[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-deepseek-harness-plugin-stack":3,"news-related-63c30bcd-3ffc-47c5-bd74-c2a9ed8f7c94":41},{"id":4,"title":5,"summary":6,"content":7,"original_url":8,"source_id":9,"tags":10,"translations":27,"news_slug":34,"published_at":35,"created_at":36,"modified_at":37,"is_published":38,"publish_type":39,"image_url":14,"view_count":40},"63c30bcd-3ffc-47c5-bd74-c2a9ed8f7c94","DeepSeek Harness 预览版开源:Agent 被拆成可插拔的插件栈,模型只负责想、Harness 负责做事","DeepSeek 发布 Harness 开发者预览版,把 Agent 拆成可插拔的插件栈,所有能力(模型、工具、技能、会话、沙箱、调度、UI)都能替换或重组;学术基础来自 arXiv 2608.25512 的 Cordis 论文。","最近 Agent 工程化的热度从「模型能力」滑向「模型怎么在真实环境里不被卡住」。DeepSeek 在 9 月初把这件事做成了开源:把 Agent 拆成 Model + Harness 两层,模型负责想,Harness 负责调用工具、管理会话、对接沙箱、调度子代理。\n\n## 一切皆插件的 Cordis 内核\n\nHarness 代码已在 GitHub 开源(github.com\u002Fdeepseek-ai\u002Fdeepseek-harness),核心是 Cordis 内核——把\"插件挂载、卸载、依赖管理\"做成基本动作的元框架。每个能力(models、tools、skills、sessions、sandboxes、storage、loops、scheduling、UI)都是 Cordis 插件,通过配置就能换或重组,不用改源码。\n\n学术基础是 arXiv 2608.25512(「A Programming Paradigm for Spatiotemporal Composability」,8 月 26 日提交,92 页),作者包括 DeepSeek-AI 与北大。论文提出两个正交维度:时间可组合(移除组件能完整回滚副作用)和空间可组合(声明并响应组件间依赖),用统一的\"上下文类型\"做中介,实现组件间安全交错。\n\n## 四种 Runtime Mode\n\n- **Standard**:完整 coding agent,文件编辑、shell、Web 搜索、skills、规划、子代理、workflows 全开,是默认形态。\n- **Code**:工具通过 Code Mode SDK 暴露成 TypeScript,模型把多步操作合并成一段代码执行,减少工具往返。\n- **Minimal**:只剩 bash + str_replace_editor 两个工具,给\"最小环境 benchmark 模型\"用。\n- **Creator**:Standard 之上加运行时检查、插件实验、preset 编辑,给做自定义 Agent 预设的开发者。\n\n四档位是 DeepSeek 对\"Agent 应该多复杂\"的回答——从 benchmark 用的最小壳,到生产环境的全套工具链,再到给构建者用的元层。\n\n## 可观测性是隐含卖点\n\n每个 run 进只追加的 session log:系统提示、推理、工具调用、子代理调度、上下文注入,按来源可检索;Trajectory 视图支持在同一份事件流上 resume、fork、搜索、replay。这把 Agent 的\"思考—行动\"轨迹做成可重放的结构化日志,对调模型行为、做离线评测、做训练数据回灌更友好。\n\n## 所以呢?\n\nDeepSeek 不是第一个做开源 Agent harness 的(参考 Databricks Omnigent、Anthropic SDK、LangChain 风格框架),但它是少数把 Agent = Model + Harness 这条线喊得最清楚、把\"插件组合\"做成一等公民、又把可观测性默认开在底层的实现。配合 arXiv 论文对可组合性的形式化,Agent 卡在环境交互层的团队,去看 Cordis 的 effect\u002Fcoeffect 设计,会比从零搭框架更省力。\n\n引用:[DeepSeek Harness](https:\u002F\u002Fwww.deepseek.com\u002Fharness\u002Fen\u002F)、[arXiv:2608.25512](https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.25512)、[GitHub 仓库](https:\u002F\u002Fgithub.com\u002Fdeepseek-ai\u002Fdeepseek-harness)","https:\u002F\u002Fwww.deepseek.com\u002Fharness\u002Fen\u002F","4194681c-1a38-405d-a917-40e1dc2622ea",[11,15,18,21,24],{"id":12,"name":13,"slug":13,"description":14,"color":14},"6ad31a14-c0da-42df-81fd-564281f768db","agentic-ai",null,{"id":16,"name":17,"slug":17,"description":14,"color":14},"e82b2d09-81b2-43d1-977e-e018443b3c14","coding-agent",{"id":19,"name":20,"slug":20,"description":14,"color":14},"b52db7e9-7c58-42c3-9536-5132cb2f8f72","deepseek",{"id":22,"name":23,"slug":23,"description":14,"color":14},"01598627-1ea6-4b27-a5d8-874971571a71","llm",{"id":25,"name":26,"slug":26,"description":14,"color":14},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[28],{"id":29,"lang":30,"title":31,"summary":32,"content":33},"e42b3b17-490f-4f71-a424-12def5fc073c","en","DeepSeek Harness goes open source: Agent as a pluggable plugin stack","DeepSeek released the Harness developer preview, splitting the Agent into a pluggable plugin stack. Every capability (model, tools, skills, sessions, sandboxes, scheduling, UI) can be swapped or recomposed. The academic foundation is the Cordis paper at arXiv 2608.25512.","The center of gravity in Agent engineering is sliding from \"model capability\" to \"how the model does not get stuck in real environments.\" DeepSeek packaged that idea into open source in early September: split the Agent into Model + Harness, where the model thinks and the Harness handles tool calls, session management, sandbox integration, and sub-agent scheduling.\n\n## Everything is a plugin: the Cordis kernel\n\nThe Harness code is open on GitHub (github.com\u002Fdeepseek-ai\u002Fdeepseek-harness). Its core is the Cordis kernel, a meta-framework that turns plugin mounting, unmounting, and dependency management into primitive operations. Every capability (models, tools, skills, sessions, sandboxes, storage, loops, scheduling, UI) is a Cordis plugin; you can swap or recompose anything in config, no source changes needed.