Skip to main content

Class: PdfLoader

Defined in: packages/agentos/src/cognition/memory/io/ingestion/PdfLoader.ts:93

Document loader for PDF files.

Extraction tiers​

  1. unpdf — always used as the primary extraction engine. Performs pure-JS PDF text layer extraction with no native binaries required.
  2. OCR fallback (optional) — supplied at construction time and engaged automatically when unpdf yields sparse text (< 50 chars per page on average), indicating a scanned document.
  3. Docling fallback (optional) — when provided, takes precedence over both unpdf and OCR, yielding the highest-fidelity extraction at the cost of requiring a Python runtime.

Implements​

Example​

const ocrLoader    = createOcrPdfLoader();   // null if tesseract.js absent
const doclingLoader = createDoclingLoader(); // null if docling absent
const pdfLoader = new PdfLoader(ocrLoader, doclingLoader);
const doc = await pdfLoader.load('/reports/q3.pdf');

Implements​

Constructors​

Constructor​

new PdfLoader(ocrLoader?, doclingLoader?): PdfLoader

Defined in: packages/agentos/src/cognition/memory/io/ingestion/PdfLoader.ts:116

Creates a new PdfLoader.

Parameters​

ocrLoader?​

Optional OCR fallback (for example from createOcrPdfLoader()).

IDocumentLoader | null

doclingLoader?​

Optional Docling loader (for example from createDoclingLoader()).

IDocumentLoader | null

Returns​

PdfLoader

Properties​

supportedExtensions​

readonly supportedExtensions: string[]

Defined in: packages/agentos/src/cognition/memory/io/ingestion/PdfLoader.ts:95

File extensions this loader handles, each with a leading dot.

Used by LoaderRegistry to route file paths to the correct loader.

Example​

['.md', '.mdx']

Implementation of​

IDocumentLoader.supportedExtensions

Methods​

canLoad()​

canLoad(source): boolean

Defined in: packages/agentos/src/cognition/memory/io/ingestion/PdfLoader.ts:129

Returns true when this loader is capable of handling source.

For string sources the check is purely extension-based. For Buffer sources the loader may inspect magic bytes when relevant.

Parameters​

source​

Absolute file path or raw bytes.

string | Buffer

Returns​

boolean

Implementation of​

IDocumentLoader.canLoad

load()​

load(source, options?): Promise<LoadedDocument>

Defined in: packages/agentos/src/cognition/memory/io/ingestion/PdfLoader.ts:148

Parses source and returns a normalised LoadedDocument.

When source is a string the loader treats it as an absolute (or resolvable) file path and reads the file from disk. When source is a Buffer the loader parses the bytes directly and derives as much metadata as possible from the buffer content alone.

Parameters​

source​

Absolute file path OR raw document bytes.

string | Buffer

options?​

LoadOptions

Optional hints such as a format override.

Returns​

Promise<LoadedDocument>

A promise resolving to the fully-populated LoadedDocument.

Throws​

When the file cannot be read or the format is not parsable.

Implementation of​

IDocumentLoader.load