cmd/cognitive-worker/multimodal.go

@@ -2,6 +2,7 @@ package main
 
 import (
 	"encoding/json"
+	"fmt"
 	"io"
 	"log"
 	"net/http"
@@ -38,7 +39,8 @@ func (g multimodalGateway) handler(operation capability.MultimodalOperation) htt
 				http.Error(w, "local image generation capability unavailable", http.StatusBadGateway)
 				return
 			}
-			image, err := g.imageGenerator.Generate(r.Context(), g.imageModel, r.FormValue("prompt"), r.FormValue("size"))
+			prompt := r.FormValue("prompt")
+			image, err := g.imageGenerator.Generate(r.Context(), g.imageModel, prompt, imageRequestSize(prompt, r.FormValue("size")))
 			if err != nil || strings.TrimSpace(image.DataBase64) == "" {
 				logMultimodalFailure(operation, err)
 				http.Error(w, "local image generation capability unavailable", http.StatusBadGateway)
@@ -96,6 +98,40 @@ func (g multimodalGateway) handler(operation capability.MultimodalOperation) htt
 	}
 }
 
+// imageRequestSize is a transport-level normalization for the local image
+// capability.  An explicit, valid size always wins.  When the caller only
+// expresses a conventional output-quality target, the capability translates it
+// to a supported provider size without exposing provider workflow details to
+// the interaction core.  This policy is deliberately local and replaceable
+// with the image capability implementation.
+func imageRequestSize(prompt, requested string) string {
+	if isImageSize(requested) {
+		return strings.TrimSpace(requested)
+	}
+	words := strings.Fields(strings.ToLower(prompt))
+	for index, word := range words {
+		switch word {
+		case "4k", "uhd":
+			return "2048x2048"
+		case "hd":
+			return "1024x1024"
+		case "ultra":
+			if index+1 < len(words) && words[index+1] == "hd" {
+				return "2048x2048"
+			}
+		}
+	}
+	return ""
+}
+
+func isImageSize(value string) bool {
+	var width, height int
+	if _, err := fmt.Sscanf(strings.TrimSpace(value), "%dx%d", &width, &height); err != nil {
+		return false
+	}
+	return width >= 256 && height >= 256 && width <= 2048 && height <= 2048
+}
+
 func logMultimodalFailure(operation capability.MultimodalOperation, err error) {
 	// Keep provider details in server diagnostics; never return them to chat.
 	log.Printf("multimodal operation %s failed: %v", operation, err)

cmd/cognitive-worker/multimodal_test.go

@@ -47,6 +47,27 @@ func TestMultimodalImageGenerationFailsClosedWhenSkillIsUnbound(t *testing.T) {
 	}
 }
 
+func TestImageRequestSizeUsesExplicitSizeOrQualityTarget(t *testing.T) {
+	for _, test := range []struct {
+		name      string
+		prompt    string
+		requested string
+		want      string
+	}{
+		{name: "explicit", prompt: "нарисуй птицу", requested: "768x512", want: "768x512"},
+		{name: "hd", prompt: "нарисуй птицу в HD", want: "1024x1024"},
+		{name: "uhd", prompt: "draw a bird in UHD", want: "2048x2048"},
+		{name: "ordinary", prompt: "нарисуй птицу", want: ""},
+		{name: "invalid explicit", prompt: "нарисуй птицу", requested: "100x100", want: ""},
+	} {
+		t.Run(test.name, func(t *testing.T) {
+			if got := imageRequestSize(test.prompt, test.requested); got != test.want {
+				t.Fatalf("imageRequestSize(%q, %q) = %q, want %q", test.prompt, test.requested, got, test.want)
+			}
+		})
+	}
+}
+
 func TestMultimodalVideoUnderstandingFailsClosedWhenSkillIsUnbound(t *testing.T) {
 	server := multimodalGateway{now: time.Now}
 	request := multipartRequest(t, "/v1/multimodal/video-understand", "clip.mp4", []byte("video"))