@mastra/evals 1.2.1 → 1.2.2-alpha.0

package/dist/scorers/prebuilt/index.cjs

@@ -1,6 +1,6 @@
 'use strict';
 
-var chunkAY4K3J4R_cjs = require('../../chunk-AY4K3J4R.cjs');
+var chunk33T2SZZ2_cjs = require('../../chunk-33T2SZZ2.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 = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const assistantMessage = chunk33T2SZZ2_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 = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const input = chunk33T2SZZ2_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 chunkAY4K3J4R_cjs.roundToTwoDecimals(score * options.scale);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(score * options.scale);
   }).generateReason({
     description: "Reason about the results",
     createPrompt: ({ run, results, score }) => {
       return createReasonPrompt({
-        input: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        input: chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
         score,
         results: results.analyzeStepResult.results,
         scale: options.scale
@@ -466,7 +466,7 @@ function createAnswerSimilarityScorer({
           groundTruth: ""
         });
       }
-      const output = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const output = chunk33T2SZZ2_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 chunkAY4K3J4R_cjs.roundToTwoDecimals(score * mergedOptions.scale);
+    return chunk33T2SZZ2_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 = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const output = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       const groundTruth = typeof run.groundTruth === "string" ? run.groundTruth : JSON.stringify(run.groundTruth);
       return createReasonPrompt2({
         output,
@@ -698,6 +698,10 @@ Example Responses:
 }
 
 // src/scorers/llm/faithfulness/index.ts
+var getToolInvocationContext = (output) => {
+  if (!Array.isArray(output)) return [];
+  return output.filter((message) => message?.role === "assistant").flatMap((message) => message?.content?.toolInvocations ?? []).filter((toolCall) => toolCall.state === "result").map((toolCall) => JSON.stringify(toolCall.result));
+};
 function createFaithfulnessScorer({
   model,
   options
@@ -717,17 +721,14 @@ function createFaithfulnessScorer({
       claims: zod.z.array(zod.z.string())
     }),
     createPrompt: ({ run }) => {
-      const prompt = createFaithfulnessExtractPrompt({ output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
+      const prompt = createFaithfulnessExtractPrompt({ output: chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
       return prompt;
     }
   }).analyze({
     description: "Score the relevance of the statements to the input",
     outputSchema: zod.z.object({ verdicts: zod.z.array(zod.z.object({ verdict: zod.z.string(), reason: zod.z.string() })) }),
     createPrompt: ({ results, run }) => {
-      const assistantMessage = run.output.find(({ role }) => role === "assistant");
-      const context = options?.context ?? assistantMessage?.content?.toolInvocations?.map(
-        (toolCall) => toolCall.state === "result" ? JSON.stringify(toolCall.result) : ""
-      ) ?? [];
+      const context = options?.context ?? getToolInvocationContext(run.output);
       const prompt = createFaithfulnessAnalyzePrompt({
         claims: results.preprocessStepResult?.claims || [],
         context
@@ -741,15 +742,14 @@ function createFaithfulnessScorer({
       return 0;
     }
     const score = supportedClaims / totalClaims * (options?.scale || 1);
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
+    return chunk33T2SZZ2_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: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
-        context: assistantMessage?.content?.toolInvocations?.map((toolCall) => JSON.stringify(toolCall)) || [],
+        input: chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        context: options?.context ?? getToolInvocationContext(run.output),
         score,
         scale: options?.scale || 1,
         verdicts: results.analyzeStepResult?.verdicts || []
@@ -881,13 +881,13 @@ function createBiasScorer({ model, options }) {
     outputSchema: zod.z.object({
       opinions: zod.z.array(zod.z.string())
     }),
-    createPrompt: ({ run }) => createBiasExtractPrompt({ output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" })
+    createPrompt: ({ run }) => createBiasExtractPrompt({ output: chunk33T2SZZ2_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: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        output: chunk33T2SZZ2_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 chunkAY4K3J4R_cjs.roundToTwoDecimals(score * (options?.scale || 1));
