🏠 Upgraded framework positioning: underlying architecture + high-level tools side by side
- Underlying Advantage: LangGraph has always been characterized by "low-level, no hidden logic", which is suitable for production environments. Enterprise users (e.g. Uber, LinkedIn) use it to flexibly build customized AI Agents.
- New high-level tools: Prebuilt Agents are now available, such as templates for structured data extraction, multi-Agents collaboration, etc., so that newcomers can get started quickly while retaining the freedom of underlying modifications.
💡 Version 0.3 update highlights: splits and extensions
- Functionality split: simple Agents tools that were built-in (e.g. create_react_agent) were moved to a separate package, langgraph-prebuilt, to keep the main framework simple.
- Four new pre-built Agents have been added:
- Trustcall: A Reliable Tool for Structured Data Extraction
- Supervisor: Multi-Agents Collaborative Control Architecture
- LangMem: long-term memory function
- Swarm: multiple Agents working together ("swarm" mode)
🤗 Promote a shared community ecology
- Encourage contributions: developers can follow the guidelines to create their own pre-built Agents to join the official registry, similar to the 700+ community integrations LangChain already has!
- Users benefit: they can either use off-the-shelf solutions directly or adjust them based on transparent code, or even participate in eco-construction, creating a virtuous cycle.
English 
简体中文 
日本語 
Deutsch 
Português do Brasil