feat: support image quality control

2026-05-28 00:17:27 +08:00
parent 1d3a4f1da9
commit f5c69b9852
9 changed files with 243 additions and 9 deletions
@@ -214,7 +214,7 @@ type ImageGenerationRequest struct {
 	Prompt         string `json:"prompt" example:"A watercolor robot reading a book"`
 	N              int    `json:"n,omitempty" example:"1"`
 	Size           string `json:"size,omitempty" example:"1024x1024"`
-	Quality        string `json:"quality,omitempty" example:"standard"`
+	Quality        string `json:"quality,omitempty" example:"auto"`
 	ResponseFormat string `json:"response_format,omitempty" example:"url"`
 	RunMode        string `json:"runMode,omitempty" example:"simulation"`
 }
@@ -226,6 +226,7 @@ type ImageEditRequest struct {
 	Mask           string `json:"mask,omitempty" example:"https://example.com/mask.png"`
 	N              int    `json:"n,omitempty" example:"1"`
 	Size           string `json:"size,omitempty" example:"1024x1024"`
+	Quality        string `json:"quality,omitempty" example:"auto"`
 	ResponseFormat string `json:"response_format,omitempty" example:"url"`
 	RunMode        string `json:"runMode,omitempty" example:"simulation"`
 }
@@ -63,6 +63,7 @@ func NewParamProcessorChain() ParamProcessorChain {
 			durationProcessor{},
 			audioProcessor{},
 			imageCountProcessor{},
+			imageQualityProcessor{},
 		},
 	}
 }
@@ -378,3 +378,64 @@ func (imageCountProcessor) Process(params map[string]any, modelType string, cont
 	params["n"] = count
 	return true
 }
+
+type imageQualityProcessor struct{}
+
+func (imageQualityProcessor) Name() string { return "ImageQualityProcessor" }
+
+var openAICompatibleImageQualities = map[string]struct{}{
+	"low":    {},
+	"medium": {},
+	"high":   {},
+	"auto":   {},
+}
+
+func (imageQualityProcessor) ShouldProcess(params map[string]any, modelType string, context *paramProcessContext) bool {
+	if modelType != "image_generate" && modelType != "image_edit" {
+		return false
+	}
+	_, ok := params["quality"]
+	return ok
+}
+
+func (imageQualityProcessor) Process(params map[string]any, modelType string, context *paramProcessContext) bool {
+	capability := capabilityForType(context.modelCapability, modelType)
+	quality := stringFromAny(params["quality"])
+	if supportsImageQualityControl(capability) && isOpenAICompatibleImageQuality(quality) {
+		return true
+	}
+
+	before := params["quality"]
+	delete(params, "quality")
+	context.recordChange(
+		"ImageQualityProcessor",
+		"remove",
+		"quality",
+		before,
+		nil,
+		"模型能力未开启生成质量控制，已移除 quality 参数。",
+		capabilityPath(modelType, "support_quality_control"),
+		capabilityValue(context.modelCapability, modelType, "support_quality_control"),
+	)
+	return true
+}
+
+func supportsImageQualityControl(capability map[string]any) bool {
+	if capability == nil {
+		return false
+	}
+	for _, key := range []string{"support_quality_control", "supportQualityControl", "quality_control", "qualityControl", "quality"} {
+		if boolFromAny(capability[key]) {
+			return true
+		}
+	}
+	return false
+}
+
+func isOpenAICompatibleImageQuality(value string) bool {
+	if value == "" {
+		return false
+	}
+	_, ok := openAICompatibleImageQualities[value]
+	return ok
+}
@@ -660,3 +660,56 @@ func TestParamProcessorImageResolutionAndOutputCount(t *testing.T) {
 		t.Fatalf("image count should be capped to 4, got %+v", processed["n"])
 	}
 }
+
+func TestParamProcessorImageQualityControl(t *testing.T) {
+	body := map[string]any{
+		"model":   "mock-image",
+		"prompt":  "draw",
+		"quality": "high",
+	}
+
+	unsupported := preprocessRequestWithLog("images.generations", body, store.RuntimeModelCandidate{
+		ModelType: "image_generate",
+		Capabilities: map[string]any{
+			"image_generate": map[string]any{
+				"output_resolutions": []any{"1K"},
+			},
+		},
+	})
+	if _, ok := unsupported.Body["quality"]; ok {