+    return chunk33T2SZZ2_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: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
+      const prompt = createHallucinationExtractPrompt({ output: chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "" });
       return prompt;
     }
   }).analyze({
@@ -1145,7 +1145,7 @@ function createHallucinationScorer({
       return 0;
     }
     const score = contradictedStatements / totalStatements * (options?.scale || 1);
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
+    return chunk33T2SZZ2_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: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "",
+        input: chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunk33T2SZZ2_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: chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "",
-        output: chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? ""
+        input: chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "",
+        output: chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? ""
       });
       return prompt;
     }
@@ -1288,7 +1288,7 @@ function createToxicityScorer({
       }
     }
     const score = toxicityCount / numberOfVerdicts;
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score * (options?.scale || 1));
+    return chunk33T2SZZ2_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 } = chunkAY4K3J4R_cjs.extractToolCalls(run.output);
+    const { tools: actualTools, toolCallInfos } = chunk33T2SZZ2_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 = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const userInput = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunk33T2SZZ2_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 chunkAY4K3J4R_cjs.roundToTwoDecimals(appropriateToolCalls / totalToolCalls);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(appropriateToolCalls / totalToolCalls);
   }).generateReason({
     description: "Generate human-readable explanation of tool selection evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const userInput = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const evaluations = results.analyzeStepResult?.evaluations || [];
       const missingTools = results.analyzeStepResult?.missingTools || [];
       return createReasonPrompt3({
@@ -1636,6 +1636,16 @@ var DEFAULT_PENALTIES = {
   MAX_MISSING_CONTEXT_PENALTY: 0.5
   // Maximum 50% penalty for missing context
 };
+var getContext = ({
+  input,
+  output,
+  options
+}) => {
+  if (options.contextExtractor && chunk33T2SZZ2_cjs.isScorerRunInputForAgent(input) && chunk33T2SZZ2_cjs.isScorerRunOutputForAgent(output)) {
+    return options.contextExtractor(input, output);
+  }
+  return options.context ?? [];
+};
 function createContextRelevanceScorerLLM({
   model,
   options
@@ -1659,9 +1669,9 @@ function createContextRelevanceScorerLLM({
     description: "Analyze the relevance and utility of provided context",
     outputSchema: analyzeOutputSchema3,
     createPrompt: ({ run }) => {
-      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;
+      const userQuery = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const context = getContext({ input: run.input, output: run.output, options });
       if (context.length === 0) {
         return createAnalyzePrompt3({
           userQuery,
@@ -1677,7 +1687,7 @@ function createContextRelevanceScorerLLM({
     }
   }).generateScore(({ results, run }) => {
     const evaluations = results.analyzeStepResult?.evaluations || [];
-    const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
+    const context = getContext({ input: run.input, output: run.output, options });
     if (context.length === 0) {
       return 1 * (options.scale || 1);
     }
@@ -1708,12 +1718,12 @@ 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 chunkAY4K3J4R_cjs.roundToTwoDecimals(scaledScore);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(scaledScore);
   }).generateReason({
     description: "Generate human-readable explanation of context relevance evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const context = options.contextExtractor ? options.contextExtractor(run.input, run.output) : options.context;
+      const userQuery = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const context = getContext({ input: run.input, output: run.output, options });
       if (context.length === 0) {
         return `No context was available for evaluation. The agent response was generated without any supporting context. Score: ${score}`;
       }
@@ -1860,6 +1870,16 @@ var contextRelevanceOutputSchema = zod.z.object({
     })
   )
 });
+var getContext2 = ({
+  input,
+  output,
+  options
+}) => {
