danielmiessler/Fabric · warning · Error

${fileName}: the conversion returned no text.

Error message

${fileName}: the conversion returned no text.

What it means

interpretPdfResult throws when the PDF is not classified as scanned/image-based but result.markdown is empty or whitespace-only — the inspector found a text layer yet extracted zero usable text. This is a data-quality outcome of the WASM conversion, not a crash.

Source

Thrown at web/src/lib/services/PdfConversionService.ts:18

import type { PdfProcessResult } from '@firecrawl/pdf-inspector-wasm';
import type { PdfRequest, PdfResponse } from '../workers/pdf-inspector.worker';

export interface PdfConversion {
	markdown: string;
	warning?: string;
}

// Maps the raw worker result to UI behavior. Throws when there is no usable text.
export function interpretPdfResult(result: PdfProcessResult, fileName: string): PdfConversion {
	if (result.pdfType === 'Scanned' || result.pdfType === 'ImageBased') {
		throw new Error(
			`${fileName} contains no machine-readable text. OCR is necessary, and the Fabric web interface does not include OCR.`
		);
	}
	const markdown = result.markdown;
	if (!markdown || markdown.trim().length === 0) {
		throw new Error(`${fileName}: the conversion returned no text.`);
	}
	const warnings: string[] = [];
	if (result.pdfType === 'Mixed') {
		warnings.push(
			`${fileName}: pages ${result.pagesNeedingOcr.join(', ')} contain no machine-readable text. OCR is necessary for those pages, and their content is not included.`
		);
	}
	if (result.hasEncodingIssues) {
		warnings.push(`Some text in ${fileName} did not decode correctly.`);
	}
	return warnings.length > 0 ? { markdown, warning: warnings.join(' ') } : { markdown };
}

export class PdfConversionService {
	private worker: Worker | null = null;
	private nextId = 1;
	private pending = new Map<
		number,

View on GitHub (pinned to 338b89cfe9)

Solutions

1. Try re-exporting or re-saving the PDF from another tool (reprints a sane text layer)
2. If the file is effectively image-based despite classification, OCR it externally as in the scanned case
3. Report files where classification says text-bearing but extraction is empty — that is an inspector gap

if (!result.markdown || result.markdown.trim().length === 0) {
  // classify as 'unusable text layer' before interpretPdfResult throws
}

function isEmptyConversionError(e: unknown): boolean {
  return e instanceof Error && e.message.includes('returned no text');
}

try { conversion = interpretPdfResult(result, fileName); }
catch (e) {
  if (isEmptyConversionError(e)) { toast.warn('This PDF has no extractable text. Re-export or OCR it.'); return; }
  throw e;
}

When it happens

Trigger: PDFs whose 'text' is only invisible/empty glyph runs; fonts with broken ToUnicode maps yielding whitespace; a text layer consisting solely of whitespace or control characters.

Common situations: Designer PDFs with outlined/curved text; PDFs where text is drawn as vector paths; partially corrupt files that parse but yield nothing.

Related errors

• ${fileName} contains no machine-readable text. OCR is necess
• Invalid response format: missing vendors data
• Invalid session file format
• Pattern validation failed: {validation_message}

AI-assisted analysis of danielmiessler/Fabric@338b89cfe9 (2026-08-15). Data as JSON: /api/errors/bfb9c8c546b9078c. Report an issue: GitHub.