LangChain's new Indexing API makes it easy to load and keep in sync documents from any source into a vector store, avoiding duplicated content and re-computing embeddings over unchanged content. It uses a record manager to track document writes, hashes are computed for each document, and information such as write time and source id is stored in the record manager. Cleanup modes allow users to pick the behavior they want when re-indexing documents into a vector store. The API supports various vector stores, including ElasticsearchStore, and can be used to index documents with incremental changes.