+  if (options.contextExtractor && chunk33T2SZZ2_cjs.isScorerRunInputForAgent(input) && chunk33T2SZZ2_cjs.isScorerRunOutputForAgent(output)) {
+    return options.contextExtractor(input, output);
+  }
+  return options.context ?? [];
+};
 function createContextPrecisionScorer({
   model,
   options
@@ -1883,9 +1903,9 @@ function createContextPrecisionScorer({
     description: "Evaluate the relevance of each context piece for generating the expected output",
     outputSchema: contextRelevanceOutputSchema,
     createPrompt: ({ run }) => {
-      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;
+      const input = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const output = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const context = getContext2({ input: run.input, output: run.output, options });
       if (context.length === 0) {
         throw new Error("No context available for evaluation");
       }
@@ -1917,13 +1937,13 @@ function createContextPrecisionScorer({
     }
     const map = sumPrecision / relevantCount;
     const score = map * (options.scale || 1);
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(score);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(score);
   }).generateReason({
     description: "Reason about the context precision results",
     createPrompt: ({ run, results, score }) => {
-      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;
+      const input = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const output = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const context = getContext2({ input: run.input, output: run.output, options });
       return createContextPrecisionReasonPrompt({
         input,
         output,
@@ -2177,8 +2197,8 @@ function createNoiseSensitivityScorerLLM({
     description: "Analyze the impact of noise on agent response quality",
     outputSchema: analyzeOutputSchema4,
     createPrompt: ({ run }) => {
-      const originalQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const noisyResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const originalQuery = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const noisyResponse = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       if (!originalQuery || !noisyResponse) {
         throw new Error("Both original query and noisy response are required for evaluation");
       }
@@ -2221,11 +2241,11 @@ function createNoiseSensitivityScorerLLM({
     const majorIssues = analysisResult.majorIssues || [];
     const issuesPenalty = Math.min(majorIssues.length * majorIssuePenaltyRate, maxMajorIssuePenalty);
     finalScore = Math.max(0, finalScore - issuesPenalty);
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(finalScore);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(finalScore);
   }).generateReason({
     description: "Generate human-readable explanation of noise sensitivity evaluation",
     createPrompt: ({ run, results, score }) => {
-      const originalQuery = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const originalQuery = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const analysisResult = results.analyzeStepResult;
       if (!analysisResult) {
         throw new Error("Analysis step failed to produce results for reason generation");
@@ -2550,17 +2570,17 @@ function createPromptAlignmentScorerLLM({
     description: "Analyze prompt-response alignment across multiple dimensions",
     outputSchema: analyzeOutputSchema5,
     createPrompt: ({ run }) => {
-      const userPrompt = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const systemPrompt = chunkAY4K3J4R_cjs.getCombinedSystemPrompt(run.input) ?? "";
-      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
+      const userPrompt = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const systemPrompt = chunk33T2SZZ2_cjs.getCombinedSystemPrompt(run.input) ?? "";
+      const agentResponse = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output) ?? "";
       if (evaluationMode === "user" && !userPrompt) {
         throw new Error("User prompt is required for user prompt alignment scoring");
       }
       if (evaluationMode === "system" && !systemPrompt) {
         throw new Error("System prompt is required for system prompt alignment scoring");
       }
-      if (evaluationMode === "both" && (!userPrompt || !systemPrompt)) {
-        throw new Error("Both user and system prompts are required for combined alignment scoring");
+      if (evaluationMode === "both" && !userPrompt && !systemPrompt) {
+        throw new Error("A user or system prompt is required for combined alignment scoring");
       }
       if (!agentResponse) {
         throw new Error("Agent response is required for prompt alignment scoring");
