DeepSeek Harness v0.1 Open-Source Developer Preview Released
DeepSeek reportedly released the DeepSeek Harness v0.1 developer preview, abbreviated dsh, on August 13, 2026, alongside open-source code. The Juejin AI article describes Harness as an agent runtime layer that connects models with tools, sessions, permissions, storage, and interfaces, rather than as a new DeepSeek large model. Its architecture includes a replaceable model provider and plugin-based tools, but the report does not independently verify compatibility with other model providers.
The release puts DeepSeek’s agent execution and engineering layer into open source, covering the infrastructure around tools, sessions, permissions, storage, and recovery rather than model capabilities alone.
2026 年 8 月 13 日,DeepSeek 发布 DeepSeek Harness v0.1 开发者预览版(简称 dsh),同时开源源码。 看完博主「轩辕的编程宇宙」的《DeepSeek Harness 技术架构教程》第一课,我的第一反应是:这套东西的重点根本不在「DeepSeek」。原文链接:DeepSeek Harness 到底是什么?|第 1 课
一、把「DeepSeek」这个词的重量降下来
很多人第一眼会把重音放在「DeepSeek」上,以为是 DeepSeek-R1、DeepSeek-V4 之后的新模型。原文第一句就拆掉了这个误会:它和那些大模型由同一家公司推出,却不是同一种产品。
| 负责什么 | |
|---|---|
| DeepSeek 大模型 | 理解输入、推理、生成文字,或表达「我想调用某个工具」 |
| DeepSeek Harness | 把模型接入工具、会话、权限、存储和界面,让任务持续执行 |
这套分工可以用一句话概括:Agent = Model + Harness。而且,Harness 里的「DeepSeek」更像出品方署名——模型由可替换的 Provider 提供,只要有适配器,也能接别的模型。