@mastra/evals 1.2.0-alpha.0 → 1.2.0

package/dist/scorers/prebuilt/index.cjs

@@ -1,6 +1,6 @@
 'use strict';
 
-var chunkXRUR5PBK_cjs = require('../../chunk-XRUR5PBK.cjs');
+var chunkAY4K3J4R_cjs = require('../../chunk-AY4K3J4R.cjs');
 var evals = require('@mastra/core/evals');
 var zod = require('zod');
 var nlp = require('compromise');
@@ -239,14 +239,14 @@ function createAnswerRelevancyScorer({
     description: "Extract relevant statements from the LLM output",
     outputSchema: extractOutputSchema,
     createPrompt: ({ run }) => {
-      const assistantMessage = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const assistantMessage = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       return createExtractPrompt(assistantMessage);
     }
   }).analyze({
     description: "Score the relevance of the statements to the input",
     outputSchema: zod.z.object({ results: zod.z.array(zod.z.object({ result: zod.z.string(), reason: zod.z.string() })) }),
     createPrompt: ({ run, results }) => {
-      const input = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const input = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
       return createScorePrompt(JSON.stringify(input), results.preprocessStepResult?.statements || []);
     }
   }).generateScore(({ results }) => {
@@ -263,13 +263,13 @@ function createAnswerRelevancyScorer({
       }
     }
     const score = relevancyCount / numberOfResults;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score * options.scale);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score * options.scale);
   }).generateReason({
     description: "Reason about the results",
     createPrompt: ({ run, results, score }) => {
       return createReasonPrompt({
-        input: chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        input: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
         score,
         results: results.analyzeStepResult.results,
         scale: options.scale
@@ -466,7 +466,7 @@ function createAnswerSimilarityScorer({
           groundTruth: ""
         });
       }
-      const output = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const output = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const groundTruth = typeof run.groundTruth === "string" ? run.groundTruth : JSON.stringify(run.groundTruth);
       return createExtractPrompt2({
         output,
@@ -524,14 +524,14 @@ function createAnswerSimilarityScorer({
     );
     score -= extraInfoPenalty;
     score = Math.max(0, Math.min(1, score));
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score * mergedOptions.scale);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score * mergedOptions.scale);
   }).generateReason({
     description: "Generate explanation of similarity score",
     createPrompt: ({ run, results, score }) => {
       if (!run.groundTruth) {
         return "No ground truth was provided for comparison. Score is 0 by default.";
       }
-      const output = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const output = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const groundTruth = typeof run.groundTruth === "string" ? run.groundTruth : JSON.stringify(run.groundTruth);
       return createReasonPrompt2({
         output,
@@ -717,7 +717,7 @@ function createFaithfulnessScorer({
       claims: zod.z.array(zod.z.string())
     }),
     createPrompt: ({ run }) => {
-      const prompt = createFaithfulnessExtractPrompt({ output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
+      const prompt = createFaithfulnessExtractPrompt({ output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
       return prompt;
     }
   }).analyze({
@@ -741,14 +741,14 @@ function createFaithfulnessScorer({
       return 0;
     }
     const score = supportedClaims / totalClaims * (options?.scale || 1);
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
   }).generateReason({
     description: "Reason about the results",
     createPrompt: ({ run, results, score }) => {
       const assistantMessage = run.output.find(({ role }) => role === "assistant");
       const prompt = createFaithfulnessReasonPrompt({
-        input: chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        input: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
         context: assistantMessage?.content?.toolInvocations?.map((toolCall) => JSON.stringify(toolCall)) || [],
