Copyright	(c) 2025 Tushar Adhatrao
License	MIT
Maintainer	Tushar Adhatrao <[email protected]>
Stability	experimental
Safe Haskell	None
Language	Haskell2010

Langchain.DocumentLoader.Core

Contents

Document Representation
Loading Interface

Description

Implementation of LangChain's document loading abstraction, providing:

Document representation with content and metadata
Typeclass for loading/splitting documents from various sources
Integration with text splitting capabilities

For more information on document loader in the original Langchain library, see: https:/python.langchain.comdocsconceptsdocument_loaders/

Example usage:

-- Create a document
doc :: Document
doc = Document "Sample content" (fromList [("source", String "example.txt")])

-- Hypothetical file loader instance
data FileLoader = FileLoader FilePath

instance BaseLoader FileLoader where
  load (FileLoader path) = do
    content <- readFile path
    return $ Right [Document content (fromList [("source", String (T.pack path))])]

Test case patterns:

>>> mempty :: Document
Document {pageContent = "", metadata = fromList []}

>>> doc1 = Document "Hello" (fromList [("a", Number 1)])
>>> doc2 = Document " World" (fromList [("b", Bool True)])
>>> doc1 <> doc2
Document {pageContent = "Hello World", metadata = fromList [("a", Number 1), ("b", Bool True)]}

Synopsis

data Document = Document {
- pageContent :: Text
- metadata :: Map Text Value
}
class BaseLoader m where
- load :: m -> IO (Either String [Document])
- loadAndSplit :: m -> IO (Either String [Text])

Document Representation

data Document Source #

Document container with content and metadata. Used for storing loaded data and associated metadata like source URLs or page numbers.

Example:

>>> Document "Hello World" (fromList [("source", String "example.txt")])
Document {pageContent = "Hello World", metadata = fromList [("source",String "example.txt")]}

Constructors

Document
Fields pageContent :: Text The text content of the document metadata :: Map Text Value Additional metadata (e.g., source, page number)

Instances

Instances details

Monoid Document Source #	Monoid instance provides empty document: `>>>` `mempty :: Document`Document {pageContent = "", metadata = fromList []}
Instance details Defined in Langchain.DocumentLoader.Core Methods mempty :: Document # mappend :: Document -> Document -> Document # mconcat :: [Document] -> Document #
Semigroup Document Source #	Semigroup instance combines both content and metadata `>>>` `let doc1 = Document "A" (fromList [("x", Number 1)])``>>>` `let doc2 = Document "B" (fromList [("y", Bool True)])``>>>` `doc1 <> doc2`Document {pageContent = "AB", metadata = fromList [("x", Number 1), ("y", Bool True)]}
Instance details Defined in Langchain.DocumentLoader.Core Methods (<>) :: Document -> Document -> Document # sconcat :: NonEmpty Document -> Document # stimes :: Integral b => b -> Document -> Document #
Show Document Source #
Instance details Defined in Langchain.DocumentLoader.Core Methods showsPrec :: Int -> Document -> ShowS # show :: Document -> String # showList :: [Document] -> ShowS #
Eq Document Source #
Instance details Defined in Langchain.DocumentLoader.Core Methods (==) :: Document -> Document -> Bool # (/=) :: Document -> Document -> Bool #

Loading Interface

class BaseLoader m where Source #

Typeclass for document loading implementations. Implementations should define how to:

Load full documents with load
Load and split content with loadAndSplit

Example instance for text files:

instance BaseLoader FilePath where
  load path = do
    content <- readFile path
    return $ Right [Document content (fromList [("source", String (T.pack path))])]

  loadAndSplit path = do
    content <- readFile path
    return $ Right (splitText defaultCharacterSplitterOps content)

Methods

load :: m -> IO (Either String [Document]) Source #

Load all documents from the source.

loadAndSplit :: m -> IO (Either String [Text]) Source #

Load all the document and split them using recursiveCharacterSpliter

Instances

Instances details

BaseLoader DirectoryLoader Source #
Instance details Defined in Langchain.DocumentLoader.DirectoryLoader Methods load :: DirectoryLoader -> IO (Either String [Document]) Source # loadAndSplit :: DirectoryLoader -> IO (Either String [Text]) Source #
BaseLoader FileLoader Source #
Instance details Defined in Langchain.DocumentLoader.FileLoader Methods load :: FileLoader -> IO (Either String [Document]) Source # loadAndSplit :: FileLoader -> IO (Either String [Text]) Source #
BaseLoader PdfLoader Source #
Instance details Defined in Langchain.DocumentLoader.PdfLoader Methods load :: PdfLoader -> IO (Either String [Document]) Source # loadAndSplit :: PdfLoader -> IO (Either String [Text]) Source #