feat(platform): 插件协议系统 - 支持声明式渠道、工作流与参考媒体映射

web/test/canvas-model-policy.test.ts

@@ -129,6 +129,28 @@ describe("逻辑模型选择", () => {
         expect(resolveModelGenerationDefaults(config, model, "video", {}, { videoSeconds: "6" }).videoSeconds).toBe("15");
     });
 
+
+    test("渠道视频能力配置优先于旧的全局 6 秒和 1:1 配置", () => {
+        const model = "autodl-channel::MiniMax H3";
+        const profile = defaultModelCapabilityConfig("autodl-comfyui", "MiniMax H3");
+        profile.video!.duration = { selection: "range", min: 1, max: 15, step: 1, default: 15 };
+        profile.video!.defaultRatio = "16:9";
+        const config: AiConfig = {
+            ...defaultConfig,
+            videoSeconds: "6",
+            size: "1:1",
+            channels: [{ id: "autodl-channel", name: "AutoDL", baseUrl: "https://autodl.art", apiKey: "system", apiFormat: "openai", scope: "system", models: ["MiniMax H3"], modelCosts: [{ model: "MiniMax H3", capability: "video", protocol: "autodl-comfyui", billingMode: "fixed_request", unitPriceMicrocredits: 1, capabilityConfig: profile }] }],
+            models: [model],
+            videoModels: [model],
+            videoModel: model,
+            model,
+        };
+
+        const defaults = resolveModelGenerationDefaults(config, model, "video", {}, { videoSeconds: "6", size: "1:1" });
+        expect(defaults.videoSeconds).toBe("15");
+        expect(defaults.size).toBe("16:9");
+    });
+
     test("后台标注的视频模型不因内部标识缺少视频关键词而回退", () => {
         const config = policyConfig();
         const selectedModel = config.videoModels[0]!;