@agentv/core 3.4.0 → 3.5.0

package/dist/index.js

@@ -5,7 +5,6 @@ import {
   extractLastAssistantContent,
   fileExists,
   findGitRoot,
-  getHitCount,
   isAgentProvider,
   isEvaluatorKind,
   isJsonObject,
@@ -17,10 +16,10 @@ import {
   readTextFile,
   resolveFileReference,
   resolveTargetDefinition
-} from "./chunk-JO4HIAEF.js";
+} from "./chunk-EFR4JHPL.js";
 import {
   AgentvProvider
-} from "./chunk-Q52FQPKQ.js";
+} from "./chunk-W5YDZWT4.js";
 import {
   OtlpJsonFileExporter
 } from "./chunk-HFSYZHGF.js";
@@ -3752,7 +3751,7 @@ var AzureProvider = class {
     };
     this.retryConfig = config.retry;
     const azure = createAzure(buildAzureOptions(config));
-    this.model = azure(config.deploymentName);
+    this.model = azure.chat(config.deploymentName);
   }
   id;
   kind = "azure";
@@ -9784,9 +9783,11 @@ function negateScore(score) {
     ...score,
     score: negatedScore,
     verdict: negatedVerdict,
-    reasoning: score.reasoning ? `[Negated] ${score.reasoning} (original score: ${score.score.toFixed(2)})` : `[Negated] Original score: ${score.score.toFixed(2)}`,
-    hits: score.misses,
-    misses: score.hits
+    assertions: score.assertions.map((a) => ({
+      ...a,
+      passed: !a.passed,
+      evidence: a.evidence ? `[Negated] ${a.evidence}` : void 0
+    }))
   };
 }
 
@@ -10301,9 +10302,13 @@ var CodeEvaluator = class {
       );
       const parsed = parseJsonSafe(stdout);
       const score = clampScore(typeof parsed?.score === "number" ? parsed.score : 0);
-      const hits = Array.isArray(parsed?.hits) ? parsed.hits.filter(isNonEmptyString) : [];
-      const misses = Array.isArray(parsed?.misses) ? parsed.misses.filter(isNonEmptyString) : [];
-      const reasoning = typeof parsed?.reasoning === "string" ? parsed.reasoning : void 0;
+      const assertions = Array.isArray(parsed?.assertions) ? parsed.assertions.filter(
+        (a) => typeof a === "object" && a !== null && typeof a.text === "string"
+      ).map((a) => ({
+        text: String(a.text),
+        passed: Boolean(a.passed),
+        ...typeof a.evidence === "string" ? { evidence: a.evidence } : {}
+      })) : [];
       const details = parsed?.details && typeof parsed.details === "object" && !Array.isArray(parsed.details) ? parsed.details : void 0;
       const proxyUsage = getProxyUsage?.();
       const evaluatorRawRequest = {
@@ -10319,10 +10324,8 @@ var CodeEvaluator = class {
       return {
         score,
         verdict: scoreToVerdict(score),
-        hits,
-        misses,
-        expectedAspectCount: hits.length + misses.length || 1,
-        reasoning,
+        assertions,
+        expectedAspectCount: assertions.length || 1,
         evaluatorRawRequest,
         ...details ? { details } : {},
         tokenUsage: proxyUsage?.tokenUsage
@@ -10333,10 +10336,8 @@ var CodeEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`Code evaluator failed: ${message}`],
+        assertions: [{ text: `Code evaluator failed: ${message}`, passed: false }],
         expectedAspectCount: 1,
-        reasoning: message,
         evaluatorRawRequest: {
           command: this.command,
           ...this.cwd ? { cwd: this.cwd } : {},
@@ -10444,9 +10445,13 @@ Be concise and focused in your evaluation. Provide succinct, specific feedback r
 {{${TEMPLATE_VARIABLES.ANSWER}}}`;
 var freeformEvaluationSchema = z3.object({
   score: z3.number().min(0).max(1).describe("Score between 0.0 and 1.0"),
-  hits: z3.array(z3.string()).describe("Brief specific achievements").optional(),
-  misses: z3.array(z3.string()).describe("Brief failures or omissions").optional(),
-  reasoning: z3.string().describe("Concise explanation (1-2 sentences)").optional()
+  assertions: z3.array(
+    z3.object({
+      text: z3.string().describe("Brief description of what was checked"),
+      passed: z3.boolean().describe("Whether this aspect was satisfied"),
+      evidence: z3.string().describe("Concise evidence (1-2 sentences)").optional()
+    })
+  ).describe("Per-aspect evaluation results \u2014 one entry per aspect checked").optional()
 });
 var rubricCheckResultSchema = z3.object({
   id: z3.string().describe("The ID of the rubric item being checked"),
@@ -10548,17 +10553,12 @@ ${context.fileChanges}`;
         schema: freeformEvaluationSchema
       });
       const score = clampScore(data.score);
-      const hits = Array.isArray(data.hits) ? data.hits.filter(isNonEmptyString).slice(0, 4) : [];
-      const misses = Array.isArray(data.misses) ? data.misses.filter(isNonEmptyString).slice(0, 4) : [];
-      const reasoning = data.reasoning;
-      const expectedAspectCount = Math.max(hits.length + misses.length, 1);
+      const assertions = Array.isArray(data.assertions) ? data.assertions.slice(0, 8) : [];
       return {
         score,
         verdict: scoreToVerdict(score),
-        hits,
-        misses,
-        expectedAspectCount,
-        reasoning,
+        assertions,
+        expectedAspectCount: Math.max(assertions.length, 1),
         evaluatorRawRequest,
         tokenUsage
       };
@@ -10569,10 +10569,8 @@ ${context.fileChanges}`;
       return {
         score: 0,
         verdict: "skip",
-        hits: [],
-        misses: [`Grader parse failure after 3 attempts: ${message}`],
+        assertions: [{ text: `Grader parse failure after 3 attempts: ${message}`, passed: false }],
         expectedAspectCount: 1,
-        reasoning: `Grader parse failure after 3 attempts: ${message}`,
         evaluatorRawRequest
       };
     }
@@ -10602,14 +10600,12 @@ ${context.fileChanges}`;
         userPrompt: prompt,
         schema: rubricEvaluationSchema
       });
-      const { score, verdict, hits, misses } = calculateRubricScore(data, rubrics);
+      const { score, verdict, assertions } = calculateRubricScore(data, rubrics);
       return {
         score,
         verdict,
-        hits,
-        misses,
+        assertions,
         expectedAspectCount: rubrics.length,
-        reasoning: data.overall_reasoning,
         evaluatorRawRequest,
         tokenUsage
       };
@@ -10620,10 +10616,8 @@ ${context.fileChanges}`;
       return {
         score: 0,
         verdict: "skip",
-        hits: [],
-        misses: [`Grader parse failure after 3 attempts: ${message}`],
+        assertions: [{ text: `Grader parse failure after 3 attempts: ${message}`, passed: false }],
         expectedAspectCount: rubrics.length,
