@verica-app/cli 0.1.4 → 0.1.5

package/README.md

@@ -152,8 +152,12 @@ also answers `200` instead of `202`).
 - `--reuse-same-ref` — only reuse a run on the **same git ref**. Off by default: an
   identical config produces the same output distribution regardless of branch.
 - Only **completed** runs are reused (never a partial/failed one).
-- Incompatible with `--threshold` / `--baseline-ref` / `--baseline-run` — a reused
-  verdict was frozen under its own gate, so it can't honor a new one.
+- Incompatible with `--threshold` / `--baseline-ref` / `--baseline-run`. Reuse hands
+  back a _prior_ run's verdict, frozen under the gate that applied when it ran, so a
+  new `--threshold` can't be recomputed against it. `--baseline-ref` is worse than
+  stale: no-regression compares against the _last run on the ref_ — a moving target —
+  so a cached verdict can never be a fresh no-regression check. Gate on either → run
+  fresh (omit reuse).
 
 Omit `--reuse-if-unchanged` (the default) any time you want a guaranteed fresh run.
 
@@ -176,8 +180,8 @@ During `0.x` the **minor** version is the breaking lever, so pin accordingly:
 
 We bump the **minor** for any breaking change (flags, output shapes, push behavior) and
 the **patch** for additive features and fixes. **1.0** will freeze the commands, flags,
-exit codes, and output shapes under standard semver. See
-[CHANGELOG.md](./CHANGELOG.md) for what changed in each release.
+exit codes, and output shapes under standard semver. See the bundled `CHANGELOG.md`
+for what changed in each release.
 
 MIT licensed. There's no IP in the client — the engine, graders, gate, and crypto all
 run server-side behind the token API.

package/dist/cli.js

@@ -4110,7 +4110,9 @@ var runRequestSchema = external_exports.object({
    * default — an eval's output isn't a pure function of its config (generation +
    * judge are non-deterministic, the model endpoint drifts), so reuse is always
    * the caller's explicit choice and is bounded by `maxAgeHours`. Incompatible
-   * with `gate` (the cached verdict was frozen under the old gate).
+   * with `gate`: a reused verdict is frozen under its own gate (a new threshold
+   * can't be recomputed), and no-regression compares against a moving baseline
+   * (the last run on the ref), so a cache can never be a fresh gated check.
    */
   reuse: external_exports.object({
     /** Turn reuse on. The trigger — everything else is just tuning. */
@@ -4718,7 +4720,7 @@ async function main() {
   }
   if (values["reuse-if-unchanged"] && (threshold !== void 0 || values["baseline-ref"] !== void 0 || values["baseline-run"] !== void 0)) {
     throw new Error(
-      "--reuse-if-unchanged cannot be combined with --threshold / --baseline-ref / --baseline-run (a reused verdict was frozen under the prior gate)."
+      "--reuse-if-unchanged cannot be combined with --threshold / --baseline-ref / --baseline-run: a reused verdict is frozen under its own gate, and no-regression compares against a moving baseline \u2014 neither can be recomputed. Gate on these? Run fresh (drop --reuse-if-unchanged)."
     );
   }
   const opts = {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@verica-app/cli",
-  "version": "0.1.4",
+  "version": "0.1.5",
   "private": false,
   "description": "Run a Verica eval from CI and block the merge on the result.",
   "license": "MIT",