dsh-llm-wiki DSH plugin

This plugin is a thin adapter layer that exposes 7 tools for DSH agents to interact with LLM-Wiki knowledge bases. It delegates all retrieval and compilation logic to the DSH-Wiki engine via subprocess, requiring no runtime dependencies beyond Node >= 18 and Python >= 3.10. Each tool returns structured JSON results for agent consumption.

developer-tools skills-workflows

How to install the dsh-llm-wiki DSH plugin

dsh plugin --profile web add @detpecca/dsh-llm-wiki

Copying does not run this command. Review the repository and version before installing the dsh-llm-wiki DSH plugin.

dsh-llm-wiki DSH plugin data source

dsh-llm-wiki DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-llm-wiki DSH plugin can do

wiki_search: structured signal retrieval with CJK tokenization, returning weighted hits (page name, alias, tag, summary, body).
wiki_read: batch read pages by wiki path, follow wikilinks, return page contents or error indicators.
wiki_stats: get page/category/digest/error book counts.
wiki_validate: run 4 deterministic structural checks (dangling links, incomplete pages, malformed source references, index inconsistency).
wiki_fix: perform deterministic repairs (rebuild indexes, fill backlinks) without LLM; with finalize:true, run 3-round LLM-assisted fix and error book verify.
wiki_errorbook: view the error book entries (self-correction records).
wiki_ingest: compile source text into the wiki using the full algorithm 1 pipeline (requires LLM key).

Where the dsh-llm-wiki DSH plugin fits

Add a new note file to the knowledge base and immediately validate its structure.
Ask agent to search for information about a topic and retrieve the relevant pages with supporting evidence.
Run a health check on the wiki structure and let the agent fix any issues automatically.
Review the error book to understand recurring problems and track what has been corrected.

Who the dsh-llm-wiki DSH plugin is for

DeepSeek Harness users who want to manage a personal LLM-Wiki knowledge base through agent conversations.
Developers or researchers using LLM-Wiki for structured knowledge retrieval who need agent-driven maintenance.

dsh-llm-wiki DSH plugin limitations

Requires Python >= 3.10 and the DSH-Wiki engine installed separately (pip install git+https://github.com/detpecca/DSH-Wiki.git).
wiki_ingest and wiki_fix finalize require an OpenAI-compatible API key (set via config or environment variable).
The plugin is a thin adapter; all core logic depends on the DSH-Wiki engine CLI. No offline fallback for complex operations.
Subprocess output is capped at 2MB stdout / 200KB stderr; large ingest or search may hit limits.

dsh-llm-wiki DSH plugin: from the repository README

Quoted from the detpecca/dsh-llm-wiki README, the upstream source of the dsh-llm-wiki DSH plugin. Copyright remains with the original authors.

DeepSeek Harness 插件：让 agent 直接管理 **LLM-Wiki** 个人知识库——检索、阅读、统计、校验、修复、错误本、入库，共 7 个工具。 `@detpecca/dsh-llm-wiki` 是一个**薄适配层**，不重复实现任何检索/编译逻辑： ``` DSH agent ──调用工具──▶ 本插件（ToolDefinition × 7） │ ctx.subprocess.spawn ▼ python -m llm_wiki --wiki <路径> <子命令> --json │ ▼ DSH-Wiki 引擎（检索打分 / 算法 1 编译 / 校验修复的唯一权威） │ ▼ 你的 Wiki 目录 ``` - 每个工具通过 DSH 的 `subprocess` 服务调用 [DSH-Wiki](https://github.com/detpecca/DSH-Wiki) 引擎 CLI 的 `--json` 通道； - Wiki 自己的**结构化信号检索**（页名/别名/标签/摘要加权打分）和**论文算法 1 编译流程**保持唯一权威； - 插件本体**零运行时依赖**（纯 ESM，无构建步骤），Node ≥ 18。 ## 前置要求 1. 已安装 [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)（`dsh` 命令可用，或有其仓库 checkout）； 2. Python ≥ 3.10（供引擎使用）； 3. （可选）一个 OpenAI 兼容的 LLM API key——只有 `wiki_ingest` 和 `wiki_fix {finalize:true}` 需要。 ## 🚀 快速开始 **两条命令，不需要克隆任何仓库。** ### 1. 安装 Python 引擎（DSH-Wiki） ```bash pip install git+https://github.com/detpecca/DSH-Wiki.git # 唯一运行时依赖是 pyyaml ``` ### 2. 安装本插件 ```bash dsh plugin --profile web add @detpecca/dsh-llm-wiki ``` > 还没发布 npm 时

Read the full READMERepository license: MIT

dsh-llm-wiki DSH plugin questions

How do I install the dsh-llm-wiki plugin?

First install the DSH-Wiki Python engine: pip install git+https://github.com/detpecca/DSH-Wiki.git. Then install the plugin with: dsh plugin --profile web add @detpecca/dsh-llm-wiki. If not published on npm, use: dsh plugin --profile web add github:detpecca/dsh-llm-wiki. After installation, configure the wiki path in cordis.patch.yml.

Does the plugin require an API key for all tools?

No, only wiki_ingest and wiki_fix with finalize:true need an OpenAI-compatible API key. Search, read, stats, validate, errorbook, and basic fix do not require any key. You can configure the key via cordis.patch.yml or environment variables (LLM_WIKI_API_KEY, etc.).

What should I do if I get 'No module named llm_wiki'?

This means the DSH-Wiki engine is not installed in the Python interpreter specified by pythonPath. Make sure you ran pip install git+https://github.com/detpecca/DSH-Wiki.git in the correct Python environment. You can also check the error message for additional hints.

How do I point the plugin to my existing wiki directory?

Edit the cordis.patch.yml file in your DSH profile directory (usually ~/.dsh/profiles/<name>/cordis.patch.yml). Add an entry with id: llm-wiki and set wikiPath to the absolute path of your wiki root (the directory containing index.md). Restart DSH for changes to take effect.

Can I use the plugin to fix structural issues in my wiki?

Yes, call wiki_validate to detect issues, then wiki_fix to perform deterministic repairs (rebuild indexes, fill backlinks). If the problems persist, run wiki_fix with finalize:true (requires LLM key) for a deeper LLM-assisted fix cycle. After fixing, run wiki_validate again to confirm.