         score,
         scale: options?.scale || 1,
@@ -881,13 +881,13 @@ function createBiasScorer({ model, options }) {
     outputSchema: zod.z.object({
       opinions: zod.z.array(zod.z.string())
     }),
-    createPrompt: ({ run }) => createBiasExtractPrompt({ output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" })
+    createPrompt: ({ run }) => createBiasExtractPrompt({ output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" })
   }).analyze({
     description: "Score the relevance of the statements to the input",
     outputSchema: zod.z.object({ results: zod.z.array(zod.z.object({ result: zod.z.string(), reason: zod.z.string() })) }),
     createPrompt: ({ run, results }) => {
       const prompt = createBiasAnalyzePrompt({
-        output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
         opinions: results.preprocessStepResult?.opinions || []
       });
       return prompt;
@@ -898,7 +898,7 @@ function createBiasScorer({ model, options }) {
     }
     const biasedVerdicts = results.analyzeStepResult.results.filter((v) => v.result.toLowerCase() === "yes");
     const score = biasedVerdicts.length / results.analyzeStepResult.results.length;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score * (options?.scale || 1));
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score * (options?.scale || 1));
   }).generateReason({
     description: "Reason about the results",
     createPrompt: ({ score, results }) => {
@@ -1117,7 +1117,7 @@ function createHallucinationScorer({
       claims: zod.z.array(zod.z.string())
     }),
     createPrompt: ({ run }) => {
-      const prompt = createHallucinationExtractPrompt({ output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
+      const prompt = createHallucinationExtractPrompt({ output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
       return prompt;
     }
   }).analyze({
@@ -1145,7 +1145,7 @@ function createHallucinationScorer({
       return 0;
     }
     const score = contradictedStatements / totalStatements * (options?.scale || 1);
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
   }).generateReason({
     description: "Reason about the results",
     createPrompt: async ({ run, results, score }) => {
@@ -1156,8 +1156,8 @@ function createHallucinationScorer({
         context = options?.context ?? [];
       }
       const prompt = createHallucinationReasonPrompt({
-        input: chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        input: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
         context,
         score,
         scale: options?.scale || 1,
@@ -1271,8 +1271,8 @@ function createToxicityScorer({
     outputSchema: zod.z.object({ verdicts: zod.z.array(zod.z.object({ verdict: zod.z.string(), reason: zod.z.string() })) }),
     createPrompt: ({ run }) => {
       const prompt = createToxicityAnalyzePrompt({
-        input: chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? ""
+        input: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? ""
       });
       return prompt;
     }
@@ -1288,7 +1288,7 @@ function createToxicityScorer({
       }
     }
     const score = toxicityCount / numberOfVerdicts;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score * (options?.scale || 1));
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score * (options?.scale || 1));
   }).generateReason({
     description: "Reason about the results",
     createPrompt: ({ results, score }) => {
@@ -1422,7 +1422,7 @@ function createToolCallAccuracyScorerLLM({ model, availableTools }) {
     if (isInputInvalid || isOutputInvalid) {
       throw new Error("Input and output messages cannot be null or empty");
     }
-    const { tools: actualTools, toolCallInfos } = chunkXRUR5PBK_cjs.extractToolCalls(run.output);
+    const { tools: actualTools, toolCallInfos } = chunkAY4K3J4R_cjs.extractToolCalls(run.output);
     return {
       actualTools,
       hasToolCalls: actualTools.length > 0,
@@ -1432,8 +1432,8 @@ function createToolCallAccuracyScorerLLM({ model, availableTools }) {
     description: "Analyze the appropriateness of tool selections",
     outputSchema: analyzeOutputSchema2,
     createPrompt: ({ run, results }) => {
-      const userInput = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const agentResponse = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const toolsCalled = results.preprocessStepResult?.actualTools || [];
