How-to

Databricks 工程师分享:给智能体的工作文件也做分支

Why we picked it

Databricks 的工程师在 Code Summit 上讲了个实用招:agent 分头调查时,把 notes、scripts 这些工作文件也整套分支,像 git 一样管理。做 coding agent 或多分支工作流的开发者可以直接抄这套做法。

Databricks 的 ankit_math 在 AI Engineer Code Summit 上分享了一套智能体工作流的分支思路:当 agent 开出多条调查路径时,notes、plans、scripts 等工作文件也应随之分支,每条 investigation 持有自己独立的工作状态副本。

Image · AI Engineer
Body

核心做法是把整个工作目录当作可分支的状态,类似 git branch 的机制应用到 agent 的文件系统上。这样不同调查分支互不污染,回溯或合并时也更可控。

对构建多路径探索、并行实验类智能体的开发者来说,这是一套可以直接借鉴的工作状态管理方法。

Original · AI Engineer

Your agent branches. Its files need to branch with it. Notes, plans, scripts: each investigation needs its own copy. @ankit_math of @databricks on branching the entire working state. From AI Engineer Code Summit.

Industry Coordinates & Source Verification

This report is categorized under AITOP's "How-to" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Agent, agents, MCP and tools, coding assistants.

Source attribution: curated from AI Engineer (@aiDotEngineer).