
+      'doubao-seedance-2-0-fast-260128',
+      'doubao-seedance-2-0-mini-260615'
+    )
+  );
@@ -46,8 +46,10 @@ SHA-256，不集中保存响应：
 | `gemini-peak` | 32 | 8 MiB | 8 MiB | 15–30 秒 |

 多参考图视频使用正式 `content` 结构，按 60%/30%/10% 分配 3/6/9 图，包含 JPEG、PNG、
-WebP 和 4K 图片。128 组输入组合避免只命中相同缓存；每四个任务至少一个携带 4K 图片并
-要求在到达上游前完成缩放或重编码。
+WebP、合法 4K 图片和越过 Seedance 2.0 官方输入边界的 `6144x2160` 图片。官方边界为
+宽高均在 `300..6000 px` 且宽高比在 `0.4..2.5`；4K 本身不应被误判为越界。128 组输入
+组合避免只命中相同缓存；每四个任务至少一个携带越界图片，并要求在到达上游前完成缩放、
+补边或重编码。

 - `video-throughput`：1200 个任务并发提交，提交窗口不超过 10 秒。
 - `video-recovery`：96 个 2–3 分钟任务，运行中删除香港 Worker Pod，验证远程任务恢复、