       return createAnalyzePrompt2({
         userInput,
@@ -1450,11 +1450,11 @@ function createToolCallAccuracyScorerLLM({ model, availableTools }) {
     }
     const appropriateToolCalls = evaluations.filter((e) => e.wasAppropriate).length;
     const totalToolCalls = evaluations.length;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(appropriateToolCalls / totalToolCalls);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(appropriateToolCalls / totalToolCalls);
   }).generateReason({
     description: "Generate human-readable explanation of tool selection evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userInput = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const evaluations = results.analyzeStepResult?.evaluations || [];
       const missingTools = results.analyzeStepResult?.missingTools || [];
       return createReasonPrompt3({
@@ -1659,8 +1659,8 @@ function createContextRelevanceScorerLLM({
     description: "Analyze the relevance and utility of provided context",
     outputSchema: analyzeOutputSchema3,
     createPrompt: ({ run }) => {
-      const userQuery = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const agentResponse = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const userQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
       if (context.length === 0) {
         return createAnalyzePrompt3({
@@ -1708,11 +1708,11 @@ function createContextRelevanceScorerLLM({
     const missingContextPenalty = Math.min(missingContext.length * missingPenaltyRate, maxMissingPenalty);
     const finalScore = Math.max(0, relevanceScore - usagePenalty - missingContextPenalty);
     const scaledScore = finalScore * (options.scale || 1);
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(scaledScore);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(scaledScore);
   }).generateReason({
     description: "Generate human-readable explanation of context relevance evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userQuery = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const userQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
       if (context.length === 0) {
         return `No context was available for evaluation. The agent response was generated without any supporting context. Score: ${score}`;
@@ -1883,8 +1883,8 @@ function createContextPrecisionScorer({
     description: "Evaluate the relevance of each context piece for generating the expected output",
     outputSchema: contextRelevanceOutputSchema,
     createPrompt: ({ run }) => {
-      const input = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const output = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const input = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const output = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
       if (context.length === 0) {
         throw new Error("No context available for evaluation");
@@ -1917,12 +1917,12 @@ function createContextPrecisionScorer({
     }
     const map = sumPrecision / relevantCount;
     const score = map * (options.scale || 1);
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(score);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
   }).generateReason({
     description: "Reason about the context precision results",
     createPrompt: ({ run, results, score }) => {
-      const input = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const output = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const input = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const output = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
       return createContextPrecisionReasonPrompt({
         input,
@@ -2177,8 +2177,8 @@ function createNoiseSensitivityScorerLLM({
     description: "Analyze the impact of noise on agent response quality",
     outputSchema: analyzeOutputSchema4,
     createPrompt: ({ run }) => {
-      const originalQuery = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const noisyResponse = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const originalQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const noisyResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       if (!originalQuery || !noisyResponse) {
         throw new Error("Both original query and noisy response are required for evaluation");
