User repository on GitHub
Jev decides whether a batch of logs is worth acting on. Typed questions, confidence gates, nothing executed.
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Logtriage
CLI that collapses Loki log batches and asks Jev six typed questions in one call, then maps the answers in code to suppress, watch, review, notify, page or an auto-remediation candidate.
- Stars
- 3
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- jyatesdotdev
- Last verified
- Sep 2026
About this project
jev-logtriage Jev decides whether a batch of logs is worth acting on. Your code keeps the thresholds. Nothing is executed. Independent. Not an official TypeSafe AI project. Prometheus is good at conditions you already know how to write in PromQL. This is for the rest. Repeated benign warnings get suppress. A helm reconcile error and a failed ntfy push get notify. Low confidence never auto-acts. It goes to review. Needs uv and a TypeSafe API key. Loki is not required.
Quoted from the project's README. Read the full README.
Highlights
- One System One call per source asks is_routine_noise, severity, impact_scope, needs_action, auto_remediable and category.
- Gates run in code: routine noise at 0.80 suppresses, confidence below 0.50 goes to review, and high severity with priority pages.
- Nothing is executed: auto_remediate_candidate is only a label, and security is never an auto-remediation candidate.
- Runs without Loki on bundled demo fixtures, writes a JSON report and serves repeated batches from a local SQLite cache.
- Follows TypeSafe's speculative fan-out, composite scoring (0.60 severity + 0.40 impact) and confidence-routing patterns.
Quickstart
bash
export TYPESAFE_API_KEY=YOUR_API_KEY
uvx --from jev-logtriage logtriage --demoWatch out
MIT-licensed. Needs Python 3.10+ and a TypeSafe API key; Loki is optional, and the README says the code gates do not move even though numbers vary.
Questions about this entry
Short answers drawn from the linked project or page.
The gate order is designed to prevent that. Routine noise at 0.80 with severity below the page line suppresses first, needs_action under 0.50 downgrades to watch, and anything with confidence below the confidence floor, 0.50 by default, goes to review instead of acting. Only severity at or above 2.0 together with priority at or above 0.70 pages.
No. auto_remediate_candidate is only a label, and the README states the repository does not restart pods, call webhooks, or page anyone. Security is never treated as an auto-remediation candidate even when the auto_remediable score reaches 0.85 and the category would otherwise qualify.
No. The demo runs the same pipeline over bundled sanitized fixtures with uvx or pip on Python 3.10 and needs only a TypeSafe key. Loki is an optional input, and --port-forward will even start the kubectl port-forward for you when the readiness check against the local address fails.
Not as measurements, no. The README says the numbers move a little from run to run while the gates do not, and the sample table shows confidence ranging from 0.38 to 0.80 across decisions. No comparison against a baseline router or against alertmanager's own grouping is published.
Set TYPESAFE_API_KEY and run `uvx --from jev-logtriage logtriage --demo` with no clone, or `uv run logtriage --demo` from the repo. Without uv, `pip install jev-logtriage` then `logtriage --demo`. Python 3.10+ is enough, and Loki is not required.
Gates run in code, in order: routine noise below the page line becomes suppress, no action needed becomes watch, confidence below the floor becomes review, high severity and priority page, and a safe auto-remediable candidate is labeled, otherwise notify. Priority is 0.60*severity/3 plus 0.40*impact/3.
Yes. A JSON report is written to reports/triage-<timestamp>.json unless you pass --no-report, and repeat batches skip Jev and use a local SQLite cache, which --no-cache turns off.
More Jev resources like this
A CLI that turns Jev into Unix pipeline primitives: is, choose, score, filter, and guard commands return typed verdicts, stable exit codes, and versioned JSON without prompt or parsing glue.
Jev Tools & CLIs#community#python#cli
Community
sharziki/semdecidePython
sharziki/semdecidePythonA Hermes skill that calls Jev as a typed tiebreaker: the agent submits a state and a question, and Choice, Score, or Noul answers with probabilities let it act or escalate.
Jev Tools & CLIs#community#python#skills
Community
shantanugoel/ask-jev-skillPython
shantanugoel/ask-jev-skillPythonAgent-ergonomic CLI following the AXI conventions that gives coding agents Jev judgments for blocking risky tool calls, screening fetched content for prompt injection, triaging logs, flagging risky diffs, and ranking many items. Routine commands are decided locally at no cost.
Jev Tools & CLIs#community#typescript#cli
Community
shiftynick/jev-axiTypeScript
shiftynick/jev-axiTypeScriptAn optional routing layer for Hermes Agent, shipped as a Python CLI and library, that sends a request and a public skill catalogue to Jev and returns route, no_skill, or review plus confidence and candidate distributions.
Jev Tools & CLIs#community#python#routing
Community
GodsBoy/jev-agent-skill-routerPython
GodsBoy/jev-agent-skill-routerPythonA streaming grep-by-meaning CLI that judges each line with one Jev yes/no question and prints the lines above a threshold, built to sit behind tail -f, kubectl logs -f, or journalctl -f.
Jev Tools & CLIs#community#python#cli
Community
allebee/jevgrepPython
allebee/jevgrepPythonUse Jev as a Braintrust judge scorer, then inspect selected answers, confidence, and probabilities alongside the scorer answer. Includes tracing for your own calls.
Jev Tools & CLIsArticle#community#evaluation#observability
Community
From 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.