fix(canvas): 节点模型配置 - 切换模型时清理生成参数残留

修复 issue #254；集成 PR #265，保留原 PR 开发者为提交作者。

web/test/canvas-node-config-reset.test.ts

@@ -0,0 +1,78 @@
+import { describe, expect, test } from "bun:test";
+import { applyNodeConfigPatch } from "../src/lib/canvas/canvas-project-domain";
+import { CanvasNodeType, type CanvasNodeData } from "../src/types/canvas";
+
+function imageNode(metadata: Record<string, unknown> = {}): CanvasNodeData {
+    return {
+        id: "image-1",
+        type: CanvasNodeType.Image,
+        title: "图片",
+        position: { x: 0, y: 0 },
+        width: 400,
+        height: 400,
+        metadata: metadata as CanvasNodeData["metadata"],
+    };
+}
+
+describe("applyNodeConfigPatch 模型切换清理生成参数", () => {
+    test("切换模型标识时清理旧模型的能力档位参数，回落全局配置", () => {
+        const node = imageNode({
+            model: "default::gpt-image-2",
+            size: "1:1",
+            quality: "high",
+            count: 2,
+            transparentBackground: "true",
+        });
+        const next = applyNodeConfigPatch(node, { model: "default::grok-imagine-video" });
+        expect(next.metadata?.model).toBe("default::grok-imagine-video");
+        expect(next.metadata?.size).toBeUndefined();
+        expect(next.metadata?.quality).toBeUndefined();
+        expect(next.metadata?.count).toBeUndefined();
+        expect(next.metadata?.transparentBackground).toBeUndefined();
+    });
+
+    test("视频参数随模型切换一并清理", () => {
+        const node = imageNode({
+            model: "default::seedance-video",
+            seconds: "10",
+            vquality: "1080p",
+            generateAudio: "true",
+            watermark: "false",
+        });
+        const next = applyNodeConfigPatch(node, { model: "default::veo-3" });
+        expect(next.metadata?.seconds).toBeUndefined();
+        expect(next.metadata?.vquality).toBeUndefined();
+        expect(next.metadata?.generateAudio).toBeUndefined();
+        expect(next.metadata?.watermark).toBeUndefined();
+    });
+
+    test("同一模型标识的参数调整不受影响", () => {
+        const node = imageNode({ model: "default::gpt-image-2", size: "1:1", quality: "high" });
+        const next = applyNodeConfigPatch(node, { model: "default::gpt-image-2", quality: "medium" });
+        expect(next.metadata?.model).toBe("default::gpt-image-2");
+        expect(next.metadata?.quality).toBe("medium");
+        expect(next.metadata?.size).toBe("1:1");
+    });
+
+    test("非模型 patch（仅调整参数）不触发清理", () => {
+        const node = imageNode({ model: "default::gpt-image-2", size: "1:1" });
+        const next = applyNodeConfigPatch(node, { size: "16:9" });
+        expect(next.metadata?.size).toBe("16:9");
+        expect(next.metadata?.model).toBe("default::gpt-image-2");
+    });
+
+    test("模型切换时同批显式参数仍优先于清理", () => {
+        const node = imageNode({ model: "default::gpt-image-2", size: "1:1", quality: "high" });
+        const next = applyNodeConfigPatch(node, { model: "default::grok-imagine-video", size: "16:9" });
+        expect(next.metadata?.model).toBe("default::grok-imagine-video");
+        expect(next.metadata?.size).toBe("16:9");
+        expect(next.metadata?.quality).toBeUndefined();
+    });
+
+    test("内容型字段不随模型切换清理", () => {
+        const node = imageNode({ model: "default::gpt-image-2", prompt: "一只猫", composerContent: "生成一只猫" });
+        const next = applyNodeConfigPatch(node, { model: "default::grok-imagine-video" });
+        expect(next.metadata?.prompt).toBe("一只猫");
+        expect(next.metadata?.composerContent).toBe("生成一只猫");
+    });
+});