Mr-remon219/dsh-search-boost
80Last commit Aug 15, 2026
dsh-search-boost DSH plugin
dsh-search-boost is a DSH plugin that enhances the built-in web_search and registers a family of search tools. It provides free-by-default parallel engines, a fused ranking system, X/Twitter search, deep research, and parallel multi-agent research. The plugin also injects a proactive-search policy into the system prompt.
How to install the dsh-search-boost DSH plugin
dsh plugin --profile web add dsh-search-boost # latestCopying does not run this command. Review the repository and version before installing the dsh-search-boost DSH plugin.
dsh-search-boost DSH plugin data source
dsh-search-boost DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-search-boost DSH plugin can do
- Upgrades built-in web_search and web_fetch with multi-engine chain and Jina-based page reader.
- Provides fused_search for multi-engine retrieval with co-occurrence scoring and time-decay freshness.
- Offers x_search for real-time X/Twitter search with credential-free fallback chain.
- Supports deep_research step-mode driven by the main agent with coverage analysis and gaps.
- Enables research_parallel for sub-query decomposition and fan-out to native DSH subagents.
- Includes /web_change command to switch between free (keyless) and api (keyed) search layers.
- Provides search_stats command for cache and engine availability audit.
Where the dsh-search-boost DSH plugin fits
- Enhance default web_search with multiple free search engines for better coverage.
- Perform real-time Twitter search without requiring API keys using multi-engine and oEmbed.
- Conduct deep research on complex queries with iterative rounds and coverage analysis.
- Run parallel multi-agent research by decomposing queries into sub-tasks across subagents.
- Switch between free and API search layers based on cost or availability needs.
Who the dsh-search-boost DSH plugin is for
- DSH users who need more powerful and flexible web search capabilities.
- Researchers and developers who want deep research and multi-agent search workflows.
- Users who prefer free search engines and want to avoid API key dependencies.
dsh-search-boost DSH plugin limitations
- Requires DSH environment and the `dsh` CLI with `--profile` flag; missing global `dsh` command may need manual setup.
- Free engines (Antigravity CLI, Bing, DuckDuckGo, Exa MCP) are platform-dependent (Antigravity CLI only on macOS/Linux).
- Keyed engines (Tavily, Brave, Exa) need API keys configured via file or environment variables.
- x_search official path requires optional xAI credentials (SuperGrok/X Premium+ or API key) and explicit opt-in via `/x-login`.
- Dynamic plugin via `plugin-host.js` does not survive process restart.
dsh-search-boost DSH plugin: from the repository README
Quoted from the Mr-remon219/dsh-search-boost README, the upstream source of the dsh-search-boost DSH plugin. Copyright remains with the original authors.
> Search boost for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (DSH): multi-engine fused search, focused page fetching, X (Twitter) search, step-mode deep research, parallel multi-agent research, and an injected proactive-search policy. A **bundle plugin** for DSH that upgrades the built-in `web_search` and registers a family of search tools: - Free-by-default engines run **in parallel**: **Antigravity CLI / Bing / DuckDuckGo / Exa MCP (exa-free)** (all keyless), with keyed **Tavily / Brave / Exa** joining when keys are present. - **Two search layers, switched with `/web_change`**: `free` (keyless engines only — no credits, ideal for research loops) vs `api` (full pool incl. keyed Tavily / Brave / Exa). Choice persists across reloads. - Fused multi-engine ranking with cross-engine co-occurrence scoring and time-decay freshness. - Deep research driven by the main agent, and parallel research fanned out to native DSH subagents. ## Features | Capability | Description | |---|---| | **Built-in `web_search` + `web_fetch` upgrade** | Registers `WebSearchProvider` + `WebFetchProvider` and patches both seam configs (`searchProvider` + `fetchProvider`), so the bui
Read the full READMERepository license: MIT
dsh-search-boost DSH plugin questions
How do I install dsh-search-boost?
The recommended way is to run `dsh plugin --profile web add dsh-search-boost` from the command line. You must specify a profile (e.g., `web`). The command will automatically install the package and wire the patch layer. After installation, restart DSH with `dsh --profile web` to activate the plugin.
Do I need API keys to use the search engines?
No, you can use the free engines (Bing, DuckDuckGo, Exa MCP, and Antigravity CLI on macOS/Linux) without any keys. If you want to use keyed engines like Tavily, Brave, or Exa, you need to provide API keys via `~/.dsh-search-boost-keys.json` or environment variables. The plugin automatically drops engines without keys.
How does x_search work without credentials?
When no xAI credentials are provided, x_search falls back to a multi-engine route that restricts search to x.com, combined with oEmbed for full-text extraction and anonymous guest GraphQL for user profiles. This allows real-time X/Twitter search without any API keys, though without the official hosted path's advanced syntax.
What is the difference between free and api search layers?
The free layer only uses keyless engines (Antigravity CLI, Bing, DuckDuckGo, Exa MCP). The api layer adds keyed engines (Tavily, Brave, Exa) when keys are present. You can switch between layers at runtime using the `/web_change` command. The choice persists across reloads via a file.
I get 'dsh: command not found' after installing. What should I do?
The official DSH is run via `npx @deepseek-ai/dsh web`, which doesn't install a global `dsh` command. To use the `dsh plugin` command, install DSH globally with `npm install -g @deepseek-ai/dsh`, then reopen your terminal. Alternatively, you can use the install script from the repository which auto-detects npx cache paths.