fix(canvas): 模型生成 - 保留安全的上游失败原因并展示到用户卡片

web/test/generation-error-details.test.ts

@@ -0,0 +1,24 @@
+import { describe, expect, test } from "bun:test";
+import { CONTENT_MODERATION_MESSAGE, generationErrorMessage, generationFailureMetadata } from "../src/lib/generation-error";
+
+describe("safe upstream error details", () => {
+    const detail = "非常抱歉，生成的图片可能违反了关于裸露、色情或情色内容的防护限制。请重试或修改提示语。";
+    test("keeps the specific image rejection on the card", () => {
+        const message = `模型服务拒绝了请求，请检查模型和参数；上游：${detail}`;
+        expect(generationErrorMessage(new Error(message))).toBe(message);
+    });
+    test("does not replace a safe upstream message with a generic HTTP hint", () => {
+        const message = `模型服务暂时不可用（HTTP 500）；上游：${detail}`;
+        expect(generationErrorMessage(message)).toBe(message);
+        expect(generationErrorMessage("模型服务暂时不可用（HTTP 500）")).toBe("网络异常。");
+    });
+    test("keeps details when moderation metadata is stored", () => {
+        const message = `内容审核未通过；上游：${detail}`;
+        const metadata = generationFailureMetadata(message, "test prompt");
+        expect(metadata.errorDetails).toBe(`${CONTENT_MODERATION_MESSAGE}；上游：${detail}`);
+        expect(metadata.generationErrorCode).toBe("sensitive_words_detected");
+    });
+    test("upstream marker does not bypass URL hiding", () => {
+        expect(generationErrorMessage("模型服务请求失败；上游：https://private.test")).toBe("网络异常。");
+    });
+});