-        reasoning: `Grader parse failure after 3 attempts: ${message}`,
         evaluatorRawRequest
       };
     }
@@ -10648,14 +10642,12 @@ ${context.fileChanges}`;
         userPrompt: prompt,
         schema: scoreRangeEvaluationSchema
       });
-      const { score, verdict, hits, misses, details } = calculateScoreRangeResult(data, rubrics);
+      const { score, verdict, assertions, details } = calculateScoreRangeResult(data, rubrics);
       return {
         score,
         verdict,
-        hits,
-        misses,
+        assertions,
         expectedAspectCount: rubrics.length,
-        reasoning: data.overall_reasoning,
         evaluatorRawRequest,
         details,
         tokenUsage
@@ -10667,10 +10659,8 @@ ${context.fileChanges}`;
       return {
         score: 0,
         verdict: "skip",
-        hits: [],
-        misses: [`Grader parse failure after 3 attempts: ${message}`],
+        assertions: [{ text: `Grader parse failure after 3 attempts: ${message}`, passed: false }],
         expectedAspectCount: rubrics.length,
-        reasoning: `Grader parse failure after 3 attempts: ${message}`,
         evaluatorRawRequest
       };
     }
@@ -10727,8 +10717,7 @@ ${context.fileChanges}`;
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`llm-grader built-in evaluation failed: ${message}`],
+        assertions: [{ text: `llm-grader built-in evaluation failed: ${message}`, passed: false }],
         expectedAspectCount: 1,
         evaluatorRawRequest,
         details: { mode: "built-in", error: message }
@@ -10778,8 +10767,9 @@ ${context.fileChanges}`;
         return {
           score: 0,
           verdict: "fail",
-          hits: [],
-          misses: [`llm-grader ${modeLabel} returned no assistant response`],
+          assertions: [
+            { text: `llm-grader ${modeLabel} returned no assistant response`, passed: false }
+          ],
           expectedAspectCount: 1,
           evaluatorRawRequest,
           details: { mode: modeLabel, grader_target: provider.targetName }
@@ -10797,8 +10787,9 @@ ${context.fileChanges}`;
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`llm-grader ${modeLabel} evaluation failed: ${message}`],
+        assertions: [
+          { text: `llm-grader ${modeLabel} evaluation failed: ${message}`, passed: false }
+        ],
         expectedAspectCount: 1,
         evaluatorRawRequest,
         details: {
@@ -10950,29 +10941,24 @@ ${outputSchema}`;
       const parsed = parseJsonFromText(text);
       if (rubrics && rubrics.length > 0) {
         const data2 = rubricEvaluationSchema.parse(parsed);
-        const { score: score2, verdict, hits: hits2, misses: misses2 } = calculateRubricScore(data2, rubrics);
+        const { score: score2, verdict, assertions: assertions2 } = calculateRubricScore(data2, rubrics);
         return {
           score: score2,
           verdict,
-          hits: hits2,
-          misses: misses2,
+          assertions: assertions2,
           expectedAspectCount: rubrics.length,
-          reasoning: data2.overall_reasoning,
           evaluatorRawRequest,
           details
         };
       }
       const data = freeformEvaluationSchema.parse(parsed);
       const score = clampScore(data.score);
-      const hits = Array.isArray(data.hits) ? data.hits.filter(isNonEmptyString).slice(0, 4) : [];
-      const misses = Array.isArray(data.misses) ? data.misses.filter(isNonEmptyString).slice(0, 4) : [];
+      const assertions = Array.isArray(data.assertions) ? data.assertions.slice(0, 8) : [];
       return {
         score,
         verdict: scoreToVerdict(score),
-        hits,
-        misses,
-        expectedAspectCount: Math.max(hits.length + misses.length, 1),
-        reasoning: data.reasoning,
+        assertions,
+        expectedAspectCount: Math.max(assertions.length, 1),
         evaluatorRawRequest,
         details
       };
@@ -10980,8 +10966,12 @@ ${outputSchema}`;
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["Failed to parse llm-grader agent response as valid evaluation JSON"],
+        assertions: [
+          {
+            text: "Failed to parse llm-grader agent response as valid evaluation JSON",
+            passed: false
+          }
+        ],
         expectedAspectCount: 1,
         evaluatorRawRequest,
         details
@@ -11110,9 +11100,13 @@ function buildOutputSchema() {
     "",
     "{",
     '  "score": <number between 0.0 and 1.0>,',
-    '  "hits": [<array of strings, max 4 items, brief specific achievements>],',
-    '  "misses": [<array of strings, max 4 items, brief specific failures or omissions, empty if none>],',
-    '  "reasoning": "<string, concise explanation for the score, 1-2 sentences max>"',
+    '  "assertions": [',
+    "    {",
+    '      "text": "<brief description of what was checked>",',
+    '      "passed": <boolean>,',
+    '      "evidence": "<concise evidence, 1-2 sentences, optional>"',
+    "    }",
+    "  ]",
     "}"
   ].join("\n");
 }
@@ -11137,8 +11131,7 @@ function substituteVariables(template, variables) {
 }
 function calculateRubricScore(result, rubrics) {
   const rubricMap = new Map(rubrics.map((rubric) => [rubric.id, rubric]));
-  const hits = [];
-  const misses = [];
+  const assertions = [];
   let totalWeight = 0;
   let earnedWeight = 0;
   let failedRequired = false;
@@ -11148,19 +11141,20 @@ function calculateRubricScore(result, rubrics) {
       continue;
     }
     totalWeight += rubric.weight;
+    assertions.push({
+      text: `[${rubric.id}] ${rubric.outcome}`,
+      passed: check.satisfied,
+      evidence: check.reasoning
+    });
     if (check.satisfied) {
       earnedWeight += rubric.weight;
-      hits.push(`[${rubric.id}] ${rubric.outcome}: ${check.reasoning}`);
-    } else {
-      misses.push(`[${rubric.id}] ${rubric.outcome}: ${check.reasoning}`);
-      if (rubric.required) {
-        failedRequired = true;
-      }
+    } else if (rubric.required) {
+      failedRequired = true;
     }
   }
   const score = totalWeight > 0 ? Math.min(1, Math.max(0, earnedWeight / totalWeight)) : 0;
   const verdict = failedRequired ? "fail" : scoreToVerdict(score);
-  return { score, verdict, hits, misses };
+  return { score, verdict, assertions };
 }
 function buildScoreRangeOutputSchema() {
   return `You are an expert evaluator. Score the candidate answer on each criterion.
