LEEVAR reliability battery runs on your own machine — the client starts it, so there is no endpoint to ping. Last commit 21 Oct 2026.
Grade an AI agent's transcripts on 18 reliability tests. Thin evidence is NOT TESTED, not guessed.
Nothing serious here today
Today is the operative word: we check LEEVAR reliability battery every 15 minutes and re-read its code on every release. Watch it and you find out the day that stops being true.
Alternatives to LEEVAR reliability battery
same job, measured the same wayAI agents trade general work for test USDC on Base Sepolia, with escrow and signed evidence.
Analyse tests for item quality, true ability, reliability and learner mastery.
Moderated usability testing: read sessions, notes, transcripts, reports, and draft test scenarios.
Isolated macOS app tests, MCP images, OCR/ASCII frames, accessibility and evidence reports.
Read-only lookup of tested agent tools and MCP servers: scores, test evidence, deprecation.
Test the voice agents you run: scored transcripts, pass/fail verdicts, latency and WER metrics.
Local QA workflows, test evidence, screenshot maps and confidence reports for AI coding agents.
Evidence-oriented code questions, diff review and test selection via a Jev endpoint.
LEEVAR reliability battery — questions
Answers built from our own checks of this server.