\n\nThe academic foundation is arXiv 2608.25512 (\"A Programming Paradigm for Spatiotemporal Composability\", submitted August 26, 92 pages), authored by DeepSeek-AI and Peking University. The paper formalizes two orthogonal composition dimensions: temporal composability (a component's side effects can be fully reverted on removal) and spatial composability (inter-component dependencies can be declared and reactively managed). A unified \"context type\" mediates effects and coeffects, giving components safe interleaving without disturbing each other.\n\n## Four runtime modes, pick the Agent by use case\n\n- **Standard**: a full coding agent — file editing, shell, file and Web search, skills, planning, goals, sub-agents, workflows — all on. This is the default.\n- **Code**: tools are exposed as TypeScript via the Code Mode SDK; the model composes multi-step operations into a single program, reducing tool roundtrips.\n- **Minimal**: only bash and str_replace_editor remain — built for \"benchmark models in a minimal environment.\"\n- **Creator**: built on Standard plus runtime inspection, plugin experiments, and preset authoring — for developers building custom Agent presets.\n\nThese four tiers are DeepSeek's answer to \"how complex should an Agent be\" — from a minimal shell for benchmarks, to a full production toolchain, to a meta-layer for builders.\n\n## Observability as the hidden sell\n\nEvery run enters an append-only session log: system prompts, reasoning, tool calls and results, sub-agent scheduling, context injections — every step indexed by source. The Trajectory view supports resume, fork, search, and replay over the same event stream. This turns the Agent's \"think—act\" trajectory into a replayable structured log, which makes model behavior debugging, offline evaluation, and training-data backfill noticeably easier.\n\n## So what?\n\nDeepSeek is not the first to ship an open-source Agent harness (think Databricks Omnigent, the Anthropic SDK, LangChain-style frameworks), but it is one of the few that articulates Agent = Model + Harness most clearly, treats \"plugin composition\" as a first-class citizen, and ships observability on by default. Combined with the formal treatment of composability in the arXiv paper, teams whose Agents get stuck at the environment-interaction layer should look at Cordis's effect\u002Fcoeffect design before reinventing the wheel.\n\nReferences: [DeepSeek Harness developer preview](https:\u002F\u002Fwww.deepseek.com\u002Fharness\u002Fen\u002F), [arXiv:2608.25512](https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.25512), [GitHub: deepseek-ai\u002Fdeepseek-harness](https:\u002F\u002Fgithub.com\u002Fdeepseek-ai\u002Fdeepseek-harness)","deepseek-harness-plugin-stack","2026-09-05T06:00:00Z","2026-09-05T07:08:17.701224Z","2026-09-05T07:08:17.701232Z",true,"agent",25,{"items":42},[43,48,53,58,63,68],{"id":44,"title":45,"news_slug":46,"published_at":47},"c94766df-827e-4e4e-a006-b6639ec76722","DeepSeek V4-Flash-0731 转正观察:权重不动,后训练把 Agent 分数打到 V4-Pro 之上","deepseek-v4-flash-0731-agent-benchmark-official-aug2026","2026-08-01T02:00:00+00:00",{"id":49,"title":50,"news_slug":51,"published_at":52},"e3d0e837-43fc-49a6-b533-fbb75aff95d3","FastContext 4B-30B 子代理：把 60% 推理 Token 留给\"找代码\"，Coding Agent 提速 5.5%","fastcontext-msr-sub-agent-code-search-5-5pct","2026-06-20T02:01:00+00:00",{"id":54,"title":55,"news_slug":56,"published_at":57},"c77381d9-29ba-45ff-89df-855d11f90de2","Terminal-Universe:Qwen 把旧轨迹反向重建为 3.73 万个环境,27B 微调双基准 +11.9\u002F+13.8 分","terminal-universe-trajectory-environments","2026-09-04T17:10:00+00:00",{"id":59,"title":60,"news_slug":61,"published_at":62},"b1645fba-d364-47e6-97da-06868f98d987","Linux 内核 7.x 每版近 2000 个 CVE:AI 帮倒忙,维护者不堪重负","linux-kernel-cve-ai-overwhelmed","2026-09-04T00:00:00+00:00",{"id":64,"title":65,"news_slug":66,"published_at":67},"14a7f5ab-e270-461c-b862-4bde139e463f","HarnessDev 基准:让 LLM 自建 Agent Harness,代码领域仍输人类工程师","harnessdev-llm-selfbuilt-agent-harness","2026-09-03T19:10:00+00:00",{"id":69,"title":70,"news_slug":71,"published_at":72},"089195fb-7fe5-4ba9-a4bc-8e356fe5e923","BAAI把1000个GitHub仓库蒸馏成5000个技能,科研agent奖牌率31%冲到73%","baai-disco-repo-to-skill-library","2026-09-03T17:07:35+00:00"]