A Document is a LangChain data structure that keeps source text together with associated metadata. Loaders create Document objects from files and other data sources, giving later retrieval steps a consistent input. section 10.5.2
Use Documents as the working unit in a retrieval pipeline:
- Load text files with DirectoryLoader and TextLoader. section 10.5.2
- Split large Documents into smaller Documents with RecursiveCharacterTextSplitter, using chunk sizes and overlap to preserve useful context. section 10.5.2
- Store the resulting Documents in a vector store, then retrieve relevant Documents through similarity search. section 10.5.2
Proper care ensures your InkStream and RoyalQuill fountain pens write smoothly for years.
- Cleaning: Flush the nib with warm water every few weeks.
- Refilling: Use high-quality ink to prevent clogging.
- Storage: Store pens upright to avoid leaks and ensure ink flow.
Treat the supplied care instructions as source text: a loader can turn this material into a Document before it is chunked and indexed. section 10.5.2
