jegrep

jegrep</h1>
  <p><em>Jevantic grep: describe it, find it.</em></p>
  <a href="https://github.com/can1357/jegrep/actions/workflows/ci.yml"><img src="https://github.com/can1357/jegrep/actions/workflows/ci.yml/badge.svg" alt="CI" /></a>
  <a href="https://github.com/can1357/jegrep/releases/latest"><img src="https://img.shields.io/github/v/release/can1357/jegrep" alt="GitHub release" /></a>
  <a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-blue.svg" alt="License" /></a>
</div>

Natural-language search that works like `grep`. No embeddings, no index, no daemon.

- **Semantic:** Finds concepts ("where do we verify JWT tokens?"), not just strings.
- **No Index:** Searches the live tree on every run. Nothing to build, refresh, or go stale.
- **Calibrated:** Every path gets an absolute yes/no probability, so thresholds mean something and batches stay comparable.
- **Precise:** Returns files *and* line ranges as `dirname/filename:first-last`
  references, ready to paste into an editor, with original line numbers and
  merged adjacent passages.
- **Cheap:** Jev bills $0.042 per million input tokens, output free. A typical search over a few thousand files runs **$0.01–0.03**.
- **Agent-Ready:** `--json` for scripts and coding agents, `--compact` for a token-lean digest, plus a benchmark harness for regressions.

## Quick Start

1. **Install**

   With [Rust and Cargo](https://rustup.rs) installed:

   ```bash
   cargo install jegrep
   ```

   Or download the archive for your platform from
   [GitHub Releases](https://github.com/can1357/jegrep/releases/latest), extract it,
   and put `jegrep` (`jegrep.exe` on Windows) on your `PATH`. Releases include
   Linux and macOS binaries for x86-64 and ARM64, Windows x86-64 binaries, and
   `S