fix(core): Polyfill DOMMatrix when parsing PDFs in Data Loader (#31669)

packages/@n8n/ai-utilities/package.json

@@ -102,6 +102,7 @@
     "tmp-promise": "3.0.3",
     "js-tiktoken": "catalog:",
     "https-proxy-agent": "catalog:",
+    "@thednp/dommatrix": "^2.0.12",
     "pdf-parse": "catalog:",
     "proxy-from-env": "^1.1.0",
     "undici": "^6.21.0"

packages/@n8n/ai-utilities/src/__tests__/utils/loaders/n8n-pdf-loader.test.ts

@@ -225,4 +225,35 @@ describe('N8nPdfLoader', () => {
 			blobType: 'application/pdf',
 		});
 	});
+
+	// `pdf-parse` v2 is backed by pdfjs-dist, which references the `DOMMatrix`
+	// global. Node.js does not provide it, so the loader must polyfill it before
+	// parsing — otherwise pdfjs throws "DOMMatrix is not defined" on PDFs that
+	// exercise that code path.
+	describe('DOMMatrix polyfill', () => {
+		const hadDomMatrix = 'DOMMatrix' in globalThis;
+		const originalDomMatrix: unknown = Reflect.get(globalThis, 'DOMMatrix');
+
+		afterAll(() => {
+			if (hadDomMatrix) {
+				Reflect.set(globalThis, 'DOMMatrix', originalDomMatrix);
+			} else {
+				Reflect.deleteProperty(globalThis, 'DOMMatrix');
+			}
+		});
+
+		it('defines a usable DOMMatrix global before parsing when one is absent', async () => {
+			Reflect.deleteProperty(globalThis, 'DOMMatrix');
+			mockGetText.mockResolvedValue({
+				pages: [{ num: 1, text: 'page' }],
+				text: 'page',
+				total: 1,
+			});
+
+			const loader = new N8nPdfLoader(makeBlob());
+			await loader.load();
+
+			expect(typeof Reflect.get(globalThis, 'DOMMatrix')).toBe('function');
+		});
+	});
 });

packages/@n8n/ai-utilities/src/utils/loaders/n8n-pdf-loader.ts

@@ -25,6 +25,13 @@ export class N8nPdfLoader extends BufferLoader {
 	}
 
 	protected async parse(raw: Buffer, metadata: Record<string, unknown>): Promise<Document[]> {
+		// pdf-parse v2 is backed by pdfjs-dist, which expects a `DOMMatrix` global
+		// that Node.js does not provide. Polyfill it before parsing.
+		if (typeof Reflect.get(globalThis, 'DOMMatrix') === 'undefined') {
+			const { default: DOMMatrix } = await import('@thednp/dommatrix');
+			Reflect.set(globalThis, 'DOMMatrix', DOMMatrix);
+		}
+
 		const { PDFParse } = await import('pdf-parse');
 
 		// Buffer extends Uint8Array; PDFParse accepts it directly.

packages/@n8n/instance-ai/package.json

@@ -61,6 +61,7 @@
     "@n8n/utils": "workspace:*",
     "@n8n/workflow-sdk": "workspace:*",
     "@opentelemetry/api": "^1.9.0",
+    "@thednp/dommatrix": "^2.0.12",
     "csv-parse": "catalog:",
     "fast-glob": "catalog:",
     "flatted": "catalog:",

packages/@n8n/instance-ai/src/parsers/__tests__/pdf-parser.test.ts

@@ -97,4 +97,34 @@ describe('extractPdfText', () => {
 			}),
 		).rejects.toThrow(/no extractable text/);
 	});
+
+	// `pdf-parse` v2 is backed by pdfjs-dist, which references the `DOMMatrix`
+	// global. Node.js does not provide it, so the parser must polyfill it before
+	// parsing — otherwise pdfjs throws "DOMMatrix is not defined" on PDFs that
+	// exercise that code path.
+	describe('DOMMatrix polyfill', () => {
+		const hadDomMatrix = 'DOMMatrix' in globalThis;
+		const originalDomMatrix: unknown = Reflect.get(globalThis, 'DOMMatrix');
+
+		afterAll(() => {
+			if (hadDomMatrix) {
+				Reflect.set(globalThis, 'DOMMatrix', originalDomMatrix);
+			} else {
+				Reflect.deleteProperty(globalThis, 'DOMMatrix');
+			}
+		});
+
+		it('defines a usable DOMMatrix global before parsing when one is absent', async () => {
+			Reflect.deleteProperty(globalThis, 'DOMMatrix');
+			mockGetText.mockResolvedValue({ text: 'Hello world', total: 1 });
+
+			await extractPdfText({
+				data: toBase64('pdf-bytes'),
+				mimeType: 'application/pdf',
+				fileName: 'doc.pdf',
+			});
+
+			expect(typeof Reflect.get(globalThis, 'DOMMatrix')).toBe('function');
+		});
+	});
 });

packages/@n8n/instance-ai/src/parsers/pdf-parser.ts

@@ -22,6 +22,13 @@ export async function extractPdfText(attachment: AttachmentInfo): Promise<PdfExt
 		throw new Error(formatSizeLimitMessage(decoded.length));
 	}
 
+	// pdf-parse v2 is backed by pdfjs-dist, which expects a `DOMMatrix` global
+	// that Node.js does not provide. Polyfill it before parsing.
+	if (typeof Reflect.get(globalThis, 'DOMMatrix') === 'undefined') {
+		const { default: DOMMatrix } = await import('@thednp/dommatrix');
+		Reflect.set(globalThis, 'DOMMatrix', DOMMatrix);
+	}
+
 	const { PDFParse } = await import('pdf-parse');
 
 	const parser = new PDFParse({ data: decoded });

pnpm-lock.yaml

@@ -839,6 +839,9 @@ importers:
       '@n8n/utils':
         specifier: workspace:*
         version: link:../utils
+      '@thednp/dommatrix':
+        specifier: ^2.0.12
+        version: 2.0.12
       https-proxy-agent:
         specifier: 'catalog:'
         version: 7.0.6
@@ -1948,6 +1951,9 @@ importers:
       '@opentelemetry/api':
         specifier: ^1.9.0
         version: 1.9.0
+      '@thednp/dommatrix':
+        specifier: ^2.0.12
+        version: 2.0.12
       csv-parse:
         specifier: 'catalog:'
         version: 6.2.1