@@ -11180,8 +11174,7 @@ Important: The "score" must be an integer from 0 to 10 that falls within one of
 }
 function calculateScoreRangeResult(result, rubrics) {
   const rubricMap = new Map(rubrics.map((rubric) => [rubric.id, rubric]));
-  const hits = [];
-  const misses = [];
+  const assertions = [];
   const rawScores = {};
   let totalWeight = 0;
   let weightedScoreSum = 0;
@@ -11207,24 +11200,22 @@ function calculateScoreRangeResult(result, rubrics) {
     );
     const rangeDescription = matchingRange?.outcome ?? "";
     const criterionLabel = rubric.outcome ?? rubric.id;
-    const reasoningText = check.reasoning ? `: ${check.reasoning}` : "";
-    const scoreInfo = `[${rubric.id}] ${criterionLabel} - Score: ${rawScore}/10 (${rangeDescription})${reasoningText}`;
+    const passed = !(requiredMinScore !== void 0 && rawScore < requiredMinScore) && rawScore >= 7;
     if (requiredMinScore !== void 0 && rawScore < requiredMinScore) {
       failedRequired = true;
-      misses.push(scoreInfo);
-    } else if (rawScore >= 7) {
-      hits.push(scoreInfo);
-    } else {
-      misses.push(scoreInfo);
     }
+    assertions.push({
+      text: `[${rubric.id}] ${criterionLabel} - Score: ${rawScore}/10 (${rangeDescription})`,
+      passed,
+      evidence: check.reasoning
+    });
   }
   const score = totalWeight > 0 ? Math.min(1, Math.max(0, weightedScoreSum / totalWeight)) : 0;
   const verdict = failedRequired ? "fail" : scoreToVerdict(score);
   return {
     score,
     verdict,
-    hits,
-    misses,
+    assertions,
     details: {
       raw_scores: rawScores,
       normalization: "score / 10",
@@ -11400,9 +11391,7 @@ var CompositeEvaluator = class {
     let totalWeight = 0;
     let weightedSum = 0;
     let evaluatedCount = 0;
-    const allHits = [];
-    const allMisses = [];
-    const reasoningParts = [];
+    const allAssertions = [];
     const scores = [];
     for (const member of results) {
       const weight = weights?.[member.id] ?? 1;
@@ -11412,9 +11401,7 @@ var CompositeEvaluator = class {
         score: member.result.score,
         weight,
         verdict: member.result.verdict,
-        hits: [...member.result.hits],
-        misses: [...member.result.misses],
-        reasoning: member.result.reasoning,
+        assertions: [...member.result.assertions],
         evaluatorRawRequest: member.result.evaluatorRawRequest,
         scores: member.result.scores,
         details: member.result.details,
@@ -11426,20 +11413,16 @@ var CompositeEvaluator = class {
       evaluatedCount++;
       totalWeight += weight;
       weightedSum += member.result.score * weight;
-      allHits.push(...member.result.hits.map((h) => `[${member.id}] ${h}`));
-      allMisses.push(...member.result.misses.map((m) => `[${member.id}] ${m}`));
-      if (member.result.reasoning) {
-        reasoningParts.push(`${member.id}: ${member.result.reasoning}`);
-      }
+      allAssertions.push(
+        ...member.result.assertions.map((a) => ({ ...a, text: `[${member.id}] ${a.text}` }))
+      );
     }
     if (evaluatedCount === 0 && results.length > 0) {
       return {
         score: 0,
         verdict: "skip",
-        hits: [],
-        misses: [],
+        assertions: [{ text: "All evaluators skipped (infrastructure failure)", passed: false }],
         expectedAspectCount: 1,
-        reasoning: "All evaluators skipped (infrastructure failure)",
         evaluatorRawRequest: {
           aggregator: "weighted_average",
           ...weights ? { weights } : {}
@@ -11451,10 +11434,8 @@ var CompositeEvaluator = class {
     return {
       score: clampScore(finalScore),
       verdict: scoreToVerdict(finalScore),
-      hits: allHits,
-      misses: allMisses,
-      expectedAspectCount: Math.max(allHits.length + allMisses.length, 1),
-      reasoning: reasoningParts.length > 0 ? reasoningParts.join("; ") : void 0,
+      assertions: allAssertions,
+      expectedAspectCount: allAssertions.length || 1,
       evaluatorRawRequest: {
         aggregator: "weighted_average",
         ...weights ? { weights } : {}
@@ -11464,11 +11445,8 @@ var CompositeEvaluator = class {
   }
   runThreshold(results, threshold) {
     const scores = [];
-    const allHits = [];
-    const allMisses = [];
-    const reasoningParts = [];
+    const allAssertions = [];
     let passingCount = 0;
-    let borderlineCount = 0;
     let evaluatedCount = 0;
     for (const member of results) {
       scores.push({
@@ -11476,9 +11454,7 @@ var CompositeEvaluator = class {
         type: member.type,
         score: member.result.score,
         verdict: member.result.verdict,
-        hits: [...member.result.hits],
-        misses: [...member.result.misses],
-        reasoning: member.result.reasoning,
+        assertions: [...member.result.assertions],
         evaluatorRawRequest: member.result.evaluatorRawRequest,
         scores: member.result.scores,
         details: member.result.details,
@@ -11491,24 +11467,17 @@ var CompositeEvaluator = class {
       const isPassing = member.result.verdict === "pass" || member.result.verdict === "borderline";
       if (isPassing) {
         passingCount++;
-        if (member.result.verdict === "borderline") {
-          borderlineCount++;
-        }
-      }
-      allHits.push(...member.result.hits.map((h) => `[${member.id}] ${h}`));
-      allMisses.push(...member.result.misses.map((m) => `[${member.id}] ${m}`));
-      if (member.result.reasoning) {
-        reasoningParts.push(`${member.id}: ${member.result.reasoning}`);
       }
+      allAssertions.push(
+        ...member.result.assertions.map((a) => ({ ...a, text: `[${member.id}] ${a.text}` }))
+      );
     }
     if (evaluatedCount === 0 && results.length > 0) {
       return {
         score: 0,
         verdict: "skip",
-        hits: [],
-        misses: [],
+        assertions: [{ text: "All evaluators skipped (infrastructure failure)", passed: false }],
         expectedAspectCount: 1,
