User repository on GitHub
Fish-style zsh history autosuggestions ranked by Jev (TypeSafe)
- Forks
- 10
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Shell History
Fish-style zsh autosuggestions ranked by Jev: each keystroke asks which of your last 100 distinct history entries you are completing and shows the best match for one-key acceptance.
- Stars
- 119
- Language
- TypeScript
- Reading time
- 3 min
- Published by
- Community
- Author
- mrnugget
- Last verified
- Sep 2026
About this project
Fish-style autosuggestions for zsh, ranked by Jev (TypeSafe). As you type, the plugin looks at your last 100 distinct history entries and asks Jev which one you are most likely completing. The best match is shown in grey after the cursor, with its score; press → (or ^E / End) to accept it. (The demo runs against a fabricated history; regenerate it with demo/make-demo.sh, which needs vhs, ffmpeg and TYPESAFEAPIKEY.) Requirements: zsh 5.9+, Node 22+ (runs the TypeScript directly), a TypeSafe API key.
Quoted from the project's README. Read the full README.
Works forJev for Scoring
Used inJev for Software & DevTools
Highlights
- Considers the last 100 distinct history entries with one Jev request per keystroke.
- Prefix mode completes literal matches; fuzzy mode replaces when nothing starts with the input.
- The suggestion appears grey after the cursor with its score, accepted with right-arrow, ^E, or End.
- Two questions in one request: a Choice over candidate ids and a Noul for whether any of them completes the line.
Watch out
Needs zsh 5.9 or newer, Node 22 or newer, and a TypeSafe key, and each request takes roughly 0.7 to 0.9 seconds.
Questions about this entry
Short answers drawn from the linked project or page.
One request per keystroke carries the last JEV_HISTORY_LIMIT distinct commands, 100 by default, plus what you have typed so far. Literal matches never leave your machine: if any history entry starts with what you typed, only those entries are sent, and a single prefix match is suggested with no request at all. Otherwise the full candidate list goes out as fuzzy mode.
Latency is roughly 0.7 to 0.9 seconds per request and almost all of it is API time; Node startup and history parsing together are about 0.1 s. The plugin hooks line-pre-redraw, kills any in-flight request when the buffer changes, and applies a result only if the buffer is still what was typed when the request started, so fast typing does not queue up stale suggestions.
Fuzzy mode suggests only when the top score clears JEV_MIN_SCORE (0.3) and either the Noul clears JEV_THRESHOLD (0.5) or the top score clears JEV_STRONG_SCORE (0.9). Both signals are used because they fail in different places: the Noul under-fires on tiny histories while the Choice spreads out on nonsense. Nothing happens below JEV_MIN_CHARS, and the e2e test asserts nonsense yields nothing.
It is autosuggestion inside the prompt, not a separate search UI. The grey completion and its score appear after the cursor as you type, and forward-char, ^E, or End accepts it in both emacs and vi keymaps, since the widget fires whenever the cursor sits at the end of the line. There is also a standalone jev-accept-suggestion widget if you want your own binding.
Requires zsh 5.9+, Node 22+ and a TypeSafe API key. Clone the repo to ~/.zsh/jev-shell-history, run `npm install`, then in ~/.zshrc export TYPESAFE_API_KEY and source the plugin file. Suggestions are accepted with the right arrow or ^E at the end of the line.
It hooks line-pre-redraw, kills any in-flight request on each buffer change and starts src/cli.ts in the background via zle -F on a process-substitution fd, so the prompt never blocks. A result is applied only if the buffer is still what was typed when the request started.
If any history entry literally starts with the typed text, only those candidates are sent in prefix mode, and a single prefix match is suggested immediately with no request at all. Otherwise all recent candidates are sent in fuzzy mode and shown only if a Choice score and a Noul gate both clear.
More Jev resources like this
Vercel 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 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.
Jev Tools & CLIs#community#typescript#cli
Community
mizchi/jev-lintTypeScript
mizchi/jev-lintTypeScriptA CLI that grades markdown and text files against plain-Markdown rulesets with Jev, running every rule against every line in parallel and caching unchanged lines to avoid repeat API calls.
Jev Tools & CLIs#community#typescript#cli
Community
lukstei/slop-graderTypeScript
lukstei/slop-graderTypeScriptA 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 provider-neutral CLI, MCP server and Claude Code plugin that asks yes/no, pick-one, rate or rank questions about text and gets probabilities instead of paragraphs. Backends include Jev (TypeSafe or via OpenRouter), OpenAI's Decisions API, and Claude with an existing Claude Code login. MIT, Node 20+.
Jev Tools & CLIs#community#github#typescript
Community
adityajha2005/decisionkitTypeScript
adityajha2005/decisionkitTypeScriptA 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/jevgrepTypeScriptFrom 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.