jev-search (larguesa)
Experimental Python CLI for semantic line search with Jev via OpenRouter, meant to complement exact search for agents and notes.
Overview
jev-search by larguesa is a dependency-free Python CLI that finds lines by meaning to complement exact search. Exact search finds names, symbols and literal references; jev-search asks whether a line expresses the intent you are looking for even with different wording, for example finding 'we adopted open formats to make migration easier' from a query about reducing vendor dependence. Suggested uses are exploring design intentions in repository docs, finding related ideas in Markdown notes or second-brain wikis, and checking selected passages alongside Obsidian or OpenViking workflows.
The README's early evidence is modest and candid: in six simulated tasks over real technical text, lexical search plus Jev recovered 29 of 34 labelled relevant passages versus 22 of 34 for lexical alone, with no extra false positives, for about $0.0006 at 0.536 s median latency, but several extra hits were adjacent to lexical hits and no advantage over lexical search plus reading context has been demonstrated. It supports Linux and Windows (not macOS), accepts only local regular files in trusted directories, and v0.2.0 is on main without a tagged release.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Does this line express the intent described by the query? | yes · no |
How it works
- Run an exact search with your usual tools
- Run jev-search with a meaning-based query over selected files
- Jev judges candidate lines
- Read surrounding context and original sources
Strengths
- Finds paraphrased content keywords miss
- No runtime dependencies
- Careful path safety checks
- Honest reporting of limits
Caveats
- Small curated experiments only
- Not shown to beat lexical search plus context reading
- macOS unsupported; v0.2.0 not tagged
Getting started
git clone --branch main https://github.com/larguesa/jev-search.git, review it, then uv tool install . and run jev-search --helpAt a glance
- Type
- CLI
- Status
- prototype
- Language
- Python
- Runtime
- CLI (and agent skill)
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Experimental semantic line search with TypeSafe Jev via OpenRouter. Python CLI with no runtime dependencies.
logicrw/awesome-jev-projectsExperimental semantic line search with TypeSafe Jev via OpenRouter. Python CLI with no runtime dependencies.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Whether each line expresses the intent of a meaning-based query.
Does it replace grep?
No. It is meant to complement exact search.
How well does it work?
In six small tasks it found 7 more relevant passages than lexical search alone, but the README makes no general claim.
Related projects
Deterministic code-graph context for coding agents on Postgres, AGE and pgvector, with optional Jev or local re-selection of symbols.
MCP serverbetaPythonClaude Code plugin that adds Jev as an independent judge for review findings, debugging hypotheses, design options and search hits.
PluginbetaRetrieval layer for Codex that finds evidence locally, optionally ranks sanitized excerpts with Jev, and returns bounded, cited evidence.
ProjectbetaPipe grep or file lists into Jev to get every candidate scored, sorted and cut at a calibrated threshold, so agents read only what matters.
ProjectbetaTypeScriptsys1grep: a zero-dependency grep that finds lines by meaning, scoring each line with Jev, with AND/OR/NOT and cross-language queries.
ServicebetaJavaScriptSingle MCP tool that indexes a repo and has Jev rank a shortlist so a coding agent reads the right files and lines instead of grepping.
MCP serverbetaPython