-        reasoning: "All evaluators skipped (infrastructure failure)",
         evaluatorRawRequest: {
           aggregator: "threshold",
           threshold
@@ -11519,19 +11488,15 @@ var CompositeEvaluator = class {
     const totalCount = evaluatedCount;
     const score = totalCount > 0 ? passingCount / totalCount : 0;
     const pass = score >= threshold;
-    if (pass && borderlineCount > 0) {
-      reasoningParts.push(`Warning: ${borderlineCount} borderline evaluator(s) counted as passing`);
-    }
-    reasoningParts.unshift(
-      `${passingCount}/${totalCount} evaluators passed (threshold: ${threshold})`
-    );
+    allAssertions.unshift({
+      text: `${passingCount}/${totalCount} evaluators passed (threshold: ${threshold})`,
+      passed: pass
+    });
     return {
       score: clampScore(score),
       verdict: pass ? "pass" : "fail",
-      hits: allHits,
-      misses: allMisses,
-      expectedAspectCount: Math.max(allHits.length + allMisses.length, 1),
-      reasoning: reasoningParts.join("; "),
+      assertions: allAssertions,
+      expectedAspectCount: allAssertions.length || 1,
       evaluatorRawRequest: {
         aggregator: "threshold",
         threshold
@@ -11548,9 +11513,7 @@ var CompositeEvaluator = class {
       score: member.result.score,
       weight: weights?.[member.id] ?? 1,
       verdict: member.result.verdict,
-      hits: [...member.result.hits],
-      misses: [...member.result.misses],
-      reasoning: member.result.reasoning,
+      assertions: [...member.result.assertions],
       evaluatorRawRequest: member.result.evaluatorRawRequest,
       scores: member.result.scores,
       details: member.result.details
@@ -11559,17 +11522,19 @@ var CompositeEvaluator = class {
       const stdout = await executeScript(scriptPath, inputPayload, void 0, cwd);
       const parsed = parseJsonSafe(stdout);
       const score = clampScore(typeof parsed?.score === "number" ? parsed.score : 0);
-      const hits = Array.isArray(parsed?.hits) ? parsed.hits.filter(isNonEmptyString) : [];
-      const misses = Array.isArray(parsed?.misses) ? parsed.misses.filter(isNonEmptyString) : [];
-      const reasoning = typeof parsed?.reasoning === "string" ? parsed.reasoning : void 0;
+      const assertions = Array.isArray(parsed?.assertions) ? parsed.assertions.filter(
+        (a) => typeof a === "object" && a !== null && typeof a.text === "string"
+      ).map((a) => ({
+        text: String(a.text),
+        passed: Boolean(a.passed),
+        ...typeof a.evidence === "string" ? { evidence: a.evidence } : {}
+      })) : [];
       const verdict = typeof parsed?.verdict === "string" && (parsed.verdict === "pass" || parsed.verdict === "fail" || parsed.verdict === "borderline") ? parsed.verdict : scoreToVerdict(score);
       return {
         score,
         verdict,
-        hits,
-        misses,
-        expectedAspectCount: hits.length + misses.length || 1,
-        reasoning,
+        assertions,
+        expectedAspectCount: assertions.length || 1,
         evaluatorRawRequest: {
           aggregator: "code-grader",
           script: scriptPath
@@ -11581,10 +11546,8 @@ var CompositeEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`Code aggregator failed: ${message}`],
+        assertions: [{ text: `Code aggregator failed: ${message}`, passed: false }],
         expectedAspectCount: 1,
-        reasoning: message,
         evaluatorRawRequest: {
           aggregator: "code-grader",
           script: scriptPath,
@@ -11606,9 +11569,7 @@ var CompositeEvaluator = class {
       type: member.type,
       score: member.result.score,
       verdict: member.result.verdict,
-      hits: [...member.result.hits],
-      misses: [...member.result.misses],
-      reasoning: member.result.reasoning,
+      assertions: [...member.result.assertions],
       evaluatorRawRequest: member.result.evaluatorRawRequest,
       scores: member.result.scores,
       details: member.result.details
@@ -11632,16 +11593,12 @@ var CompositeEvaluator = class {
         });
         const data2 = freeformEvaluationSchema.parse(parseJsonFromText(text));
         const score2 = clampScore(data2.score);
-        const hits2 = Array.isArray(data2.hits) ? data2.hits.filter(isNonEmptyString).slice(0, 4) : [];
-        const misses2 = Array.isArray(data2.misses) ? data2.misses.filter(isNonEmptyString).slice(0, 4) : [];
-        const reasoning2 = data2.reasoning;
+        const assertions2 = Array.isArray(data2.assertions) ? data2.assertions.slice(0, 8) : [];
         return {
           score: score2,
           verdict: scoreToVerdict(score2),
-          hits: hits2,
-          misses: misses2,
-          expectedAspectCount: Math.max(hits2.length + misses2.length, 1),
-          reasoning: reasoning2,
+          assertions: assertions2,
+          expectedAspectCount: Math.max(assertions2.length, 1),
           evaluatorRawRequest,
           scores
         };
@@ -11656,16 +11613,12 @@ var CompositeEvaluator = class {
         parseJsonFromText(extractLastAssistantContent(response.output))
       );
       const score = clampScore(data.score);
-      const hits = Array.isArray(data.hits) ? data.hits.filter(isNonEmptyString).slice(0, 4) : [];
-      const misses = Array.isArray(data.misses) ? data.misses.filter(isNonEmptyString).slice(0, 4) : [];
-      const reasoning = data.reasoning;
+      const assertions = Array.isArray(data.assertions) ? data.assertions.slice(0, 8) : [];
       return {
         score,
         verdict: scoreToVerdict(score),
-        hits,
-        misses,
-        expectedAspectCount: Math.max(hits.length + misses.length, 1),
-        reasoning,
+        assertions,
+        expectedAspectCount: Math.max(assertions.length, 1),
         evaluatorRawRequest,
         scores
       };
@@ -11673,8 +11626,7 @@ var CompositeEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [],
+        assertions: [{ text: "LLM aggregator failed", passed: false }],
