Split param processor and tighten Volces frame validation

2026-05-14 00:41:06 +08:00
parent 3225833f96
commit cdf469eccf
9 changed files with 1983 additions and 1538 deletions
@@ -381,6 +381,128 @@ func TestParamProcessorVideoCapabilitiesNormalizeAndFilter(t *testing.T) {
 	}
 }

+func TestParamProcessorDowngradesReferenceImagesToFrames(t *testing.T) {
+	candidate := store.RuntimeModelCandidate{
+		ModelType: "image_to_video",
+		Capabilities: map[string]any{
+			"image_to_video": map[string]any{
+				"input_first_frame":                 true,
+				"input_first_last_frame":            true,
+				"input_reference_generate_single":   false,
+				"input_reference_generate_multiple": false,
+			},
+		},
+	}
+	body := map[string]any{
+		"model":  "Seedance",
+		"prompt": "animate it",
+		"content": []any{
+			map[string]any{"type": "text", "text": "animate it"},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/first.png"}},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/last.png"}},
+		},
+	}
+
+	result := preprocessRequestWithLog("videos.generations", body, candidate)
+	if result.Err != nil {
+		t.Fatalf("two image references should downgrade to first/last frames: %v", result.Err)
+	}
+	content := contentItems(result.Body["content"])
+	if stringFromAny(content[1]["role"]) != "first_frame" || stringFromAny(content[2]["role"]) != "last_frame" {
+		t.Fatalf("expected first/last frame downgrade, got %+v", content)
+	}
+}
+
+func TestParamProcessorDowngradesSingleReferenceImageToFirstFrame(t *testing.T) {
+	candidate := store.RuntimeModelCandidate{
+		ModelType: "image_to_video",
+		Capabilities: map[string]any{
+			"image_to_video": map[string]any{
+				"input_first_frame":                 true,
+				"input_first_last_frame":            true,
+				"input_reference_generate_single":   false,
+				"input_reference_generate_multiple": false,
+			},
+		},
+	}
+	body := map[string]any{
+		"model":  "Seedance",
+		"prompt": "animate it",
+		"content": []any{
+			map[string]any{"type": "text", "text": "animate it"},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/first.png"}},
+		},
+	}
+
+	result := preprocessRequestWithLog("videos.generations", body, candidate)
+	if result.Err != nil {
+		t.Fatalf("single image reference should downgrade to first frame: %v", result.Err)
+	}
+	content := contentItems(result.Body["content"])
+	if stringFromAny(content[1]["role"]) != "first_frame" {
+		t.Fatalf("expected first frame downgrade, got %+v", content)
+	}
+}
+
+func TestParamProcessorRejectsUnsafeReferenceImageDowngrade(t *testing.T) {
+	candidate := store.RuntimeModelCandidate{
+		ModelType: "image_to_video",
+		Capabilities: map[string]any{
+			"image_to_video": map[string]any{
+				"input_first_frame":                 true,
+				"input_first_last_frame":            false,
+				"input_reference_generate_single":   false,
+				"input_reference_generate_multiple": false,
+			},
+		},
+	}
+	body := map[string]any{
+		"model":  "Seedance",
+		"prompt": "animate it",
+		"content": []any{
+			map[string]any{"type": "text", "text": "animate it"},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/first.png"}},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/last.png"}},
+		},
+	}
+
+	result := preprocessRequestWithLog("videos.generations", body, candidate)
+	if result.Err == nil {
+		t.Fatalf("two image references should be rejected when first/last frame is unsupported")
+	}
+	if len(result.Log.Changes) == 0 || result.Log.Changes[len(result.Log.Changes)-1].Action != "reject" {
+		t.Fatalf("expected reject preprocessing log, got %+v", result.Log.Changes)
+	}
+}
+
+func TestParamProcessorRejectsVideoOrAudioReferenceDowngrade(t *testing.T) {
+	candidate := store.RuntimeModelCandidate{
+		ModelType: "image_to_video",
+		Capabilities: map[string]any{
+			"image_to_video": map[string]any{
+				"input_first_frame":                 true,
+				"input_first_last_frame":            true,
+				"input_reference_generate_single":   false,
+				"input_reference_generate_multiple": false,
+			},
+		},
+	}
+	body := map[string]any{
+		"model":  "Seedance",
+		"prompt": "animate it",
+		"content": []any{
+			map[string]any{"type": "text", "text": "animate it"},
+			map[string]any{"type": "image_url", "role": "reference_image", "image_url": map[string]any{"url": "https://example.com/first.png"}},
+			map[string]any{"type": "video_url", "role": "reference_video", "video_url": map[string]any{"url": "https://example.com/ref.mp4"}},
+		},
+	}
+
+	result := preprocessRequestWithLog("videos.generations", body, candidate)
+	if result.Err == nil {
+		t.Fatalf("video reference should be rejected instead of downgraded")
+	}
+}
+
 func TestParamProcessorDurationRangeRoundsFractionalSecondsUp(t *testing.T) {
 	body := map[string]any{
 		"model":    "Seedance",