Does Jev rewrite my code directly?

No. Jev only answers yes/no, score, and choice questions with probabilities. Your agent reads the JSON or TTY report and edits only the flagged hunks.

How do I gate CI without blocking every warning?

Run audit with JSON output, which exits 1 when findings reach the fail-on level, high by default in CI. Interactive terminals default to no failure gate.

Where do the rules live, and can I customize them?

Packs are YAML and the only definition of good code; the CLI ships a pinned core pack plus per-repo overrides. The update command previews question and threshold diffs before pulling.

Will re-runs re-bill every hunk on a large diff?

Hunk plus context is capped under 24k tokens, results cache by state hash with 24-hour expiry for rolling model aliases, and audit --json keeps humans on stderr for agent parsing.

What does an audit send to TypeSafe?

Per-hunk state: the file, language, unified-diff hunk and up to context_lines of surrounding code. Whole files are never uploaded, ignored, .env* and gitignored paths are never sent, and there is no telemetry.

How do I customize the rules jev-codes checks?

Standards packs are YAML; bundled core covers leftover debug, duplicate logic, swallowed errors and more, and per-repo overrides live in .jev-codes/standards.yaml using extends: core and overriding by question key.

How does jev-codes keep re-runs cheap?

Results cache by a hash of state plus question text, so re-runs only re-score changed hunks; jev-latest entries expire after 24 hours so a model rollover cannot serve stale answers.

CommunityJev Practices & Patterns

Jev Codes

CLI and agent commands that audit a git diff hunk-by-hunk against editable YAML standards packs, with Jev answering typed questions and the agent fixing flagged lines.

See highlights

Open github.com

Stars: 5
Language: TypeScript
License: MIT
Published by: Community
Author: Kushwho
Last verified: Oct 2026

Kushwho/jev-codes

User repository on GitHub

View on GitHub

Audit your git diff against YAML coding-standards packs using TypeSafe's Jev model, from a CLI or your AI agent's command/skill.

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

About this project

CLI + agent commands that audit a git diff against editable standards packs, scored hunk-by-hunk with TypeSafe's Jev model. Jev never writes code — it answers typed questions (yes/no, score, choice) with a probability, and your agent fixes only the flagged hunks. A 500-line diff audits in seconds for a fraction of a cent: one parallel Jev call per hunk, output tokens free.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Tool calling

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

One parallel Jev call per hunk with 8 in flight; a 500-line diff audits in seconds.
Adapters for Claude Code, Cursor, Codex CLI, opencode, and Antigravity share one loop.
Bundled core pack covers debug leftovers, swallowed errors, and comment noise.
Eval runner holds 15 hand-labeled diffs with a 6/6 high-severity gate on jev-1.13.0.

Watch out

Only added and changed lines are judged, and per-hunk state goes to the TypeSafe API with ignored and secret paths excluded. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-agent-authorization

Starter kit putting Kinde identity plus Jev judgment in front of every agent tool call, with allow, hold-for-approval, or deny decided in code and written to a ledger first.