dsh-dejaview

Before you start building a dsh plugin, the most important question is: has someone already made this? dsh-dejaview brings the core idea of DejaView—"the project truth mirror"—into the dsh plugin ecosystem, running a quick novelty check right before you register a tool function.

The workflow is straightforward: you provide a one-line description, a proposed name, or a set of keywords, and the plugin simultaneously searches three public sources—the awesome-dsh-plugin registry, the dsh-plugin GitHub topic, and npm packages tagged dsh-plugin. Results are ranked by IDF-weighted lexical similarity, so rare, distinctive terms carry more weight than ecosystem-common ones. The response is structured JSON: each candidate includes name, URL, source, similarity score, and the specific signal terms behind that score, plus a six-dimension verdict_guidance block and an uncovered_terms list that points toward differentiation opportunities.

Retrieval and scoring are handled entirely by the plugin; the harness model reads the evidence and delivers the final ruling. No extra API key is needed and no external LLM provider is required. It is designed for developers brainstorming dsh plugins who want a fast, evidence-backed scan of the existing landscape before writing a single line of code.