         expectedAspectCount: 1,
         evaluatorRawRequest,
         scores
@@ -11697,10 +11649,8 @@ var CostEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No cost data available in trace"],
+        assertions: [{ text: "No cost data available in trace", passed: false }],
         expectedAspectCount: 1,
-        reasoning: "Execution cost not reported by provider",
         evaluatorRawRequest: {
           type: "cost",
           budget,
@@ -11714,10 +11664,10 @@ var CostEvaluator = class {
     return {
       score,
       verdict: passed ? "pass" : "fail",
-      hits: passed ? [`Cost ${formatCost(costUsd)} <= ${formatCost(budget)} budget`] : [],
-      misses: passed ? [] : [`Cost ${formatCost(costUsd)} > ${formatCost(budget)} budget`],
+      assertions: [
+        passed ? { text: `Cost ${formatCost(costUsd)} <= ${formatCost(budget)} budget`, passed: true } : { text: `Cost ${formatCost(costUsd)} > ${formatCost(budget)} budget`, passed: false }
+      ],
       expectedAspectCount: 1,
-      reasoning: `Execution cost ${formatCost(costUsd)} (budget: ${formatCost(budget)})`,
       evaluatorRawRequest: {
         type: "cost",
         budget,
@@ -11750,10 +11700,8 @@ var ExecutionMetricsEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No trace summary available"],
+        assertions: [{ text: "No trace summary available", passed: false }],
         expectedAspectCount: 1,
-        reasoning: "Execution metrics not available - no trace summary provided",
         evaluatorRawRequest: {
           type: "execution-metrics",
           config: this.extractConfiguredThresholds(),
@@ -11762,116 +11710,114 @@ var ExecutionMetricsEvaluator = class {
       };
     }
     const narrowedTrace = trace;
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     const actualMetrics = {};
     if (max_tool_calls !== void 0 && narrowedTrace) {
       const toolCalls = narrowedTrace.eventCount;
       actualMetrics.tool_calls = toolCalls;
       if (toolCalls <= max_tool_calls) {
-        hits.push(`Tool calls ${toolCalls} <= ${max_tool_calls} max`);
+        assertions.push({ text: `Tool calls ${toolCalls} <= ${max_tool_calls} max`, passed: true });
       } else {
-        misses.push(`Tool calls ${toolCalls} > ${max_tool_calls} max`);
+        assertions.push({ text: `Tool calls ${toolCalls} > ${max_tool_calls} max`, passed: false });
       }
     }
     if (max_llm_calls !== void 0 && narrowedTrace) {
       const llmCalls = narrowedTrace.llmCallCount;
       if (llmCalls === void 0) {
-        misses.push("LLM call count data not available");
+        assertions.push({ text: "LLM call count data not available", passed: false });
       } else {
         actualMetrics.llm_calls = llmCalls;
         if (llmCalls <= max_llm_calls) {
-          hits.push(`LLM calls ${llmCalls} <= ${max_llm_calls} max`);
+          assertions.push({ text: `LLM calls ${llmCalls} <= ${max_llm_calls} max`, passed: true });
         } else {
-          misses.push(`LLM calls ${llmCalls} > ${max_llm_calls} max`);
+          assertions.push({ text: `LLM calls ${llmCalls} > ${max_llm_calls} max`, passed: false });
         }
       }
     }
     if (max_tokens !== void 0) {
       if (!tokenUsage) {
-        misses.push("Token usage data not available");
+        assertions.push({ text: "Token usage data not available", passed: false });
       } else {
         const totalTokens = tokenUsage.input + tokenUsage.output;
         actualMetrics.tokens = totalTokens;
         if (totalTokens <= max_tokens) {
-          hits.push(`Total tokens ${totalTokens} <= ${max_tokens} max`);
+          assertions.push({
+            text: `Total tokens ${totalTokens} <= ${max_tokens} max`,
+            passed: true
+          });
         } else {
-          misses.push(`Total tokens ${totalTokens} > ${max_tokens} max`);
+          assertions.push({
+            text: `Total tokens ${totalTokens} > ${max_tokens} max`,
+            passed: false
+          });
         }
       }
     }
     if (max_cost_usd !== void 0) {
       if (costUsd === void 0) {
-        misses.push("Cost data not available");
+        assertions.push({ text: "Cost data not available", passed: false });
       } else {
         actualMetrics.cost_usd = costUsd;
         const formatCost = (n) => `$${n.toFixed(4)}`;
         if (costUsd <= max_cost_usd) {
-          hits.push(`Cost ${formatCost(costUsd)} <= ${formatCost(max_cost_usd)} max`);
+          assertions.push({
+            text: `Cost ${formatCost(costUsd)} <= ${formatCost(max_cost_usd)} max`,
+            passed: true
+          });
         } else {
-          misses.push(`Cost ${formatCost(costUsd)} > ${formatCost(max_cost_usd)} max`);
+          assertions.push({
+            text: `Cost ${formatCost(costUsd)} > ${formatCost(max_cost_usd)} max`,
+            passed: false
+          });
         }
       }
     }
     if (max_duration_ms !== void 0) {
       if (durationMs === void 0) {
-        misses.push("Duration data not available");
+        assertions.push({ text: "Duration data not available", passed: false });
       } else {
         actualMetrics.duration_ms = durationMs;
         if (durationMs <= max_duration_ms) {
-          hits.push(`Duration ${durationMs}ms <= ${max_duration_ms}ms max`);
+          assertions.push({
+            text: `Duration ${durationMs}ms <= ${max_duration_ms}ms max`,
+            passed: true
+          });
         } else {
-          misses.push(`Duration ${durationMs}ms > ${max_duration_ms}ms max`);
+          assertions.push({
+            text: `Duration ${durationMs}ms > ${max_duration_ms}ms max`,
+            passed: false
+          });
         }
       }
     }
     if (target_exploration_ratio !== void 0 && narrowedTrace) {
       const ratio = explorationRatio(narrowedTrace);
       if (ratio === void 0) {
-        misses.push("Exploration ratio not available (no tool calls)");
