Add loop-jev (TypeSafe System One) for cheaper loop decisions - #622
cobusgreyling wants to merge 1 commit into
Conversation
Jev answers typed Choice/Score/Noul questions instead of writing prose. loop-jev uses that for model-tier routing, semantic context ranking, LLM error detection, and reasoning-trace classification. Pair with loop-gate and loop-context. The TypeSafe key is never logged or committed.
|
This PR changes paths that must run the real Fork PRs from first-time contributors start with those workflows waiting for approval. A maintainer needs to open the Checks tab and click Approve and run workflows. Until that happens, branch protection will show the PR as blocked even after a review. Content-only PRs ( — loop-engineering fork-pr-gate |
|
Thanks @cobusgreyling for contributing a tool example — visible, reviewable PRs like this grow the reference for everyone. What happens next
More ways to help — loop-engineering maintainers |
Loop Readiness AuditScore: 100/100 (L3) Strong loop readiness — good candidate for L3 with explicit gates. Top suggestions
Posted by |
Yes — Jev is now a first-class Loop Engineering tool, not a side repo only.
What
New package
tools/loop-jev(npx @cobusgreyling/loop jev …) uses TypeSafe Jev as the System One layer in front of the coding agent:routeretrieveguardclassifyturnJev assesses; code owns thresholds (same pattern as
loop-gate/loop-context: exit0proceed,2escalate).Low-confidence
frontierpicks fall back tobalancedinstead of overspending.Without
TYPESAFE_API_KEY(or~/.config/typesafe/api_key), a heuristic fallback keeps CI green.--no-fallbackrequires a live call. Secrets instateare redacted before send;doctornever prints the key.Also wired into:
loop jev …pass-throughloop_jev_route/loop_jev_guard/loop_jev_classifyskills/loop-jevCompanion showcase: https://github.com/cobusgreyling/Jev (
packages/js).Auth (do not commit)