User repository on GitHub
lint text in code by jev scorerer
- Forks
- 4
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Lint — mizchi
An ast-grep semantic linter where each rule is one natural-language question judged by Jev: 98 rules flag name/body drift, stale comments, weak tests, quiet failures, and commits whose diffs contradict their messages.
- Stars
- 120
- Language
- TypeScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- mizchi
- Last verified
- Sep 2026
About this project
A lint tool that uses Jev -- a fast classifier that answers a natural-language question with a calibrated probability instead of writing text -- to decide what a parser cannot. This function is wrong, and nothing in your toolchain will say so: The comment promises null. The body throws. The return type is Cart, not Cart | null. It type-checks and it lints clean.
Quoted from the project's README. Read the full README.
Used inJev for Software & DevTools
Highlights
- Forty-seven questions ship, most asked in several languages, for 98 rules total, each with a cutoff and a score on its own fixtures.
- Six findings over two example files cost a tenth of a cent; costs for larger runs are measured in docs/cost.md rather than extrapolated.
- The key is read only from TYPESAFE_API_KEY or TYPESAFEAI_API_KEY in the environment, never from the config file.
- jev-lint review judges only lines a diff touched; jev-lint commits judges each message against its diff and the repository's own AGENTS.md or CLAUDE.md.
- The README says it is not for what compilers, type checkers, or ESLint already decide, and notes it is distinct from huntedman/JevLint.
Quickstart
bash
export TYPESAFE_API_KEY=YOUR_API_KEY
npx -y jev-lint check examples
npx jev-lint init
npx jev-lint check src --dry-runWatch out
MIT-licensed; needs Node 24+ and npx plus a TypeSafe API key for model calls, though --dry-run, rules, replay, and cached CI runs work with no key.
Questions about this entry
Short answers drawn from the linked project or page.
Measured on this repository, about one in five. A score near a cutoff can move by a few hundredths between runs. Verdicts are cached by rule and code content, --retry 3 decides on the mean of three passes and marks what did not reproduce, and --loose shows the band under a cutoff without making it a finding.
No, and the README rules itself out of that scope. It targets claims code makes about itself that a compiler, type checker, or ESLint cannot decide — a doc comment promising null while the body throws, a test that would still pass with its named behavior broken. It is explicitly poor at defects needing knowledge of a specific API, such as .sort() defaulting to lexicographic order.
Yes — the code being judged is sent to the API, and that is what the key pays for. Only the ast-grep matcher runs locally, deciding which code is looked at. The key itself is read from the environment (TYPESAFE_API_KEY or TYPESAFEAI_API_KEY) and never from the config file, since that file belongs in version control.
Not by default. No shipped rule is set to severity: error, so neither the pull-request annotation nor the pre-commit and pre-push hooks block anything until you raise a rule yourself. --dry-run shows exactly what it would ask and the price without sending a request, and jev-lint replay re-scores a recorded run.
It picks the rules and their thresholds, much like an ESLint config. init writes every shipped rule turned on as a catalogue to prune, and a rule can be set on, off, or with a threshold and a severity such as error, where any answer at or above the threshold becomes a finding. Rules use a language/id namespace, and a bare shipped id warns because it can enable a language unintentionally.
Use jev-lint review --base main, which judges only the lines a diff touched, where findings concentrate, at a fraction of the cost. jev-lint commits --base main checks each commit message against its diff and each change against the repository's own AGENTS.md, with --staged for a change that has no message yet.
TypeScript with 23 rules and Rust with 9 are first tier, meaning every rule there carries fixtures, expectations, and an accepted baseline, while Python, Go, Moonbit, shell, markdown, JSON, text, and git rules are calibrated to the same bar but not yet promised. Your own ast-grep rule can target any language the parser handles.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Decision outputs in Grep
Hacker News: jevmem keeps Claude Code project memory by asking Jev which turns are worth saving.
avinashjetwani61 pointsJevmem – automatic project memory for Claude Code, built on JevRead the thread on Hacker NewsX: A usage router for Grok Bot
X: The compaction plugin, in Chinese
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
Fuzzy linter hook for Claude Code and Codex: after the agent edits a file, it sends the added code to Jev, which answers Noul checks against team rule packs in about 0.3 seconds, letting the agent fix violations before code review.
Jev Tools & CLIs#community#linter#claude-code
Community
ckorhonen/jev-lintHTML
ckorhonen/jev-lintHTMLA natural-language linter for Node 24+ in which an ast-grep matcher selects the code and a one-sentence Jev rule returns a calibrated verdict. Ships 99 rules covering names, comments, tests, shell scripts, commit messages, and AGENTS.md files.
Jev Tools & CLIs#cli#typescript#guardrails
Community
More Jev resources like this
A semantic grep CLI that chunks files or staged diffs into 5-60 line pieces, packs 16 chunks and 16 Jev Noul questions into one request, and prints file:line hits above a probability threshold with grep-style exit codes.
Jev Tools & CLIs#community#typescript#cli
Community
kyu1204/jgrepTypeScript
kyu1204/jgrepTypeScriptA CLI that asks Jev which files and declarations answer a repository question, then prints verbatim excerpts for a coding agent to use.
Jev Tools & CLIs#community#typescript#cli
Community
dzhng/jevgrepTypeScript
dzhng/jevgrepTypeScriptA Rust CLI that reads completed Claude Code and Codex sessions, scores each turn with Jev, and turns recurring patterns into suggested edits for your AGENTS.md, skills and tools. Ships as jta with init, analyze and report commands.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
hobbs/jev-turn-analysisRust
hobbs/jev-turn-analysisRustVercel Labs' terminal CLI for the AI SDK: its `ai evaluate` command answers named Boolean, Choice, and Score questions, with Jev as the default evaluation model and one unchanged input shared across all questions.
Jev Tools & CLIs#community#typescript#cli
Community
vercel-labs/ai-cliTypeScript
vercel-labs/ai-cliTypeScriptAn experimental Sutro CLI that evaluates CSV, Parquet, and JSONL data with Jev, asks you to label ambiguous and random audit rows, and uses GEPA to propose improved function definitions.
Jev Tools & CLIs#community#python#cli
Community
sutro-sh/jev-alignPython
sutro-sh/jev-alignPythonA local-first MCP server that gives Claude Code, Codex, Cursor, and OpenCode structured quality scores from Jev across correctness, complexity, tests, security, and other dimensions while the agent writes code.
Jev Tools & CLIs#community#typescript#mcp
Community
NiazMorshed2007/jev-reviewTypeScript
NiazMorshed2007/jev-reviewTypeScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.