Fix Tesseract workerPath in Next.js standalone container, pool worker, and add timeout

src/app/api/attachments/ocr/route.ts

@@ -7,10 +7,11 @@ export async function POST(req: NextRequest) {
     const body = await req.json().catch(() => ({}));
     const limit = Math.min(parseInt(String(body.limit || '10'), 10) || 10, 50);
 
-    // Find image or PDF attachments that do not have extractedText yet
+    // Find image or PDF attachments that do not have extractedText yet (skip tiny icons < 2 KB)
     const pendingAttachments = await prisma.attachment.findMany({
       where: {
         extractedText: null,
+        sizeBytes: { gte: 2048 },
         OR: [
           { mimeType: { startsWith: 'image/' } },
           { mimeType: 'application/pdf' },

src/lib/ocr.ts

@@ -20,25 +20,60 @@ async function extractPdfText(buffer: Buffer): Promise<string | null> {
   }
 }
 
+let sharedWorker: Tesseract.Worker | null = null;
+let sharedWorkerPromise: Promise<Tesseract.Worker> | null = null;
+
+async function getOcrWorker(): Promise<Tesseract.Worker> {
+  if (sharedWorker) return sharedWorker;
+  if (!sharedWorkerPromise) {
+    let workerPath: string | undefined;
+    try {
+      workerPath = require.resolve('tesseract.js/src/worker-script/node/index.js');
+    } catch {}
+
+    const options: Record<string, unknown> = {};
+    if (workerPath) {
+      options.workerPath = workerPath;
+    }
+
+    sharedWorkerPromise = createWorker('eng', 1, options)
+      .then((w) => {
+        sharedWorker = w;
+        return w;
+      })
+      .catch((err) => {
+        sharedWorkerPromise = null;
+        throw err;
+      });
+  }
+  return sharedWorkerPromise;
+}
+
 /**
- * Recognizes text inside an image using Tesseract OCR.
+ * Recognizes text inside an image using Tesseract OCR with pooled worker and timeout.
  */
 export async function recognizeImageOcr(imageBufferOrPath: Buffer | string): Promise<string | null> {
-  let worker: Tesseract.Worker | null = null;
   try {
-    worker = await createWorker('eng');
-    const ret = await worker.recognize(imageBufferOrPath);
+    const worker = await getOcrWorker();
+    const ret = await Promise.race([
+      worker.recognize(imageBufferOrPath),
+      new Promise<never>((_, reject) =>
+        setTimeout(() => reject(new Error('OCR recognition timeout (30s)')), 30000)
+      ),
+    ]);
     const text = ret.data?.text?.trim() || '';
     return text.length > 0 ? text.slice(0, 50000) : null;
   } catch (err) {
     console.error('Tesseract OCR error:', err);
-    return null;
-  } finally {
-    if (worker) {
+    // If worker errored, terminate and reset so next call re-spawns cleanly
+    if (sharedWorker) {
       try {
-        await worker.terminate();
+        await sharedWorker.terminate();
       } catch {}
+      sharedWorker = null;
+      sharedWorkerPromise = null;
     }
+    return null;
   }
 }
 
@@ -82,6 +117,10 @@ export async function extractTextFromAttachment(
     lowerName.endsWith('.bmp') ||
     lowerName.endsWith('.tiff')
   ) {
+    // Skip tiny tracking pixels / spacer icons under 2 KB
+    if (buffer.length < 2048) {
+      return null;
+    }
     return await recognizeImageOcr(buffer);
   }