-
Notifications
You must be signed in to change notification settings - Fork 1
Overview
JeVerifier keeps a codebase maintainable and its docs consistent while Claude works on it. It puts TypeSafe's Jev, a fast classifier, beside a Claude coding session to run checks that would be too costly to have a frontier model run on every change: every function against the project's architectural rules, and every statement in the docs against the others. It also picks the docs and code a task needs and digests sessions for handover. Claude reviews what Jev surfaces, once, and later runs show only what changed. Token savings are a modest, secondary benefit (Token savings).
Jev only selects, ranks and labels. It never writes summaries or code, so it cannot invent content; its failure mode is leaving something out, which every part below measures.
| Part | Command | Source |
|---|---|---|
| Code rule check | jeverifier wiki lint |
lint.py |
| Doc contradiction check | jeverifier wiki check / accept / verdict / selftest |
consistency/ |
| Living wiki in each repo | jeverifier wiki init / map / coverage |
wiki.py |
| Reading list for a task |
jeverifier ctx find (MCP context_find) |
ctx.py |
| Session digest for handover |
jeverifier ctx digest, jeverifier wiki log
|
ctx.py, wiki.py |
| API keys without env vars | jeverifier keys gui |
keys.py |
How the parts work: How it works. What they catch: Measurements. What they save: Token savings.