       }
@@ -2221,11 +2221,11 @@ function createNoiseSensitivityScorerLLM({
     const majorIssues = analysisResult.majorIssues || [];
     const issuesPenalty = Math.min(majorIssues.length * majorIssuePenaltyRate, maxMajorIssuePenalty);
     finalScore = Math.max(0, finalScore - issuesPenalty);
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(finalScore);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(finalScore);
   }).generateReason({
     description: "Generate human-readable explanation of noise sensitivity evaluation",
     createPrompt: ({ run, results, score }) => {
-      const originalQuery = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const originalQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const analysisResult = results.analyzeStepResult;
       if (!analysisResult) {
         throw new Error("Analysis step failed to produce results for reason generation");
@@ -2550,9 +2550,9 @@ function createPromptAlignmentScorerLLM({
     description: "Analyze prompt-response alignment across multiple dimensions",
     outputSchema: analyzeOutputSchema5,
     createPrompt: ({ run }) => {
-      const userPrompt = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const systemPrompt = chunkXRUR5PBK_cjs.getCombinedSystemPrompt(run.input) ?? "";
-      const agentResponse = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const userPrompt = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const systemPrompt = chunkAY4K3J4R_cjs.getCombinedSystemPrompt(run.input) ?? "";
+      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       if (evaluationMode === "user" && !userPrompt) {
         throw new Error("User prompt is required for user prompt alignment scoring");
       }
@@ -2588,12 +2588,12 @@ function createPromptAlignmentScorerLLM({
       weightedScore = userScore * SCORING_WEIGHTS.BOTH.USER_WEIGHT + systemScore * SCORING_WEIGHTS.BOTH.SYSTEM_WEIGHT;
     }
     const finalScore = weightedScore * scale;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(finalScore);
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(finalScore);
   }).generateReason({
     description: "Generate human-readable explanation of prompt alignment evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userPrompt = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const systemPrompt = chunkXRUR5PBK_cjs.getCombinedSystemPrompt(run.input) ?? "";
+      const userPrompt = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const systemPrompt = chunkAY4K3J4R_cjs.getCombinedSystemPrompt(run.input) ?? "";
       const analysis = results.analyzeStepResult;
       if (!analysis) {
         return `Unable to analyze prompt alignment. Score: ${score}`;
@@ -2744,7 +2744,8 @@ function formatExpectedSteps(steps, indent = 0) {
   const prefix = "  ".repeat(indent);
   return steps.map((step, i) => {
     const typeStr = step.stepType ? `[${step.stepType}] ` : "";
-    const dataStr = step.data ? ` (data: ${JSON.stringify(step.data)})` : "";
+    const { name: _, stepType: _t, children: _c, ...fields } = step;
+    const dataStr = Object.keys(fields).length > 0 ? ` (${JSON.stringify(fields)})` : "";
     let line = `${prefix}${i + 1}. ${typeStr}${step.name}${dataStr}`;
     if (step.children?.steps && step.children.steps.length > 0) {
       line += `
@@ -2773,22 +2774,15 @@ function createTrajectoryAccuracyScorerLLM({
       if (Array.isArray(staticExpectedTrajectory)) {
         expectedSteps = staticExpectedTrajectory;
       } else {
-        expectedSteps = staticExpectedTrajectory.steps.map((s) => {
-          const result = { name: s.name, stepType: s.stepType };
-          const data = {};
-          if ((s.stepType === "tool_call" || s.stepType === "mcp_tool_call") && s.toolArgs !== void 0)
-            data.input = s.toolArgs;
-          if ((s.stepType === "tool_call" || s.stepType === "mcp_tool_call") && s.toolResult !== void 0)
-            data.output = s.toolResult;
-          if (s.stepType === "workflow_step" && s.output !== void 0) data.output = s.output;
-          if (Object.keys(data).length > 0) result.data = data;
-          if (s.children && s.children.length > 0) {
-            result.children = {
-              steps: s.children.map((c) => ({ name: c.name, stepType: c.stepType }))
-            };
+        const toExpectedStep = (s) => {
