deepagents-in-action

deepagents-in-action is a Chinese-language open course on building agents with the Deep Agents framework from LangChain and LangGraph. Its 16 chapters, each with text, video and a companion experiment template, progress from a first simple agent through task planning, a virtual file system, sub-agent collaboration and context management.

The course tracks the Deep Agents framework's own version changes rather than describing a single frozen release: it explains why a version's default behavior changed, such as a lighter and more transparent harness in a later version that cuts baseline input tokens for simple turns, and how to verify a migration using evaluations and execution traces before upgrading. A companion project-scaffolding tool sets up each chapter's runnable environment from a template, and two accompanying AI coding-assistant skills give guidance specific to the LangChain framework and to reading execution traces from LangSmith.