fix(pdf): flag scanned PDFs in pdf-to-text and serve text as UTF-8 (#603)

When a PDF has no text layer (scanned or image-only), pdf-to-text now
returns a 422 that points at the OCR tool instead of a silent empty file,
and text downloads carry charset=utf-8 so UTF-8 Arabic renders correctly
when the .txt is viewed inline.

Fixes #589

tests/unit/api/document-route-processors.test.ts

@@ -31,7 +31,7 @@ vi.mock("../../../apps/api/src/config.js", () => ({
 vi.mock("@snapotter/doc-engine", () => ({
   htmlToPdfPy: vi.fn(),
   pdfFlattenPy: vi.fn(),
-  pdfTextPy: vi.fn(async () => ({ chars: 12 })),
+  pdfTextPy: vi.fn(async () => ({ chars: 12, hasText: true })),
   qpdfAvailable: vi.fn(() => false),
   qpdfCheck: vi.fn(),
   qpdfPageCount: vi.fn(),
@@ -105,6 +105,19 @@ describe("document route processors", () => {
     });
   });
 
+  it("rejects a PDF with no text layer and points the user to OCR", async () => {
+    // A scanned/image-only PDF yields hasText=false; the route must reject with
+    // an OCR handoff instead of returning a silent empty .txt (#589).
+    vi.mocked(pdfTextPy).mockResolvedValueOnce({ chars: 0, hasText: false });
+    registerPdfToText(createMockApp());
+    const config = getToolConfig("pdf-to-text");
+
+    await withScratch(async (scratchDir) => {
+      const ctx = createCtx(scratchDir, "scanned.pdf");
+      await expect(config?.processV2?.(ctx)).rejects.toThrow(/text layer/i);
+    });
+  });
+
   it("flattens PDFs through pdfFlattenPy", async () => {
     registerFlattenPdf(createMockApp());
     const config = getToolConfig("flatten-pdf");