+        assertions.push({ text: "Exploration ratio not available (no tool calls)", passed: false });
       } else {
         actualMetrics.exploration_ratio = ratio;
         const diff = Math.abs(ratio - target_exploration_ratio);
         if (diff <= exploration_tolerance) {
-          hits.push(
-            `Exploration ratio ${ratio.toFixed(2)} within tolerance of target ${target_exploration_ratio}`
-          );
+          assertions.push({
+            text: `Exploration ratio ${ratio.toFixed(2)} within tolerance of target ${target_exploration_ratio}`,
+            passed: true
+          });
         } else {
-          misses.push(
-            `Exploration ratio ${ratio.toFixed(2)} outside tolerance of target ${target_exploration_ratio} (diff: ${diff.toFixed(2)}, tolerance: ${exploration_tolerance})`
-          );
+          assertions.push({
+            text: `Exploration ratio ${ratio.toFixed(2)} outside tolerance of target ${target_exploration_ratio} (diff: ${diff.toFixed(2)}, tolerance: ${exploration_tolerance})`,
+            passed: false
+          });
         }
       }
     }
-    const totalChecks = hits.length + misses.length;
-    const score = totalChecks > 0 ? hits.length / totalChecks : 0;
-    const reasoningParts = [];
-    if (actualMetrics.tool_calls !== void 0) {
-      reasoningParts.push(`tool_calls=${actualMetrics.tool_calls}`);
-    }
-    if (actualMetrics.llm_calls !== void 0) {
-      reasoningParts.push(`llm_calls=${actualMetrics.llm_calls}`);
-    }
-    if (actualMetrics.tokens !== void 0) {
-      reasoningParts.push(`tokens=${actualMetrics.tokens}`);
-    }
-    if (actualMetrics.cost_usd !== void 0) {
-      reasoningParts.push(`cost=$${actualMetrics.cost_usd.toFixed(4)}`);
-    }
-    if (actualMetrics.duration_ms !== void 0) {
-      reasoningParts.push(`duration=${actualMetrics.duration_ms}ms`);
-    }
-    if (actualMetrics.exploration_ratio !== void 0) {
-      reasoningParts.push(`exploration_ratio=${actualMetrics.exploration_ratio.toFixed(2)}`);
-    }
-    const reasoning = reasoningParts.length > 0 ? `execution-metrics ${reasoningParts.join(", ")}` : "No metrics evaluated";
+    const totalChecks = assertions.length;
+    const passedCount = assertions.filter((a) => a.passed).length;
+    const score = totalChecks > 0 ? passedCount / totalChecks : 0;
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: totalChecks || 1,
-      reasoning,
       evaluatorRawRequest: {
         type: "execution-metrics",
         config: this.extractConfiguredThresholds(),
@@ -11975,10 +11921,8 @@ var FieldAccuracyEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["Failed to parse candidate answer as JSON"],
-        expectedAspectCount: this.config.fields.length,
-        reasoning: "Candidate answer is not valid JSON"
+        assertions: [{ text: "Failed to parse candidate answer as JSON", passed: false }],
+        expectedAspectCount: this.config.fields.length
       };
     }
     const expectedData = this.extractExpectedData(evalCase.expected_output);
@@ -11986,10 +11930,8 @@ var FieldAccuracyEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No expected data found in expected_output"],
-        expectedAspectCount: this.config.fields.length,
-        reasoning: "Could not extract expected data from expected_output"
+        assertions: [{ text: "No expected data found in expected_output", passed: false }],
+        expectedAspectCount: this.config.fields.length
       };
     }
     const fieldResults = [];
@@ -12207,18 +12149,14 @@ var FieldAccuracyEvaluator = class {
    */
   aggregateResults(results) {
     const aggregation = this.config.aggregation ?? "weighted_average";
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     for (const result of results) {
-      if (result.hit) {
-        hits.push(result.message);
-      } else {
-        misses.push(result.message);
-      }
+      assertions.push({ text: result.message, passed: result.hit });
     }
     let score;
     if (aggregation === "all_or_nothing") {
-      score = misses.length === 0 ? 1 : 0;
+      const hasFailed = assertions.some((a) => !a.passed);
+      score = hasFailed ? 0 : 1;
     } else {
       const totalWeight = results.reduce((sum, r) => sum + r.weight, 0);
       if (totalWeight === 0) {
@@ -12228,15 +12166,11 @@ var FieldAccuracyEvaluator = class {
         score = weightedSum / totalWeight;
       }
     }
-    const reasoning = `${hits.length}/${results.length} fields matched`;
     return {
       score: clampScore(score),
       verdict: scoreToVerdict(score),
-      hits: hits.slice(0, 4),
-      // Cap at 4 to keep output concise
-      misses: misses.slice(0, 4),
-      expectedAspectCount: results.length,
-      reasoning
+      assertions,
+      expectedAspectCount: results.length
     };
   }
 };
@@ -12345,10 +12279,8 @@ var LatencyEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No duration data available in trace"],
+        assertions: [{ text: "No duration data available in trace", passed: false }],
         expectedAspectCount: 1,
-        reasoning: "Execution duration not reported by provider",
         evaluatorRawRequest: {
           type: "latency",
           threshold,
@@ -12361,10 +12293,10 @@ var LatencyEvaluator = class {
     return {
       score,
       verdict: passed ? "pass" : "fail",
-      hits: passed ? [`Duration ${durationMs}ms <= ${threshold}ms threshold`] : [],
-      misses: passed ? [] : [`Duration ${durationMs}ms > ${threshold}ms threshold`],
+      assertions: [
+        passed ? { text: `Duration ${durationMs}ms <= ${threshold}ms threshold`, passed: true } : { text: `Duration ${durationMs}ms > ${threshold}ms threshold`, passed: false }
+      ],
       expectedAspectCount: 1,
-      reasoning: `Execution took ${durationMs}ms (threshold: ${threshold}ms)`,
       evaluatorRawRequest: {
         type: "latency",
         threshold,
@@ -12440,23 +12372,25 @@ var SkillTriggerEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: [
