@axlsdk/eval 0.2.0 → 0.4.0

package/README.md

@@ -92,6 +92,7 @@ import { runEval } from '@axlsdk/eval';
 
 const results = await runEval(
   {
+    workflow: 'my-workflow',  // label stored in results for comparison
     dataset: ds,
     scorers: [exactMatch, qualityJudge],
     concurrency: 5,
@@ -103,7 +104,7 @@ const results = await runEval(
 );
 
 console.log(results.summary);
-// { exact-match: { mean: 0.85, min: 0, max: 1 }, quality: { mean: 0.92, ... } }
+// { count: 3, failures: 0, scorers: { 'exact-match': { mean: 0.85, min: 0, max: 1, p50: 1, p95: 1 }, ... } }
 ```
 
 ### `evalCompare(baseline, candidate)`
@@ -116,10 +117,10 @@ import { evalCompare } from '@axlsdk/eval';
 const comparison = evalCompare(baselineResults, candidateResults);
 
 console.log(comparison.regressions);
-// [{ item: ..., scorer: 'exact-match', baseline: 1, candidate: 0 }]
+// [{ input: ..., scorer: 'exact-match', baselineScore: 1, candidateScore: 0, delta: -1 }]
 
 console.log(comparison.improvements);
-// [{ item: ..., scorer: 'quality', baseline: 0.5, candidate: 0.9 }]
+// [{ input: ..., scorer: 'quality', baselineScore: 0.5, candidateScore: 0.9, delta: 0.4 }]
 ```
 
 ### `defineEval(config)`
@@ -130,7 +131,7 @@ Register an eval for CLI discovery:
 import { defineEval } from '@axlsdk/eval';
 
 export default defineEval({
-  name: 'math-eval',
+  workflow: 'math-workflow',
   dataset: ds,
   scorers: [exactMatch],
 });
@@ -142,10 +143,13 @@ Run evaluations from the command line:
 
 ```bash
 # Run an eval file
-npx axl-eval run ./evals/math.ts
+npx axl eval ./evals/math.ts
+
+# Save results to a file
+npx axl eval ./evals/math.ts --output ./results/baseline.json
 
 # Compare two results
-npx axl-eval compare ./results/baseline.json ./results/candidate.json
+npx axl eval compare ./results/baseline.json ./results/candidate.json
 ```
 
 ## Integration with AxlRuntime

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@axlsdk/eval",
-  "version": "0.2.0",
+  "version": "0.4.0",
   "description": "Evaluation framework for Axl agentic workflows",
   "type": "module",
   "main": "./dist/index.cjs",
@@ -25,7 +25,7 @@
     "dist"
   ],
   "dependencies": {
-    "@axlsdk/axl": "0.2.0"
+    "@axlsdk/axl": "0.4.0"
   },
   "devDependencies": {
     "@types/node": "^25.2.3",