+          const { durationMs: _, metadata: _m, children, ...rest } = s;
+          const result = rest;
+          if (children && children.length > 0) {
+            result.children = { steps: children.map(toExpectedStep) };
           }
           return result;
-        });
+        };
+        expectedSteps = staticExpectedTrajectory.steps.map(toExpectedStep);
       }
     } else if (run.expectedTrajectory) {
       const expectation = run.expectedTrajectory;
@@ -2804,8 +2798,8 @@ function createTrajectoryAccuracyScorerLLM({
     description: "Analyze the quality and appropriateness of the agent trajectory",
     outputSchema: analyzeOutputSchema6,
     createPrompt: ({ run, results }) => {
-      const userInput = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const agentResponse = chunkXRUR5PBK_cjs.getAssistantMessageFromRunOutput(run.output.rawOutput) ?? "";
+      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output.rawOutput) ?? "";
       return createAnalyzePrompt6({
         userInput,
         agentResponse,
@@ -2830,11 +2824,11 @@ function createTrajectoryAccuracyScorerLLM({
     const necessityScore = necessarySteps / totalSteps;
     const orderScore = orderedSteps / totalSteps;
     const score = necessityScore * 0.6 + orderScore * 0.3 - missingPenalty * 0.1;
-    return chunkXRUR5PBK_cjs.roundToTwoDecimals(Math.max(0, Math.min(1, score)));
+    return chunkAY4K3J4R_cjs.roundToTwoDecimals(Math.max(0, Math.min(1, score)));
   }).generateReason({
     description: "Generate human-readable explanation of trajectory evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userInput = chunkXRUR5PBK_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const stepEvaluations = results.analyzeStepResult?.stepEvaluations || [];
       const missingSteps = results.analyzeStepResult?.missingSteps || [];
       const extraSteps = results.analyzeStepResult?.extraSteps || [];
@@ -2897,18 +2891,18 @@ function createCompletenessScorer() {
     type: "agent"
   }).preprocess(async ({ run }) => {
     const isInputInvalid = !run.input || run.input.inputMessages.some((i) => {
-      const content = chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i);
+      const content = chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i);
       return content === null || content === void 0;
     });
     const isOutputInvalid = !run.output || run.output.some((i) => {
-      const content = chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i);
+      const content = chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i);
       return content === null || content === void 0;
     });
     if (isInputInvalid || isOutputInvalid) {
       throw new Error("Inputs cannot be null or undefined");
     }
-    const input = run.input?.inputMessages.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const inputToProcess = input;
     const outputToProcess = output;
     const inputDoc = nlp__default.default(inputToProcess.trim());
@@ -3013,8 +3007,8 @@ function createTextualDifferenceScorer() {
     description: "Calculate textual difference between input and output using sequence matching algorithms.",
     type: "agent"
   }).preprocess(async ({ run }) => {
-    const input = run.input?.inputMessages?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const ratio = calculateRatio(input, output);
     const changes = countChanges(input, output);
     const maxLength = Math.max(input.length, output.length);
@@ -3037,8 +3031,8 @@ function createKeywordCoverageScorer() {
     description: 'Leverage the nlp method from "compromise" to extract elements from the input and output and calculate the coverage.',
     type: "agent"
   }).preprocess(async ({ run }) => {
-    const input = run.input?.inputMessages?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     if (!input && !output) {
       return {
         result: {
@@ -3091,8 +3085,8 @@ function createContentSimilarityScorer({ ignoreCase, ignoreWhitespace } = { igno
     description: "Calculates content similarity between input and output messages using string comparison algorithms.",
     type: "agent"
   }).preprocess(async ({ run }) => {
-    let processedInput = run.input?.inputMessages.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    let processedOutput = run.output.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    let processedInput = run.input?.inputMessages.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    let processedOutput = run.output.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     if (ignoreCase) {
