jevkit-core
v0.2.0
Published
Shared substrate for jevkit: the .jevl record format, canonical request digests, question normalization, and token budgets for TypeSafe's Jev.
Maintainers
Readme
jevkit-core
Shared substrate for jevkit, a set of tools for building on TypeSafe's Jev. Nothing in this package calls the API.
.jevlrecord format — one JSON Lines format that serves as test cassette, drift golden set, benchmark instance, and calibration observation. See the spec.- Canonical digests — RFC 8785 canonicalization, so the same request always
hashes the same way.
record_id()includes the model;request_id()does not, which is what lets you pair a golden record with its replay on a new model version. - Question normalization — one view over SDK objects and plain dicts alike.
- Token budgets — estimates against Jev's two documented limits: 64k for state plus all questions, 32k for state plus the longest single question.
Unofficial and unaffiliated with TypeSafe.
Install
npm install jevkit-coreLicense
MIT
