@ls-stack/agent-eval 0.34.0 → 0.35.0

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/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@ls-stack/agent-eval",
3
- "version": "0.34.0",
3
+ "version": "0.35.0",
4
4
  "type": "module",
5
5
  "bin": {
6
6
  "agent-evals": "./dist/bin.mjs"
@@ -298,11 +298,12 @@ See `EvalScoreDef` / `EvalManualScoreDef` in the types for the full shape
298
298
  ## Outputs, columns, trace display
299
299
 
300
300
  - `setEvalOutput(key, value)` writes reviewable data for the case. Values are
301
- plain data (strings, numbers, booleans, JSON-safe objects) plus native
302
- `Blob`/`File` or `FileRef` variants for media columns. Stored output values
303
- are not coerced by `columns.format`; column formats only control
304
- visualization. Native `Blob`/`File` values are copied to run artifacts
305
- because saved run files are JSON. Inside `execute`, prefer the context
301
+ stored as received: primitives, objects/arrays, explicit file refs, and
302
+ native `Blob`/`File` values. `columns.format` only controls visualization.
303
+ Non-JSON runtime values such as `Date`, `Map`, `Set`, `BigInt`, typed arrays,
304
+ and class instances use the tagged value serializer instead of a string
305
+ fallback. Native `Blob`/`File` values are copied to run artifacts because
306
+ saved run files are JSON. Inside `execute`, prefer the context
306
307
  `setOutput(key, value)` helper when writing schema-backed outputs; it is
307
308
  typed from the eval's outputs generic. Keep `setEvalOutput` for shared
308
309
  workflow code that does not receive the execute context.
@@ -1,3 +0,0 @@
1
- import "./runOrchestration-CYhn4DJ0.mjs";
2
- import "./cli-CG66f0GN.mjs";
3
- export {};