+		t.Fatalf("quality should be removed when capability does not support it: %+v", unsupported.Body)
+	}
+	if len(unsupported.Log.Changes) == 0 || unsupported.Log.Changes[len(unsupported.Log.Changes)-1].CapabilityPath != "capabilities.image_generate.support_quality_control" {
+		t.Fatalf("expected quality removal to be logged against support_quality_control, got %+v", unsupported.Log.Changes)
+	}
+
+	supported := preprocessRequest("images.generations", body, store.RuntimeModelCandidate{
+		ModelType: "image_generate",
+		Capabilities: map[string]any{
+			"image_generate": map[string]any{
+				"support_quality_control": true,
+				"output_resolutions":      []any{"1K"},
+			},
+		},
+	})
+	if supported["quality"] != "high" {
+		t.Fatalf("quality should be retained when capability supports it: %+v", supported)
+	}
+
+	incompatible := preprocessRequest("images.generations", map[string]any{
+		"model":   "mock-image",
+		"prompt":  "draw",
+		"quality": "standard",
+	}, store.RuntimeModelCandidate{
+		ModelType: "image_generate",
+		Capabilities: map[string]any{
+			"image_generate": map[string]any{
+				"support_quality_control": true,
+				"output_resolutions":      []any{"1K"},
+			},
+		},
+	})
+	if _, ok := incompatible["quality"]; ok {
+		t.Fatalf("OpenAI-compatible GPT image quality should reject standard: %+v", incompatible)
+	}
+}
@@ -0,0 +1,60 @@
+-- GPT Image 系列支持 OpenAI-compatible quality 参数；其他图像模型默认不声明，
+-- runner 会在参数预处理时移除未支持模型上的 quality。
+
+CREATE OR REPLACE FUNCTION pg_temp._tmp_enable_image_quality_control(capabilities jsonb)
+RETURNS jsonb AS $$
+DECLARE
+  out jsonb := COALESCE(capabilities, '{}'::jsonb);
+BEGIN
+  IF out ? 'image_generate' THEN
+    out := jsonb_set(out, '{image_generate,support_quality_control}', 'true'::jsonb, true);
+  END IF;
+  IF out ? 'image_edit' THEN
+    out := jsonb_set(out, '{image_edit,support_quality_control}', 'true'::jsonb, true);
+  END IF;
+  RETURN out;
+END;
+$$ LANGUAGE plpgsql;
+
+UPDATE base_model_catalog
+SET capabilities = pg_temp._tmp_enable_image_quality_control(capabilities),
+    default_snapshot = CASE
+      WHEN COALESCE(default_snapshot, '{}'::jsonb) = '{}'::jsonb THEN default_snapshot
+      WHEN jsonb_typeof(default_snapshot->'metadata'->'rawModel'->'capabilities') = 'object' THEN jsonb_set(
+        jsonb_set(
+          default_snapshot,
+          '{capabilities}',
+          pg_temp._tmp_enable_image_quality_control(COALESCE(default_snapshot->'capabilities', '{}'::jsonb)),
+          true
+        ),
+        '{metadata,rawModel,capabilities}',
+        pg_temp._tmp_enable_image_quality_control(COALESCE(default_snapshot->'metadata'->'rawModel'->'capabilities', '{}'::jsonb)),
+        true
+      )
+      ELSE jsonb_set(
+        default_snapshot,
+        '{capabilities}',
+        pg_temp._tmp_enable_image_quality_control(COALESCE(default_snapshot->'capabilities', '{}'::jsonb)),
+        true
+      )
+    END,
+    metadata = CASE
+      WHEN jsonb_typeof(metadata->'rawModel'->'capabilities') = 'object' THEN jsonb_set(
+        metadata,
+        '{rawModel,capabilities}',
+        pg_temp._tmp_enable_image_quality_control(COALESCE(metadata->'rawModel'->'capabilities', '{}'::jsonb)),
+        true
+      )
+      ELSE metadata
+    END,
+    updated_at = now()
+WHERE provider_model_name IN ('gpt-image-1', 'gpt-image-1.5', 'gpt-image-2')
+  AND capabilities ?| ARRAY['image_generate', 'image_edit'];
+
+UPDATE platform_models
+SET capabilities = pg_temp._tmp_enable_image_quality_control(capabilities),
+    updated_at = now()
+WHERE COALESCE(NULLIF(provider_model_name, ''), model_name) IN ('gpt-image-1', 'gpt-image-1.5', 'gpt-image-2')
+  AND capabilities ?| ARRAY['image_generate', 'image_edit'];
+
+DROP FUNCTION pg_temp._tmp_enable_image_quality_control(jsonb);