Skip to content
  • Home
  • Plugin manual
  • dsh-web-search-pro

anweat/dsh-web-search-pro

91Last commit Aug 14, 2026

dsh-web-search-pro DSH plugin

dsh-web-search-pro is an enhanced, persistent web search plugin for DeepSeek Harness (DSH). It bundles 9 tools that enable models to perform multi-engine web searches, fetch and render pages, search across 20 platforms (e.g., GitHub, Bilibili, YouTube, Reddit), and manage search history with caching. The plugin also supports custom platform definitions and login state persistence for Chinese social media sites.

browser-webskills-workflows

How to install the dsh-web-search-pro DSH plugin

dsh plugin --profile web add dsh-web-search-pro   # 自动装 dsh-browser（dependency

Copying does not run this command. Review the repository and version before installing the dsh-web-search-pro DSH plugin.

dsh-web-search-pro DSH plugin data source

dsh-web-search-pro DSH plugin snapshot date: Aug 16, 2026

discovered

What the dsh-web-search-pro DSH plugin can do

  • Multi-engine web search with RRF fusion and dual-layer caching (memory + SQLite).
  • Page fetching with fallback: Jina Reader → HTTP + rule extraction → Playwright rendering.
  • Platform search across 20 sites (GitHub, Bilibili, YouTube, V2EX, XiaoHongshu, Twitter, Reddit, Instagram, Facebook, RSS, Zhihu, Weibo, Douban, Tieba, Douyin, Kuaishou).
  • Playwright full-page screenshot, HTML, and text snapshot saving.
  • Persistent search history with filtering, replay, and export to JSON.
  • Custom platform definitions via settings.yaml (URL template, selector, optional cookie).
  • Rule-based site extraction (script-cat style) with persistent storage.
  • Dependency detection and installation tool for external binaries (gh, bili-cli, yt-dlp, etc.).
  • Cache clearing and search statistics.
  • Login state management for Chinese platforms using Playwright (save once, reuse).

Where the dsh-web-search-pro DSH plugin fits

  • AI models needing real-time web search results from multiple engines with fallback.
  • Automated data collection from social media platforms (e.g., Bilibili, Twitter, Reddit) for analysis.
  • Building a persistent knowledge base by fetching and caching web pages with snapshots.
  • Searching across Chinese platforms (Zhihu, Weibo, Douban, etc.) with authenticated browsing.
  • Custom site scraping without code changes by defining extraction rules and cookie-based sessions.

Who the dsh-web-search-pro DSH plugin is for

  • Developers using DeepSeek Harness to build AI agents with web search capabilities.
  • Researchers and analysts who need to aggregate search results from multiple sources and platforms.
  • Hobbyists who want to automate web data collection and persist it for later use.

dsh-web-search-pro DSH plugin limitations

  • Requires external dependencies (gh, bili-cli, yt-dlp, opencli, agent-reach, playwright) to be installed manually for platform-specific backends.
  • Chinese social media platforms (Zhihu, Weibo, Douban, etc.) need login state saved via Playwright; publicly accessible APIs are blocked.
  • Some features need API keys (Exa, Jina) configured via environment variables or settings.yaml.
  • The plugin is early-stage; version compatibility issues may arise with DSH versions (e.g., 0.1.0-rc.5 vs rc.6).
  • Playwright and Chromium must be installed for page rendering and screenshots.
  • Custom platform definitions require manual configuration of URL templates and selectors.

dsh-web-search-pro DSH plugin: from the repository README

Quoted from the anweat/dsh-web-search-pro README, the upstream source of the dsh-web-search-pro DSH plugin. Copyright remains with the original authors.

增强型、可持久化的扩展网页搜索插件 for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness)（DSH）。 一个 DSH **bundle 插件**，把多引擎网页搜索、平台搜索、持久化缓存、脚本猫式按站提取、Playwright 渲染打包成模型可直接调用的 9 个工具。灵感来自 [MediaCrawler](https://github.com/NanmiCoder/MediaCrawler)、[Agent-Reach](https://github.com/Panniantong/Agent-Reach)、脚本猫/油猴 userscript、opencli 与 playwright。 ## 安装 ```bash dsh plugin --profile web add dsh-web-search-pro # 自动装 dsh-browser（dependency）+ 自动挂载 browser 行（本 patch） # 或本地目录 / tarball： dsh plugin --profile web add ./dsh-web-search-pro # 重启（web profile 关闭了 HMR）： dsh --profile web ``` > dsh-browser 需先发布到 npm（本地测试可用 `dsh plugin --profile web add ../dsh-browser ../dsh-web-search-pro` 一条命令显式列两个）。 > 依赖 `@deepseek-ai/*` 已发布到 npm（`^0.1.0-rc.6`，与社区 dsh-cc-tui 一致）。 > 若你的 harness 是本地源码 checkout（如 `0.1.0-rc.5`），版本号可能有出入——用 > `dsh plugin --profile web add ./<path>` 并在 profile 的 `pnpm-workspace.yaml` > 里对齐版本后重装即可。 ## 工具（9 个） | 工具 | 作用 | |---|---| | `web_search_pro` | 多引擎搜索 + RRF 融合 + 内存/SQLite 双层缓存 + 历史 | | `web_fetch_pro` | 可读化抓取（Jina → HTTP+规则抽取 → Playwright 兜底）+ 快照缓存 | | `web_platform_search` | 20 平台：GitHub/B站/YouTube/V2EX/小红书/Twitter/Reddit/IG/FB/RSS + 知乎/微博/豆瓣/贴吧/抖音/快手（Playwright 登录态） | | `web_snap

Read the full READMERepository license: MIT

dsh-web-search-pro DSH plugin questions

How do I install dsh-web-search-pro?

Run the command `dsh plugin --profile web add dsh-web-search-pro` in your DSH environment. This will automatically install the plugin and its dependency dsh-browser. If you are using a local directory or tarball, use `dsh plugin --profile web add ./dsh-web-search-pro` instead. Remember to restart DSH with `dsh --profile web` after installation.

Do I need any API keys to use this plugin?

Yes, some search engines require API keys. You need to set `exaApiKey` and `jinaApiKey` in your `$DSH_HOME/settings.yaml` under the `web-search-pro:` section, or use environment variables `EXA_API_KEY` and `JINA_API_KEY`. The DuckDuckGo and Bing engines are free and do not need keys.

Can I search Chinese social media like Weibo or Zhihu?

Yes, but you need to save a login session first using Playwright. Run `node scripts/save-login.mjs all login-state.json` to log in once. Then set `playwright.storageStatePath` in settings.yaml. After that, the plugin will use the saved cookies to search those platforms. Note that public APIs are blocked, so browser automation is required.

What external tools are required for platform-specific searches?

The plugin depends on several external tools: `gh` for GitHub, `bili-cli` for Bilibili, `yt-dlp` for YouTube, `opencli` for XiaoHongshu/Twitter/Reddit/Instagram/Facebook, `agent-reach` for some backends, and `playwright` for rendering. You can install them via `winget`, `choco`, `pip`, `npm`, or `uv`. The plugin includes a `web_deps` tool to detect and install missing dependencies.

How can I clear the search cache or view history?

The plugin provides three dedicated tools: `web_history` to view and filter search history (with replay and export to JSON), `web_cache_clear` to clear all cached search results, and `web_search_stats` to show storage statistics. You can call these tools directly from your DSH model.