@@ -2588,12 +2608,12 @@ function createPromptAlignmentScorerLLM({
       weightedScore = userScore * SCORING_WEIGHTS.BOTH.USER_WEIGHT + systemScore * SCORING_WEIGHTS.BOTH.SYSTEM_WEIGHT;
     }
     const finalScore = weightedScore * scale;
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(finalScore);
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(finalScore);
   }).generateReason({
     description: "Generate human-readable explanation of prompt alignment evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userPrompt = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const systemPrompt = chunkAY4K3J4R_cjs.getCombinedSystemPrompt(run.input) ?? "";
+      const userPrompt = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const systemPrompt = chunk33T2SZZ2_cjs.getCombinedSystemPrompt(run.input) ?? "";
       const analysis = results.analyzeStepResult;
       if (!analysis) {
         return `Unable to analyze prompt alignment. Score: ${score}`;
@@ -2798,8 +2818,8 @@ function createTrajectoryAccuracyScorerLLM({
     description: "Analyze the quality and appropriateness of the agent trajectory",
     outputSchema: analyzeOutputSchema6,
     createPrompt: ({ run, results }) => {
-      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
-      const agentResponse = chunkAY4K3J4R_cjs.getAssistantMessageFromRunOutput(run.output.rawOutput) ?? "";
+      const userInput = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const agentResponse = chunk33T2SZZ2_cjs.getAssistantMessageFromRunOutput(run.output.rawOutput) ?? "";
       return createAnalyzePrompt6({
         userInput,
         agentResponse,
@@ -2824,11 +2844,11 @@ function createTrajectoryAccuracyScorerLLM({
     const necessityScore = necessarySteps / totalSteps;
     const orderScore = orderedSteps / totalSteps;
     const score = necessityScore * 0.6 + orderScore * 0.3 - missingPenalty * 0.1;
-    return chunkAY4K3J4R_cjs.roundToTwoDecimals(Math.max(0, Math.min(1, score)));
+    return chunk33T2SZZ2_cjs.roundToTwoDecimals(Math.max(0, Math.min(1, score)));
   }).generateReason({
     description: "Generate human-readable explanation of trajectory evaluation",
     createPrompt: ({ run, results, score }) => {
-      const userInput = chunkAY4K3J4R_cjs.getUserMessageFromRunInput(run.input) ?? "";
+      const userInput = chunk33T2SZZ2_cjs.getUserMessageFromRunInput(run.input) ?? "";
       const stepEvaluations = results.analyzeStepResult?.stepEvaluations || [];
       const missingSteps = results.analyzeStepResult?.missingSteps || [];
       const extraSteps = results.analyzeStepResult?.extraSteps || [];
@@ -2891,18 +2911,18 @@ function createCompletenessScorer() {
     type: "agent"
   }).preprocess(async ({ run }) => {
     const isInputInvalid = !run.input || run.input.inputMessages.some((i) => {
-      const content = chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i);
+      const content = chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i);
       return content === null || content === void 0;
     });
     const isOutputInvalid = !run.output || run.output.some((i) => {
-      const content = chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i);
+      const content = chunk33T2SZZ2_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) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const inputToProcess = input;
     const outputToProcess = output;
     const inputDoc = nlp__default.default(inputToProcess.trim());
@@ -3007,8 +3027,8 @@ 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) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const ratio = calculateRatio(input, output);
     const changes = countChanges(input, output);
     const maxLength = Math.max(input.length, output.length);
@@ -3031,8 +3051,8 @@ 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) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    const output = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const input = run.input?.inputMessages?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const output = run.output?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     if (!input && !output) {
       return {
         result: {
@@ -3085,8 +3105,8 @@ 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) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
-    let processedOutput = run.output.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    let processedInput = run.input?.inputMessages.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    let processedOutput = run.output.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     if (ignoreCase) {
       processedInput = processedInput.toLowerCase();
       processedOutput = processedOutput.toLowerCase();
@@ -3116,7 +3136,7 @@ function createToneScorer(config = {}) {
