Small decisions. Interesting possibilities.Submit contentSubmit
Jevgrep preview

Jevgrep

A code search CLI that asks Jev typed relevance and scope questions about candidate source blocks.

Added to Jevfast

How it uses Jev

On a CLI search request, Jevgrep sends a query, source path, code text and declaration context as state. Jev answers Boolean relevance, scope and reference questions for candidate evidence blocks; the CLI uses those judgments to choose useful source context.

What Jev decides

Illustrative evidence requestExample answers · not a recorded Jev response · Source ↗
Question 1 · q0
YOUR APP
INSTRUCTION

Is this candidate source block evidence for the query?

STATE

Query, path, source text, declarations, and optional guidance or selected evidence.

JEV · BOOLEAN
YesNo
Question 2 · scope0
YOUR APP
INSTRUCTION

Is the candidate at the target scope for the query?

STATE

Query, path, source text, declarations, and optional guidance or selected evidence.

JEV · BOOLEAN
YesNo

App workflow

  1. Receive a query

    A user or coding agent asks a question about a codebase.

  2. Collect candidates

    The CLI narrows files and declarations and prepares source blocks.

  3. Judge evidence

    Jev answers typed relevance and scope questions for those blocks.

  4. Return context

    The CLI presents the selected code evidence to the caller.

Why it is interesting

It decomposes semantic code search into inspectable judgments over concrete code blocks, including whether a match answers the query and whether it is the right scope.