-          shouldTrigger ? evidence || `Skill "${skillName}" triggered as expected` : `Skill "${skillName}" correctly did not trigger`
+        assertions: [
+          {
+            text: shouldTrigger ? evidence || `Skill "${skillName}" triggered as expected` : `Skill "${skillName}" correctly did not trigger`,
+            passed: true
+          }
         ],
-        misses: [],
-        expectedAspectCount: 1,
-        reasoning: shouldTrigger ? "Skill triggered correctly" : "No false trigger"
+        expectedAspectCount: 1
       };
     }
     return {
       score: 0,
       verdict: "fail",
-      hits: [],
-      misses: [
-        shouldTrigger ? firstTool ? `First tool was "${firstTool.tool}" \u2014 not a skill/read tool for "${skillName}"` : "No tool calls recorded" : evidence || `Skill "${skillName}" triggered unexpectedly`
+      assertions: [
+        {
+          text: shouldTrigger ? firstTool ? `First tool was "${firstTool.tool}" \u2014 not a skill/read tool for "${skillName}"` : "No tool calls recorded" : evidence || `Skill "${skillName}" triggered unexpectedly`,
+          passed: false
+        }
       ],
-      expectedAspectCount: 1,
-      reasoning: shouldTrigger ? `Skill "${skillName}" was not triggered` : "False trigger: skill fired when it should not have"
+      expectedAspectCount: 1
     };
   }
 };
@@ -12625,10 +12559,8 @@ var TokenUsageEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No token usage data available in trace"],
+        assertions: [{ text: "No token usage data available in trace", passed: false }],
         expectedAspectCount,
-        reasoning: "Token usage not reported by provider",
         evaluatorRawRequest: {
           type: "token-usage",
           max_total: maxTotal ?? null,
@@ -12642,37 +12574,34 @@ var TokenUsageEvaluator = class {
     const output = usage.output;
     const cached = usage.cached ?? 0;
     const total = input + output + cached;
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     if (typeof maxInput === "number") {
       if (input <= maxInput) {
-        hits.push(`Input tokens ${input} <= ${maxInput}`);
+        assertions.push({ text: `Input tokens ${input} <= ${maxInput}`, passed: true });
       } else {
-        misses.push(`Input tokens ${input} > ${maxInput}`);
+        assertions.push({ text: `Input tokens ${input} > ${maxInput}`, passed: false });
       }
     }
     if (typeof maxOutput === "number") {
       if (output <= maxOutput) {
-        hits.push(`Output tokens ${output} <= ${maxOutput}`);
+        assertions.push({ text: `Output tokens ${output} <= ${maxOutput}`, passed: true });
       } else {
-        misses.push(`Output tokens ${output} > ${maxOutput}`);
+        assertions.push({ text: `Output tokens ${output} > ${maxOutput}`, passed: false });
       }
     }
     if (typeof maxTotal === "number") {
       if (total <= maxTotal) {
-        hits.push(`Total tokens ${total} <= ${maxTotal}`);
+        assertions.push({ text: `Total tokens ${total} <= ${maxTotal}`, passed: true });
       } else {
-        misses.push(`Total tokens ${total} > ${maxTotal}`);
+        assertions.push({ text: `Total tokens ${total} > ${maxTotal}`, passed: false });
       }
     }
-    const passed = misses.length === 0;
+    const passed = assertions.every((a) => a.passed);
     return {
       score: passed ? 1 : 0,
       verdict: passed ? "pass" : "fail",
-      hits,
-      misses,
+      assertions,
       expectedAspectCount,
-      reasoning: `token-usage input=${input}, output=${output}, cached=${cached}, total=${total}`,
       evaluatorRawRequest: {
         type: "token-usage",
         max_total: maxTotal ?? null,
@@ -12772,8 +12701,7 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: ["No trace available for evaluation"],
+        assertions: [{ text: "No trace available for evaluation", passed: false }],
         expectedAspectCount: 1
       };
     }
@@ -12784,8 +12712,7 @@ var ToolTrajectoryEvaluator = class {
           return {
             score: 0,
             verdict: "fail",
-            hits: [],
-            misses: ["No trace available for evaluation"],
+            assertions: [{ text: "No trace available for evaluation", passed: false }],
             expectedAspectCount: 1
           };
         }
@@ -12803,8 +12730,7 @@ var ToolTrajectoryEvaluator = class {
         return {
           score: 0,
           verdict: "fail",
-          hits: [],
-          misses: [`Unknown mode: ${this.config.mode}`],
+          assertions: [{ text: `Unknown mode: ${this.config.mode}`, passed: false }],
           expectedAspectCount: 1
         };
     }
@@ -12853,28 +12779,32 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: ["No tool requirements specified"],
-        misses: [],
+        assertions: [{ text: "No tool requirements specified", passed: true }],
         expectedAspectCount: 0
       };
     }
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     for (const toolName of toolNames) {
       const required = minimums[toolName];
       const actual = summary.toolCallsByName[toolName] ?? 0;
       if (actual >= required) {
-        hits.push(`${toolName}: called ${actual} times (required >=${required})`);
+        assertions.push({
+          text: `${toolName}: called ${actual} times (required >=${required})`,
+          passed: true
+        });
       } else {
-        misses.push(`${toolName}: called ${actual} times (required >=${required})`);
+        assertions.push({
+          text: `${toolName}: called ${actual} times (required >=${required})`,
+          passed: false
+        });
       }
     }
-    const score = hits.length / toolNames.length;
+    const passedCount = assertions.filter((a) => a.passed).length;
+    const score = passedCount / toolNames.length;
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: toolNames.length
     };
   }
@@ -12884,13 +12814,11 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: ["No tool sequence specified"],