       processedInput = processedInput.toLowerCase();
       processedOutput = processedOutput.toLowerCase();
@@ -3122,7 +3116,7 @@ function createToneScorer(config = {}) {
     type: "agent"
   }).preprocess(async ({ run }) => {
     const sentiment = new Sentiment__default.default();
-    const agentMessage = run.output?.map((i) => chunkXRUR5PBK_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const agentMessage = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const responseSentiment = sentiment.analyze(agentMessage);
     if (referenceTone) {
       const referenceSentiment = sentiment.analyze(referenceTone);
@@ -3209,7 +3203,7 @@ function createToolCallAccuracyScorerCode(options) {
     if (isInputInvalid || isOutputInvalid) {
       throw new Error("Input and output messages cannot be null or empty");
     }
-    const { tools: actualTools, toolCallInfos } = chunkXRUR5PBK_cjs.extractToolCalls(run.output);
+    const { tools: actualTools, toolCallInfos } = chunkAY4K3J4R_cjs.extractToolCalls(run.output);
     const correctToolCalled = expectedTool ? strictMode ? actualTools.length === 1 && actualTools[0] === expectedTool : actualTools.includes(expectedTool) : false;
     return {
       expectedTool,
@@ -3235,18 +3229,11 @@ function createToolCallAccuracyScorerCode(options) {
   });
 }
 function trajectoryStepToExpectedStep(step) {
-  const result = { name: step.name, stepType: step.stepType };
-  const data = {};
-  if (step.stepType === "tool_call" || step.stepType === "mcp_tool_call") {
-    if (step.toolArgs !== void 0) data.input = step.toolArgs;
-    if (step.toolResult !== void 0) data.output = step.toolResult;
-  } else if (step.stepType === "workflow_step") {
-    if (step.output !== void 0) data.output = step.output;
-  }
-  if (Object.keys(data).length > 0) result.data = data;
-  if (step.children && step.children.length > 0) {
+  const { durationMs: _, metadata: _m, children, ...rest } = step;
+  const result = rest;
+  if (children && children.length > 0) {
     result.children = {
-      steps: step.children.map(trajectoryStepToExpectedStep)
+      steps: children.map(trajectoryStepToExpectedStep)
     };
   }
   return result;
@@ -3257,15 +3244,14 @@ function expectationToExpectedSteps(expectation) {
 }
 function createTrajectoryAccuracyScorerCode(options = {}) {
   const { expectedTrajectory: staticExpectedTrajectory, comparisonOptions = {} } = options;
-  const { ordering, strictOrder, compareStepData = false, allowRepeatedSteps = true } = comparisonOptions;
-  const resolvedOrdering = ordering ?? (strictOrder ? "strict" : "relaxed");
+  const { ordering = "relaxed", allowRepeatedSteps = true } = comparisonOptions;
   const staticExpectedSteps = staticExpectedTrajectory ? Array.isArray(staticExpectedTrajectory) && staticExpectedTrajectory.length > 0 && !("steps" in staticExpectedTrajectory[0] || false) ? staticExpectedTrajectory : "steps" in staticExpectedTrajectory ? staticExpectedTrajectory.steps.map(trajectoryStepToExpectedStep) : void 0 : void 0;
   const getDescription = () => {
     if (staticExpectedSteps) {
       const expectedStepNames = staticExpectedSteps.map((s) => s.name).join(" \u2192 ");
-      return `Evaluates whether the trajectory matches the expected path: [${expectedStepNames}] (${resolvedOrdering} ordering)`;
+      return `Evaluates whether the trajectory matches the expected path: [${expectedStepNames}] (${ordering} ordering)`;
     }
-    return `Evaluates trajectory accuracy against expected trajectory from dataset items (${resolvedOrdering} ordering)`;
+    return `Evaluates trajectory accuracy against expected trajectory from dataset items (${ordering} ordering)`;
   };
   return evals.createScorer({
     id: "code-trajectory-accuracy-scorer",
@@ -3290,15 +3276,13 @@ function createTrajectoryAccuracyScorerCode(options = {}) {
       };
     }
     const itemExpectation = run.expectedTrajectory;
-    const effectiveOrdering = itemExpectation?.ordering ?? resolvedOrdering;
-    const effectiveCompareData = itemExpectation?.compareStepData ?? compareStepData;
+    const effectiveOrdering = itemExpectation?.ordering ?? ordering;
