How Deep Agents Manages Context: Offloading and Summarization

Summary

This video details 'context rot,' the degradation of an AI agent's reasoning and accuracy due to excessive context accumulation. Deep Agents, an open-source agent harness, addresses this through two primary context engineering techniques: offloading and summarization. Offloading intercepts massive inputs (like tool results or long documents) by saving them to a virtual file system and only providing the model with a preview. Summarization, conversely, runs before model calls to condense the entire context window, keeping the model focused while preserving the full history in a summarized format.

Technical details

• Deep Agents Architecture 29s

  Deep Agents is an open-source agent harness built with context engineering techniques, including offloading, summarization, skills, subagents, and long-term memory, designed for long-running tasks. (0:29)

• Offloading Thresholds 133s

  Tool results over 20,000 tokens are offloaded to `large_tool_results`. Human messages over 50,000 tokens are offloaded to `conversation_history`. (1:33)

• Summarization Thresholds 237s

  Summarization is triggered by default when token usage reaches 85%. The check runs before every model call, and the process involves summarizing content above the 'keep cutoff' line. (2:37)

Mentioned resources

• Introduction to Deep Agents (free LangChain Academy course) (Course)
• Context engineering in Deep Agents (docs) (Documentation)
• Context Management for Deep Agents (blog) (Blog Post)

Channel & topics

• LangChain
• Large Language Models (LLMs)
• Agent Architecture
• Context Management
• Context Rot
• LangChain
• Deep Agents
• Introduction to Deep Agents (free LangChain Academy course)
• Context engineering in Deep Agents (docs)
• Context Management for Deep Agents (blog)