LangChain details Deep Agents context management
What happened
On October 15, LangChain described how Deep Agents manage context through middleware, combining file offloading, summarization, on-demand skill loading, subagents and long-term memory to reduce context buildup. Tool results over about 20,000 tokens and user messages over 50,000 tokens are saved to files. When context use hits the default 85% threshold, older messages are compressed into a summary while the original record stays readable. The write-up focuses on keeping information accessible without leaving everything in the context window.
Written by AI from the coverage · updated 2 hours ago
Coverage
Follow the reports to see the story from different sides.
- AI HOT · Tips & opinionsLangChain 详解 Deep Agents 如何记住信息而不塞满上下文窗口
LangChain 详解 Deep Agents 如何通过中间件管理上下文，结合文件卸载、摘要、技能按需加载、子智能体和长期记忆减少上下文堆积。工具结果超过约 20,000 tokens、用户消息超过 50,000 tokens 时会保存到文件；上下文使用率达到默认 85% 阈值时，旧消息被压缩为摘要，原始记录仍可读取。
Heat over time
Not enough continuous observations to draw a trend yet.