@mastra/evals 0.1.0-alpha.47 → 0.1.0-alpha.49

package/CHANGELOG.md

@@ -1,5 +1,21 @@
 # @mastra/evals
 
+## 0.1.0-alpha.49
+
+### Patch Changes
+
+- Updated dependencies [66a5392]
+  - @mastra/core@0.2.0-alpha.107
+
+## 0.1.0-alpha.48
+
+### Patch Changes
+
+- a8a459a: Updated Evals table UI
+- Updated dependencies [6f2c0f5]
+- Updated dependencies [a8a459a]
+  - @mastra/core@0.2.0-alpha.106
+
 ## 0.1.0-alpha.47
 
 ### Patch Changes

package/dist/metrics/llm/index.js

@@ -1251,7 +1251,7 @@ var ToxicityJudge = class extends MastraAgentJudge {
         reason: z.string()
       })
     });
-    return result.object;
+    return result.object.reason;
   }
 };
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mastra/evals",
-  "version": "0.1.0-alpha.47",
+  "version": "0.1.0-alpha.49",
   "description": "",
   "type": "module",
   "main": "dist/index.js",
@@ -37,7 +37,7 @@
     "sentiment": "^5.0.2",
     "string-similarity": "^4.0.4",
     "zod": "^3.24.1",
-    "@mastra/core": "^0.2.0-alpha.105"
+    "@mastra/core": "^0.2.0-alpha.107"
   },
   "peerDependencies": {
     "ai": "^4.0.0"

package/src/metrics/llm/context-position/index.test.ts

@@ -1,7 +1,7 @@
 import { createOpenAI } from '@ai-sdk/openai';
 import { describe, it, expect } from 'vitest';
 
-import { TestCaseWithContext } from '../utils';
+import { isCloserTo, TestCaseWithContext } from '../utils';
 
 import { ContextPositionMetric } from './index';
 
@@ -184,7 +184,7 @@ describe(
         context: testCase.context,
       });
       const result = await metric.measure(testCase.input, testCase.output);
-      expect(result.score).toBeCloseTo(testCase.expectedResult.score, 1);
+      expect(isCloserTo(result.score, testCase.expectedResult.score, 0.5)).toBe(true);
     });
 
     it('should handle mixed relevance with good ordering', async () => {

package/src/metrics/llm/toxicity/metricJudge.ts

@@ -34,6 +34,6 @@ export class ToxicityJudge extends MastraAgentJudge {
       }),
     });
 
-    return result.object;
+    return result.object.reason;
   }
 }