Ted Hisokawa
August 16, 2024, 3:49 p.m.
LangChain revises integration documents with a standardized format and improved API references for faster information retrieval.
A big part of the LangChain ecosystem is its extensive collection of integrations. LangChain offers over 1,000 integrations for LLMs, vector stores, tools, document loaders, and more. Today, the LangChain team announced a major overhaul of its integration documentation in Python and JavaScript to make it more useful and accessible to the community, according to the LangChain blog.
Standardized content for all integration pages
Over the past year and a half, the LangChain community has contributed over 1,000 open source integrations, including chat models, vector stores, tools, and retrievers. As the number of integrations has grown and best practices have evolved, many documentation pages have become out of date.
Key integrations now follow a standardized template that highlights common features for each category (e.g., models, vector stores, retrievers). For example, chat model pages start with a table showing whether an integration supports features like tool calls and multimodal input, followed by installation and basic invocation examples.
With these redesigned integration pages, we want to help developers quickly understand what an integration can do and how it is used.
While some advanced, integration-specific examples remain on these pages, more emphasis has been placed on linking to tutorials and API references to keep the content up to date and avoid repetition.
New index pages for optimized search
To make it easier for developers to find the integrations they need, LangChain has optimized the index pages for each type of integration. Combined with a smaller sidebar, these index pages now contain tables similar to those on the individual integration pages, making it easy to quickly identify integrations with the desired functionality.
These “features” tables are currently sorted by a combination of factors, including usage in LangSmith traces and package downloads. LangChain plans to explore more ways to highlight and showcase emerging integrations in the future.
Improved API references
The new documentation pages feature improved API references for Python and JavaScript.
For Python, more explanations and usage examples have been added to the docstrings. The structure and formatting have been updated to be more modern and user-friendly, including a navigable sidebar with methods and attributes for all classes.
For JavaScript, the API reference pages have been made less intimidating by hiding the sidebar by default and filtering out less relevant methods and other build artifacts. Popular chat model and vector storage pages have been enhanced with various usage examples, and visibility of useful runtime and constructor definitions and important methods has been generally improved.
The goal of this ongoing work is to make the API references standalone, valuable resources for the LangChain community.
Image source: Shutterstock