     const effectiveAllowRepeated = itemExpectation?.allowRepeatedSteps ?? allowRepeatedSteps;
-    const comparison = chunkXRUR5PBK_cjs.compareTrajectories(
+    const comparison = chunkAY4K3J4R_cjs.compareTrajectories(
       actualTrajectory,
       { steps: resolvedExpectedSteps },
       {
         ordering: effectiveOrdering,
-        compareStepData: effectiveCompareData,
         allowRepeatedSteps: effectiveAllowRepeated
       }
     );
@@ -3317,7 +3301,7 @@ function createTrajectoryAccuracyScorerCode(options = {}) {
     return preprocessResult.comparison.score;
   });
 }
-function evaluateNestedExpectations(expectedSteps, actualSteps) {
+function evaluateNestedExpectations(expectedSteps, actualSteps, weights = { accuracy: 0.4, efficiency: 0.3, toolFailures: 0.2, blacklist: 0.1 }) {
   const results = [];
   const matchedIndices = /* @__PURE__ */ new Set();
   for (const expectedStep of expectedSteps) {
@@ -3352,47 +3336,47 @@ function evaluateNestedExpectations(expectedSteps, actualSteps) {
     const childConfig = expectedStep.children;
     let accuracy;
     if (childConfig.steps && childConfig.steps.length > 0) {
-      accuracy = chunkXRUR5PBK_cjs.compareTrajectories(
+      accuracy = chunkAY4K3J4R_cjs.compareTrajectories(
         childTrajectory,
         { steps: childConfig.steps },
         {
           ordering: childConfig.ordering ?? "relaxed",
-          compareStepData: childConfig.compareStepData ?? false,
           allowRepeatedSteps: childConfig.allowRepeatedSteps ?? true
         }
       );
     }
     const hasEfficiencyConfig = childConfig.maxSteps !== void 0 || childConfig.maxTotalTokens !== void 0 || childConfig.maxTotalDurationMs !== void 0 || childConfig.noRedundantCalls !== void 0;
-    const efficiency = hasEfficiencyConfig ? chunkXRUR5PBK_cjs.checkTrajectoryEfficiency(childTrajectory, {
+    const efficiency = hasEfficiencyConfig ? chunkAY4K3J4R_cjs.checkTrajectoryEfficiency(childTrajectory, {
       maxSteps: childConfig.maxSteps,
       maxTotalTokens: childConfig.maxTotalTokens,
       maxTotalDurationMs: childConfig.maxTotalDurationMs,
       noRedundantCalls: childConfig.noRedundantCalls ?? true
     }) : void 0;
     const hasBlacklistConfig = childConfig.blacklistedTools && childConfig.blacklistedTools.length > 0 || childConfig.blacklistedSequences && childConfig.blacklistedSequences.length > 0;
-    const blacklist = hasBlacklistConfig ? chunkXRUR5PBK_cjs.checkTrajectoryBlacklist(childTrajectory, {
+    const blacklist = hasBlacklistConfig ? chunkAY4K3J4R_cjs.checkTrajectoryBlacklist(childTrajectory, {
       blacklistedTools: childConfig.blacklistedTools,
       blacklistedSequences: childConfig.blacklistedSequences
     }) : void 0;
-    const toolFailures = chunkXRUR5PBK_cjs.analyzeToolFailures(childTrajectory, {
+    const toolFailures = chunkAY4K3J4R_cjs.analyzeToolFailures(childTrajectory, {
       maxRetriesPerTool: childConfig.maxRetriesPerTool ?? 2
     });
-    const nested = childConfig.steps ? evaluateNestedExpectations(childConfig.steps, actualStep.children) : [];
+    const nested = childConfig.steps ? evaluateNestedExpectations(childConfig.steps, actualStep.children, weights) : [];
     const scores = [];
-    if (accuracy) scores.push({ weight: 0.4, value: accuracy.score });
-    if (efficiency) scores.push({ weight: 0.3, value: efficiency.score });
-    if (toolFailures && toolFailures.patterns.length > 0) scores.push({ weight: 0.2, value: toolFailures.score });
+    if (accuracy) scores.push({ weight: weights.accuracy, value: accuracy.score });
+    if (efficiency) scores.push({ weight: weights.efficiency, value: efficiency.score });
+    if (toolFailures && toolFailures.patterns.length > 0)
+      scores.push({ weight: weights.toolFailures, value: toolFailures.score });
     if (blacklist) {
       if (blacklist.score === 0) {
         results.push({ stepName: expectedStep.name, score: 0, accuracy, efficiency, blacklist, toolFailures, nested });
         continue;
       }
-      scores.push({ weight: 0.1, value: blacklist.score });
+      scores.push({ weight: weights.blacklist, value: blacklist.score });
     }
     let levelScore = 1;
     if (scores.length > 0) {
       const totalWeight = scores.reduce((sum, s) => sum + s.weight, 0);
-      levelScore = scores.reduce((sum, s) => sum + s.weight / totalWeight * s.value, 0);
