dsh-obsidian
在终端中运行以下命令:
dsh plugin install Harzva/dsh-obsidian
将以下提示词粘贴到 DeepSeek Harness 对话框中:
在终端运行 dsh plugin install Harzva/dsh-obsidian 即可安装,插件源码位于 https://github.com/Harzva/dsh-obsidian
插件介绍
Obsidian's graph view is beautiful, but it only indexes .md files and [[wiki links]] inside your vault. It will never walk through your code repositories on its own. When you juggle a dozen local projects, hand-crafting a note and dependency edge for each one is tedious and easy to get wrong. dsh-obsidian turns that into two clicks in the DSH settings tab: hit Scan Projects, then Generate Knowledge Graph. The plugin auto-detects projects via package.json, Cargo.toml, go.mod, pyproject.toml, and .git, extracts dependencies, tech stack, root-directory areas, and keyword topics, and builds every edge with deterministic rules. No invented associations.
The output is a set of native .md notes that Obsidian indexes the moment you open the vault. Hub files under graph/, tech/, areas/, and topics/ become cluster centroids in the graph view; dependencies render as connecting edges; and graph/index.md ties the whole structure together. Transport is dual-mode: Obsidian's Local REST API (loopback, Bearer auth) when the app is running, or a plain-filesystem fallback when it is not. Incremental writes via a dedicated state file mean only plugin-managed files are ever rewritten or removed. Everything runs fully offline, and no data ever leaves your machine. A mini force-directed dependency preview sits right in the DSH settings tab for a quick sanity check.
If you lean on Obsidian as your second brain and simultaneously maintain multiple local codebases, dsh-obsidian fills the missing bridge from repository manifests to a living, link-rich knowledge graph without ever phoning home.
使用场景
- 将多个本地代码项目的依赖与技术栈自动转译为 Obsidian wiki 笔记,图谱视图一键呈现知识网络
- 在 DSH 设置页完成扫描与生成两步操作,无需手动编辑 Markdown 或维护索引文件
- Obsidian 未运行时通过文件系统回退通道增量写入 vault,全程离线且数据不出本机
适合人员
- 以 Obsidian 为第二大脑、希望把代码上下文纳入知识网络的开发者
- 同时维护多个本地仓库(npm / Cargo / Go / Python)的工程师
- 偏好全离线工具链、要求数据不离开本机的隐私敏感用户