Merge remote-tracking branch 'origin/main' into fix/v1397-tz

src/app/api/labs/ocr/extract/__tests__/route.test.ts

@@ -229,11 +229,15 @@ describe("POST /api/labs/ocr/extract — vision PDF rasterization", () => {
     vi.mocked(rasterizePdf).mockResolvedValue({
       ok: true,
       images: [{ mediaType: "image/jpeg", dataBase64: "cGFnZQ==" }],
+      pageCount: 1,
     });
     vi.mocked(runOcrExtraction).mockResolvedValue({ rows: [] } as never);
 
     const res = await POST(pdfReq());
     expect(res.status).toBe(200);
+    // The whole PDF was read: no coverage note on the response.
+    const whole = (await res.json()) as { data: Record<string, unknown> };
+    expect(whole.data).not.toHaveProperty("pageCoverage");
     expect(rasterizePdf).toHaveBeenCalledOnce();
     // The rendered page images flow through as `input_image`s; no native PDF
     // document block is sent for a non-Anthropic provider.
@@ -245,6 +249,26 @@ describe("POST /api/labs/ocr/extract — vision PDF rasterization", () => {
     );
   });
 
+  it("says on the response when only the first pages of a long PDF were read", async () => {
+    const tenPages = Array.from({ length: 10 }, () => ({
+      mediaType: "image/jpeg" as const,
+      dataBase64: "cGFnZQ==",
+    }));
+    vi.mocked(rasterizePdf).mockResolvedValue({
+      ok: true,
+      images: tenPages,
+      pageCount: 23,
+    });
+    vi.mocked(runOcrExtraction).mockResolvedValue({ rows: [] } as never);
+
+    const res = await POST(pdfReq());
+    expect(res.status).toBe(200);
+    const body = (await res.json()) as {
+      data: { pageCoverage?: { read: number; total: number } };
+    };
+    expect(body.data.pageCoverage).toEqual({ read: 10, total: 23 });
+  });
+
   it("hands the slot back when the upload is not an image or PDF", async () => {
     const form = new FormData();
     form.append(

src/app/api/labs/ocr/extract/route.ts

@@ -374,6 +374,7 @@ async function handleVisionExtract(
       dataBase64: string;
     }[];
     let documents: { mediaType: "application/pdf"; dataBase64: string }[];
+    let pageCoverage: { read: number; total: number } | undefined;
 
     if (mime === "application/pdf") {
       if (pick.pdfSupported) {
@@ -404,6 +405,20 @@ async function handleVisionExtract(
         }
         images = raster.images;
         documents = [];
+        // A PDF longer than the page cap is read from its first pages only.
+        // That used to live in the wide event alone; the person reviewing the
+        // rows is the one who needs to know the rest of the report was not
+        // read.
+        if (raster.pageCount > raster.images.length) {
+          pageCoverage = {
+            read: raster.images.length,
+            total: raster.pageCount,
+          };
+          annotate({
+            action: { name: "labs.ocr.pagesCapped" },
+            meta: { read: raster.images.length, total: raster.pageCount },
+          });
+        }
       }
     } else {
       images = [{ mediaType: mime, dataBase64: buffer.toString("base64") }];
@@ -431,7 +446,7 @@ async function handleVisionExtract(
         0,
         { servedBy: pick.entry.providerType, reservedOwner: reservation.owner },
       );
-      return apiSuccess(result);
+      return apiSuccess(pageCoverage ? { ...result, pageCoverage } : result);
     } catch (err) {
       // Provider/extraction failure — the call may still have burned tokens, so
       // reconcile against the reserved estimate rather than refunding in full.