@leostera/evalkit 0.0.1 → 0.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +1 -1
- package/dist/cli.mjs +4 -5
- package/dist/cli.mjs.map +1 -1
- package/dist/dashboard/assets/index-BYSEeNaZ.js +30 -0
- package/dist/dashboard/assets/index-fUDsY6z-.css +2 -0
- package/dist/dashboard/index.html +2 -2
- package/dist/{index-BwNYw4-_.d.mts → index-M5eRlMfd.d.mts} +5 -3
- package/dist/index-M5eRlMfd.d.mts.map +1 -0
- package/dist/index.d.mts +1 -1
- package/dist/index.mjs +1 -1
- package/dist/runner.d.mts +1 -1
- package/dist/runner.mjs +1 -1
- package/dist/{src-DJDem6nZ.mjs → src-DluldGKn.mjs} +3 -3
- package/dist/{src-DJDem6nZ.mjs.map → src-DluldGKn.mjs.map} +1 -1
- package/dist/{src-CopGdrwX.mjs → src-DrwwgaPn.mjs} +5 -3
- package/dist/src-DrwwgaPn.mjs.map +1 -0
- package/package.json +4 -2
- package/dist/dashboard/assets/index-B2P48MWY.css +0 -1
- package/dist/dashboard/assets/index-Z5UamLUC.js +0 -261
- package/dist/index-BwNYw4-_.d.mts.map +0 -1
- package/dist/src-CopGdrwX.mjs.map +0 -1
package/README.md
CHANGED
|
@@ -43,7 +43,7 @@ Runs create `_evalkit-results/` and `_evalkit-sandbox/` in the generated project
|
|
|
43
43
|
|
|
44
44
|
## Explore more
|
|
45
45
|
|
|
46
|
-
- [Get started](
|
|
46
|
+
- [Get started](sites/evalkit.leostera.dev/src/content/docs/index.md) and the [manual](sites/evalkit.leostera.dev/src/content/docs/manual/) explain adapters, fixtures, scoring, reports, and the CLI.
|
|
47
47
|
- [Interleaved scenario](examples/interleaved-scenario/) has a passing write/revise flow, an intentional failfast failure, and a provider-free fake judge agent.
|
|
48
48
|
- [Configured matrix](examples/configured-matrix/) shows a runnable parameter sweep without model credentials.
|
|
49
49
|
- [HACKING.md](HACKING.md) covers the repository layout, builds, and tests for contributors.
|
package/dist/cli.mjs
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
#!/usr/bin/env bun
|
|
2
|
-
import { F as
|
|
3
|
-
import { a as readTrialEvents, c as readTrialSummary, i as readRunSummary, l as localReportStore, n as runMatrix, o as readTrialManifest, r as readRunManifest, t as runEval } from "./src-
|
|
2
|
+
import { F as defineEvalMatrix, I as canonicalParameters, m as registerEvals, n as authoringId } from "./src-DrwwgaPn.mjs";
|
|
3
|
+
import { a as readTrialEvents, c as readTrialSummary, i as readRunSummary, l as localReportStore, n as runMatrix, o as readTrialManifest, r as readRunManifest, t as runEval } from "./src-DluldGKn.mjs";
|
|
4
4
|
import * as Schema from "effect/Schema";
|
|
5
5
|
import { Effect, Option } from "effect";
|
|
6
6
|
import { mkdir, readFile, readdir, stat, writeFile } from "node:fs/promises";
|
|
@@ -8663,7 +8663,7 @@ Commit \`bun.lock\` to pin the resolved EvalKit Git revision.
|
|
|
8663
8663
|
}
|
|
8664
8664
|
//#endregion
|
|
8665
8665
|
//#region ../cli/src/new-project.ts
|
|
8666
|
-
const npmDependency = "^0.0
|
|
8666
|
+
const npmDependency = "^0.1.0";
|
|
8667
8667
|
const ignored = [
|
|
8668
8668
|
"node_modules/",
|
|
8669
8669
|
"_evalkit-results/",
|
|
@@ -8778,7 +8778,6 @@ async function writeStarterReadme(root, name, inPlace) {
|
|
|
8778
8778
|
}
|
|
8779
8779
|
//#endregion
|
|
8780
8780
|
//#region ../cli/src/index.ts
|
|
8781
|
-
process.env.NO_COLOR === void 0 && (process.env.FORCE_COLOR === "1" || process.stdout.isTTY);
|
|
8782
8781
|
let projectRoot = process.cwd();
|
|
8783
8782
|
let reportRoot = resolve(projectRoot, "_evalkit-results");
|
|
8784
8783
|
const sandboxRoot = resolve(projectRoot, "_evalkit-sandbox");
|
|
@@ -9045,7 +9044,7 @@ const cli = run(make$1("evalkit").pipe(withSubcommands([
|
|
|
9045
9044
|
dashboardCommand
|
|
9046
9045
|
])), {
|
|
9047
9046
|
name: "EvalKit",
|
|
9048
|
-
version: "0.0
|
|
9047
|
+
version: "0.1.0"
|
|
9049
9048
|
});
|
|
9050
9049
|
const argv = process.argv.slice();
|
|
9051
9050
|
if (argv[2] === "help") argv.splice(2, 1, "--help");
|