+      levelScore = totalWeight > 0 ? scores.reduce((sum, s) => sum + s.weight / totalWeight * s.value, 0) : 1;
     }
     let finalScore = levelScore;
     if (nested.length > 0) {
@@ -3417,7 +3401,13 @@ function evaluateNestedExpectations(expectedSteps, actualSteps) {
   return results;
 }
 function createTrajectoryScorerCode(options = {}) {
-  const { defaults = {} } = options;
+  const { defaults = {}, weights: userWeights = {} } = options;
+  const w = {
+    accuracy: Math.max(0, userWeights.accuracy ?? 0.4),
+    efficiency: Math.max(0, userWeights.efficiency ?? 0.3),
+    toolFailures: Math.max(0, userWeights.toolFailures ?? 0.2),
+    blacklist: Math.max(0, userWeights.blacklist ?? 0.1)
+  };
   return evals.createScorer({
     id: "code-trajectory-scorer",
     name: "Trajectory Scorer",
@@ -3432,32 +3422,31 @@ function createTrajectoryScorerCode(options = {}) {
     }
     let accuracy;
     if (config.steps && config.steps.length > 0) {
-      accuracy = chunkXRUR5PBK_cjs.compareTrajectories(
+      accuracy = chunkAY4K3J4R_cjs.compareTrajectories(
         actualTrajectory,
         { steps: config.steps },
         {
           ordering: config.ordering ?? "relaxed",
-          compareStepData: config.compareStepData ?? false,
           allowRepeatedSteps: config.allowRepeatedSteps ?? true
         }
       );
     }
     const hasEfficiencyConfig = config.maxSteps !== void 0 || config.maxTotalTokens !== void 0 || config.maxTotalDurationMs !== void 0 || config.noRedundantCalls !== void 0;
-    const efficiency = hasEfficiencyConfig ? chunkXRUR5PBK_cjs.checkTrajectoryEfficiency(actualTrajectory, {
+    const efficiency = hasEfficiencyConfig ? chunkAY4K3J4R_cjs.checkTrajectoryEfficiency(actualTrajectory, {
       maxSteps: config.maxSteps,
       maxTotalTokens: config.maxTotalTokens,
       maxTotalDurationMs: config.maxTotalDurationMs,
       noRedundantCalls: config.noRedundantCalls ?? true
     }) : void 0;
     const hasBlacklistConfig = config.blacklistedTools && config.blacklistedTools.length > 0 || config.blacklistedSequences && config.blacklistedSequences.length > 0;
-    const blacklist = hasBlacklistConfig ? chunkXRUR5PBK_cjs.checkTrajectoryBlacklist(actualTrajectory, {
+    const blacklist = hasBlacklistConfig ? chunkAY4K3J4R_cjs.checkTrajectoryBlacklist(actualTrajectory, {
       blacklistedTools: config.blacklistedTools,
       blacklistedSequences: config.blacklistedSequences
     }) : void 0;
-    const toolFailures = chunkXRUR5PBK_cjs.analyzeToolFailures(actualTrajectory, {
+    const toolFailures = chunkAY4K3J4R_cjs.analyzeToolFailures(actualTrajectory, {
       maxRetriesPerTool: config.maxRetriesPerTool ?? 2
     });
-    const nested = config.steps && config.steps.length > 0 ? evaluateNestedExpectations(config.steps, actualTrajectory.steps) : void 0;
+    const nested = config.steps && config.steps.length > 0 ? evaluateNestedExpectations(config.steps, actualTrajectory.steps, w) : void 0;
     return {
       accuracy,
       efficiency,
@@ -3473,16 +3462,16 @@ function createTrajectoryScorerCode(options = {}) {
     }
     const scores = [];
     if (accuracy) {
-      scores.push({ weight: 0.4, value: accuracy.score });
+      scores.push({ weight: w.accuracy, value: accuracy.score });
     }
     if (efficiency) {
-      scores.push({ weight: 0.3, value: efficiency.score });
+      scores.push({ weight: w.efficiency, value: efficiency.score });
     }
     if (toolFailures && toolFailures.patterns.length > 0) {
-      scores.push({ weight: 0.2, value: toolFailures.score });
+      scores.push({ weight: w.toolFailures, value: toolFailures.score });
     }
     if (blacklist) {
-      scores.push({ weight: 0.1, value: blacklist.score });
+      scores.push({ weight: w.blacklist, value: blacklist.score });
     }
     if (scores.length === 0 && !nested) {
       return 1;
@@ -3490,7 +3479,7 @@ function createTrajectoryScorerCode(options = {}) {
     let levelScore = 1;
     if (scores.length > 0) {
       const totalWeight = scores.reduce((sum, s) => sum + s.weight, 0);
-      levelScore = scores.reduce((sum, s) => sum + s.weight / totalWeight * s.value, 0);
+      levelScore = totalWeight > 0 ? scores.reduce((sum, s) => sum + s.weight / totalWeight * s.value, 0) : 1;
     }
     if (nested && nested.length > 0) {
       const hasNestedBlacklistViolation = nested.some((r) => r.blacklist && r.blacklist.score === 0);