     type: "agent"
   }).preprocess(async ({ run }) => {
     const sentiment = new Sentiment__default.default();
-    const agentMessage = run.output?.map((i) => chunkAY4K3J4R_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
+    const agentMessage = run.output?.map((i) => chunk33T2SZZ2_cjs.getTextContentFromMastraDBMessage(i)).join(", ") || "";
     const responseSentiment = sentiment.analyze(agentMessage);
     if (referenceTone) {
       const referenceSentiment = sentiment.analyze(referenceTone);
@@ -3203,7 +3223,7 @@ function createToolCallAccuracyScorerCode(options) {
     if (isInputInvalid || isOutputInvalid) {
       throw new Error("Input and output messages cannot be null or empty");
     }
-    const { tools: actualTools, toolCallInfos } = chunkAY4K3J4R_cjs.extractToolCalls(run.output);
+    const { tools: actualTools, toolCallInfos } = chunk33T2SZZ2_cjs.extractToolCalls(run.output);
     const correctToolCalled = expectedTool ? strictMode ? actualTools.length === 1 && actualTools[0] === expectedTool : actualTools.includes(expectedTool) : false;
     return {
       expectedTool,
@@ -3278,7 +3298,7 @@ function createTrajectoryAccuracyScorerCode(options = {}) {
     const itemExpectation = run.expectedTrajectory;
     const effectiveOrdering = itemExpectation?.ordering ?? ordering;
     const effectiveAllowRepeated = itemExpectation?.allowRepeatedSteps ?? allowRepeatedSteps;
-    const comparison = chunkAY4K3J4R_cjs.compareTrajectories(
+    const comparison = chunk33T2SZZ2_cjs.compareTrajectories(
       actualTrajectory,
       { steps: resolvedExpectedSteps },
       {
@@ -3336,7 +3356,7 @@ function evaluateNestedExpectations(expectedSteps, actualSteps, weights = { accu
     const childConfig = expectedStep.children;
     let accuracy;
     if (childConfig.steps && childConfig.steps.length > 0) {
-      accuracy = chunkAY4K3J4R_cjs.compareTrajectories(
+      accuracy = chunk33T2SZZ2_cjs.compareTrajectories(
         childTrajectory,
         { steps: childConfig.steps },
         {
@@ -3346,18 +3366,18 @@ function evaluateNestedExpectations(expectedSteps, actualSteps, weights = { accu
       );
     }
     const hasEfficiencyConfig = childConfig.maxSteps !== void 0 || childConfig.maxTotalTokens !== void 0 || childConfig.maxTotalDurationMs !== void 0 || childConfig.noRedundantCalls !== void 0;
-    const efficiency = hasEfficiencyConfig ? chunkAY4K3J4R_cjs.checkTrajectoryEfficiency(childTrajectory, {
+    const efficiency = hasEfficiencyConfig ? chunk33T2SZZ2_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 ? chunkAY4K3J4R_cjs.checkTrajectoryBlacklist(childTrajectory, {
+    const blacklist = hasBlacklistConfig ? chunk33T2SZZ2_cjs.checkTrajectoryBlacklist(childTrajectory, {
       blacklistedTools: childConfig.blacklistedTools,
       blacklistedSequences: childConfig.blacklistedSequences
     }) : void 0;
-    const toolFailures = chunkAY4K3J4R_cjs.analyzeToolFailures(childTrajectory, {
+    const toolFailures = chunk33T2SZZ2_cjs.analyzeToolFailures(childTrajectory, {
       maxRetriesPerTool: childConfig.maxRetriesPerTool ?? 2
     });
     const nested = childConfig.steps ? evaluateNestedExpectations(childConfig.steps, actualStep.children, weights) : [];
@@ -3422,7 +3442,7 @@ function createTrajectoryScorerCode(options = {}) {
     }
     let accuracy;
     if (config.steps && config.steps.length > 0) {
-      accuracy = chunkAY4K3J4R_cjs.compareTrajectories(
+      accuracy = chunk33T2SZZ2_cjs.compareTrajectories(
         actualTrajectory,
         { steps: config.steps },
         {
@@ -3432,18 +3452,18 @@ function createTrajectoryScorerCode(options = {}) {
       );
     }
     const hasEfficiencyConfig = config.maxSteps !== void 0 || config.maxTotalTokens !== void 0 || config.maxTotalDurationMs !== void 0 || config.noRedundantCalls !== void 0;
-    const efficiency = hasEfficiencyConfig ? chunkAY4K3J4R_cjs.checkTrajectoryEfficiency(actualTrajectory, {
+    const efficiency = hasEfficiencyConfig ? chunk33T2SZZ2_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 ? chunkAY4K3J4R_cjs.checkTrajectoryBlacklist(actualTrajectory, {
+    const blacklist = hasBlacklistConfig ? chunk33T2SZZ2_cjs.checkTrajectoryBlacklist(actualTrajectory, {
       blacklistedTools: config.blacklistedTools,
       blacklistedSequences: config.blacklistedSequences
     }) : void 0;
-    const toolFailures = chunkAY4K3J4R_cjs.analyzeToolFailures(actualTrajectory, {
+    const toolFailures = chunk33T2SZZ2_cjs.analyzeToolFailures(actualTrajectory, {
       maxRetriesPerTool: config.maxRetriesPerTool ?? 2
     });
     const nested = config.steps && config.steps.length > 0 ? evaluateNestedExpectations(config.steps, actualTrajectory.steps, w) : void 0;