Skip to main content
Version: 0.0.5 (stable)

agrag.loaders

Document loaders: turn files, directories and raw text into Documents.

The docling loaders live in agrag.loaders.docling. PDF and image files need the docling extra.

Modules:

  • base – The Loader interface: reads one source and yields Document objects.
  • chat – Chat reader: JSON Lines or JSON messages, one section for each message.
  • common – Shared helpers for the corpus readers.
  • decode – The four-step decode pipeline for source bytes.
  • defaults – The loaders that the corpus registry holds by default.
  • docling – The docling loaders.
  • errors – Errors that the ingestion layer raises.
  • loader_registry – The extension-to-loader registry.
  • prose – Prose readers: plain text and XML.
  • records – Record readers: CSV, TSV, JSON Lines, and JSON.
  • types – Plumbing types for the document loaders.
  • walk – Corpus walk, batching, and resumable streaming.

Classes:

  • ChatLoader – Reads a file of chat messages as one document with a section for each message.
  • CsvLoader – Reads CSV and TSV files as one record document per row.
  • DecodeError – The source bytes do not decode to text.
  • DocumentConversionError – A loader could not parse or convert a source's content.
  • DocumentTooLargeError – A prose source is larger than the configured byte limit.
  • ErrorPolicy – The action to take when one source in a batch fails.
  • IngestResult – The result of one ingest call.
  • IngestionError – The base class for every ingestion error.
  • JsonLoader – Reads JSON files, disambiguating arrays from objects.
  • JsonlLoader – Reads JSON Lines files as one document per line.
  • LoadStats – Running tally of a corpus walk.
  • LoaderRegistry – Maps each source extension to the one loader that reads it.
  • MalformedRecordError – One record in a record-family source does not parse.
  • MissingExtraError – A loader exists for this format, but its package extra is not installed.
  • ReadOptions – Per-source reader configuration.
  • TextLoader – Reads plain-text and log files as one document each.
  • UnsupportedFormatError – No registered loader can read this source's format.
  • XmlLoader – Reads an XML file as the text of its elements.

Functions:

Attributes: