fix(confidence-engine): sanitize reasoning error responses

tests/app/api/cases-overview-route.test.js

@@ -1,4 +1,4 @@
-import { beforeEach, describe, expect, it, vi } from "vitest";
+import { describe, expect, it, vi } from "vitest";
 
 const mockSynthesize = vi.fn();
 
@@ -11,6 +11,10 @@ vi.mock("@/lib/llm/provider.js", () => ({
   getProviderModelName: () => "gpt-5.6-terra",
 }));
 
+vi.mock("@/lib/supabase/api-auth.js", () => ({
+  withAuthenticatedApi: (handler) => handler,
+}));
+
 describe("POST /api/cases/overview provider routing", () => {
   beforeEach(() => mockSynthesize.mockClear());
 
@@ -26,4 +30,36 @@ describe("POST /api/cases/overview provider routing", () => {
     expect(response.status).toBe(200);
     expect(mockSynthesize.mock.calls[0][1]).toMatchObject({ modelName: "gpt-5.6-terra" });
   });
+
+  it("sanitizes provider failure details", async () => {
+    const error = Object.assign(
+      new Error("Ollama /api/generate returned 500 from private host"),
+      { statusCode: 502 },
+    );
+    mockSynthesize.mockImplementationOnce(async () => {
+      throw error;
+    });
+
+    const { POST } = await import("@/app/api/cases/overview/route.js");
+    const response = await POST(new Request("http://localhost/api/cases/overview", {
+      method: "POST",
+      headers: { "Content-Type": "application/json" },
+      body: JSON.stringify({ situationGraph: { nodes: [], edges: [] } }),
+    }));
+
+    expect(response.status).toBe(502);
+    const rawData = await response.text();
+    const data = JSON.parse(rawData);
+
+    expect(data.success).toBe(false);
+    expect(data.stage).toBe("provider");
+    expect(data).toEqual({
+      success: false,
+      stage: "provider",
+      error: "Reasoning request could not be completed.",
+    });
+
+    // Prove raw provider diagnostics are NOT exposed at the route boundary
+    expect(rawData).not.toContain(error.message);
+  });
 });