v587d/dsh-search-first
Forces multi-source web search before the model answers: on agent/pre-step, when Search First is enabled, it runs query extraction (by a configurable sub-model) → a unified search engine (AnySearch / TinyFish / Tavily) → fetch → organize → cache and runs/ persistence, then injects the organized live leads to guide verification before returning control to the agent loop. Final answers stay with the LLM, which is guided toward standardized custom HTML citation tags.
Search First forces multi-source web search before the model answers. When enabled it runs on agent/pre-step: query extraction by a configurable sub-model, then a unified search engine (AnySearch / TinyFish / Tavily), fetch, organize, cache and runs/ persistence, then injects the organized live leads to guide verification before returning control to the agent loop — the final answer still belongs to the LLM. It guides the host model toward standardized custom HTML citation tags, models and ranks multi-source search results, and selectively fetches pages back to the host model.
Install
dsh plugin --profile web add https://github.com/v587d/dsh-search-first.gitCopied verbatim from the README's Installation section ('install from the GitHub repo'). Prerequisites are listed there: a mounted dsh web profile (dsh ≥ 0.1.1-rc.2) and Node ≥ 22.
Compatibility
A Cordis plugin for the DSH web profile; it enforces search on agent/pre-step when enabled, and final answers stay with the LLM.
Details
- Repo: v587d/dsh-search-first
- Category: Tools & Capabilities
- Stars: 1
- Version: GitHub repo v587d/dsh-search-first (1★), installed as a https://github.com/...git spec to the repo itself.
Alternatives
1624318455/dsh-plugin-tavily · 6Mikao9/dsh-wsl-workspace · 863683348/dsh-plugin-academic-writing