@outputai/evals 0.11.0 → 0.11.1-next.2223fa5.0

package/dist/judge.d.ts

@@ -1,7 +1,8 @@
 import { EvaluationVerdictResult, EvaluationNumberResult, EvaluationBooleanResult, EvaluationStringResult, z } from '@outputai/core';
+import { type PromptVariables } from '@outputai/llm';
 export type JudgeArgs = {
     prompt: string;
-    variables?: Record<string, string | number | boolean>;
+    variables?: PromptVariables;
     schema?: z.ZodType;
 };
 export declare function judgeVerdict({ prompt, variables, schema }: JudgeArgs): Promise<EvaluationVerdictResult>;

package/dist/judge.js

@@ -1,31 +1,31 @@
 import { EvaluationVerdictResult, EvaluationNumberResult, EvaluationBooleanResult, EvaluationStringResult, z } from '@outputai/core';
 import { Path } from '@outputai/core/sdk/helpers';
-import { generateText, Output } from '@outputai/llm';
+import { generateText, aiSdk } from '@outputai/llm';
 const verdictSchema = z.object({ verdict: z.enum(['pass', 'partial', 'fail']), reasoning: z.string() });
 const scoreSchema = z.object({ score: z.number(), reasoning: z.string() });
 const booleanSchema = z.object({ result: z.boolean(), reasoning: z.string() });
 const labelSchema = z.object({ label: z.string(), reasoning: z.string() });
 export async function judgeVerdict({ prompt, variables, schema = verdictSchema }) {
     const promptDir = Path.resolveInvocationDir();
-    const response = await generateText({ prompt, variables, promptDir, output: Output.object({ schema }) });
+    const response = await generateText({ prompt, variables, promptDir, output: aiSdk.Output.object({ schema }) });
     const result = response.output;
     return new EvaluationVerdictResult({ value: result.verdict, confidence: 0.9, reasoning: result.reasoning });
 }
 export async function judgeScore({ prompt, variables, schema = scoreSchema }) {
     const promptDir = Path.resolveInvocationDir();
-    const response = await generateText({ prompt, variables, promptDir, output: Output.object({ schema }) });
+    const response = await generateText({ prompt, variables, promptDir, output: aiSdk.Output.object({ schema }) });
     const result = response.output;
     return new EvaluationNumberResult({ value: result.score, confidence: 0.9, reasoning: result.reasoning });
 }
 export async function judgeBoolean({ prompt, variables, schema = booleanSchema }) {
     const promptDir = Path.resolveInvocationDir();
-    const response = await generateText({ prompt, variables, promptDir, output: Output.object({ schema }) });
+    const response = await generateText({ prompt, variables, promptDir, output: aiSdk.Output.object({ schema }) });
     const result = response.output;
     return new EvaluationBooleanResult({ value: result.result, confidence: 0.9, reasoning: result.reasoning });
 }
 export async function judgeLabel({ prompt, variables, schema = labelSchema }) {
     const promptDir = Path.resolveInvocationDir();
-    const response = await generateText({ prompt, variables, promptDir, output: Output.object({ schema }) });
+    const response = await generateText({ prompt, variables, promptDir, output: aiSdk.Output.object({ schema }) });
     const result = response.output;
     return new EvaluationStringResult({ value: result.label, confidence: 0.9, reasoning: result.reasoning });
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@outputai/evals",
-  "version": "0.11.0",
+  "version": "0.11.1-next.2223fa5.0",
   "description": "Offline evaluation framework for Output.ai workflows",
   "type": "module",
   "main": "dist/index.js",
@@ -19,8 +19,8 @@
     "./dist"
   ],
   "dependencies": {
-    "@outputai/core": "0.11.0",
-    "@outputai/llm": "0.11.0"
+    "@outputai/core": "0.11.1-next.2223fa5.0",
+    "@outputai/llm": "0.11.1-next.2223fa5.0"
   },
   "license": "Apache-2.0",
   "publishConfig": {