What it is
A Neovim plugin that splits the buffer into functions with Treesitter, asks a plain-language question about each, and ranks the answers by probability in quickfix. For developers searching code by intent rather than pattern.
How it uses Jev
Jev scores each function against one plain-language question, returning a calibrated probability. The plugin sends one question per function, packs requests under a token budget, and uses the probabilities to rank hits in quickfix and show them as virtual text.
Primitives:score
Technique worth stealing
Use a decision model to score each code unit against a plain-language question, then rank by probability.
Try it
Install with lazy.nvim, set TYPESAFE_API_KEY, then run :Jev <question> [glob].
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.90
- Shows a System One patternunclear0.44
- Handles uncertaintyno0.05
- Measuredno0.39
- Runnableyes0.97
- Worth recommendingunclear0.45
- Model replicano0.09
- Problem scopescore on a 0–2 scale1.49
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.