The instrument, kept apart from the lever.
EvalBook exists because the other Bookbag products needed a measurement they did not take themselves. CodeBook runs the work and MadeBook governs it; neither should grade its own homework, and neither should trust a grade it cannot re-check.
01Why a separate product
The design question was never “how do we evaluate”, it was “who may act on the result”.
CodeBook already had station evaluators and factory scorers. MadeBook already had evidence with trust levels. What the suite lacked was one place where traces, datasets, scorers, suites and reports are kept as records with a digest, run by a component that has no authority of its own — so that a passing report is a fact a consumer can fetch and re-hash, not a claim it has to believe.
EvalBook is that place. It serves CodeBook factories and any customer application alike, reads identity, organizations, billing, files and models from Bookbag SSO, and talks to its siblings over a signed, idempotent peer protocol in which anything not written down is not promised.
02Principles
Three things it will not trade.
01
Measure, do not authorize
A report is a measurement with a digest. CodeBook, MadeBook and people decide from it. EvalBook asks before it changes its own protected configuration, and never activates an author’s own change.
02
Evidence has a source
Every metric carries measured, manual, estimated or unavailable, and its denominators. A judge’s score is a judge’s score, calibrated against people. An absent number is a dash.
03
Nothing of yours runs here
The API and the worker execute no customer code. Executable scorers run in a runner you install, outbound only, in a sandbox with no network and a read-only harness.
03Part of Bookbag
One account for every Bookbag product.
EvalBook shares one account service with CodeBook, MadeBook and the rest of the suite. Sign in once and move between them; organizations, people and roles are managed in one place, so somebody who leaves the organization leaves every product at the same moment. Platform administration follows SSO’s per-product role; EvalBook has no screen that grants it.
Free at launch
CodeBook executes. EvalBook measures. MadeBook authorizes.
Traces in, scorers over them, reports with a digest, and impact with its denominators. Nothing it reports is authority: it measures, and the people and products around it decide.
Sign in with your Bookbag account · the organization is the tenant