-        misses: [],
+        assertions: [{ text: "No tool sequence specified", passed: true }],
         expectedAspectCount: 0
       };
     }
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     const warnings = [];
     let actualIndex = 0;
     let sequenceHits = 0;
@@ -12910,16 +12838,20 @@ var ToolTrajectoryEvaluator = class {
         const actualCall = toolCalls[actualIndex];
         if (actualCall.name === expectedTool) {
           if (argsMatch(expectedItem.args, actualCall.args, mode)) {
-            hits.push(`Found ${expectedTool} at position ${actualIndex}`);
+            assertions.push({
+              text: `Found ${expectedTool} at position ${actualIndex}`,
+              passed: true
+            });
             sequenceHits++;
             matchedCall = actualCall;
             actualIndex++;
             found = true;
             break;
           }
-          misses.push(
-            `Expected ${expectedTool} at position ${i}: tool found at ${actualIndex} but args mismatch`
-          );
+          assertions.push({
+            text: `Expected ${expectedTool} at position ${i}: tool found at ${actualIndex} but args mismatch`,
+            passed: false
+          });
           actualIndex++;
           argsMismatch = true;
           break;
@@ -12927,7 +12859,10 @@ var ToolTrajectoryEvaluator = class {
         actualIndex++;
       }
       if (!found && !argsMismatch) {
-        misses.push(`Expected ${expectedTool} at position ${i}, not found in remaining trace`);
+        assertions.push({
+          text: `Expected ${expectedTool} at position ${i}, not found in remaining trace`,
+          passed: false
+        });
       }
       if (found && matchedCall) {
         const latencyResult = checkLatency(
@@ -12936,10 +12871,10 @@ var ToolTrajectoryEvaluator = class {
           matchedCall.durationMs
         );
         if (latencyResult.status === "pass") {
-          hits.push(latencyResult.message);
+          assertions.push({ text: latencyResult.message, passed: true });
           latencyHits++;
         } else if (latencyResult.status === "fail") {
-          misses.push(latencyResult.message);
+          assertions.push({ text: latencyResult.message, passed: false });
         } else if (latencyResult.message) {
           warnings.push(latencyResult.message);
           latencySkips++;
@@ -12955,8 +12890,7 @@ var ToolTrajectoryEvaluator = class {
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: totalAssertions
     };
   }
@@ -12966,13 +12900,11 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: ["No tool sequence specified"],
-        misses: [],
+        assertions: [{ text: "No tool sequence specified", passed: true }],
         expectedAspectCount: 0
       };
     }
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     const warnings = [];
     let sequenceHits = 0;
     let latencyHits = 0;
@@ -12981,7 +12913,10 @@ var ToolTrajectoryEvaluator = class {
       (item) => item.maxDurationMs !== void 0
     ).length;
     if (toolCalls.length !== expected.length) {
-      misses.push(`Expected ${expected.length} tool calls, got ${toolCalls.length}`);
+      assertions.push({
+        text: `Expected ${expected.length} tool calls, got ${toolCalls.length}`,
+        passed: false
+      });
     }
     const checkLength = Math.min(expected.length, toolCalls.length);
     for (let i = 0; i < checkLength; i++) {
@@ -12993,14 +12928,17 @@ var ToolTrajectoryEvaluator = class {
       let sequenceMatched = false;
       if (actualTool === expectedTool) {
         if (argsMatch(expectedItem.args, actualCall.args, mode)) {
-          hits.push(`Position ${i}: ${expectedTool}`);
+          assertions.push({ text: `Position ${i}: ${expectedTool}`, passed: true });
           sequenceHits++;
           sequenceMatched = true;
         } else {
-          misses.push(`Position ${i}: ${expectedTool} args mismatch`);
+          assertions.push({ text: `Position ${i}: ${expectedTool} args mismatch`, passed: false });
         }
       } else {
-        misses.push(`Position ${i}: expected ${expectedTool}, got ${actualTool}`);
+        assertions.push({
+          text: `Position ${i}: expected ${expectedTool}, got ${actualTool}`,
+          passed: false
+        });
       }
       if (sequenceMatched) {
         const latencyResult = checkLatency(
@@ -13009,10 +12947,10 @@ var ToolTrajectoryEvaluator = class {
           actualCall.durationMs
         );
         if (latencyResult.status === "pass") {
-          hits.push(latencyResult.message);
+          assertions.push({ text: latencyResult.message, passed: true });
           latencyHits++;
         } else if (latencyResult.status === "fail") {
-          misses.push(latencyResult.message);
+          assertions.push({ text: latencyResult.message, passed: false });
         } else if (latencyResult.message) {
           warnings.push(latencyResult.message);
           latencySkips++;
@@ -13020,7 +12958,10 @@ var ToolTrajectoryEvaluator = class {
       }
     }
     for (let i = checkLength; i < expected.length; i++) {
-      misses.push(`Position ${i}: expected ${expected[i].tool}, got nothing`);
+      assertions.push({
+        text: `Position ${i}: expected ${expected[i].tool}, got nothing`,
+        passed: false
+      });
     }
     for (const warning of warnings) {
       console.warn(`[tool-trajectory] ${warning}`);
@@ -13031,8 +12972,7 @@ var ToolTrajectoryEvaluator = class {
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: totalAssertions
     };
   }
@@ -13047,13 +12987,11 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: ["No expected tools specified"],
-        misses: [],
+        assertions: [{ text: "No expected tools specified", passed: true }],
         expectedAspectCount: 0
       };
     }
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     const consumed = /* @__PURE__ */ new Set();
     for (let i = 0; i < expected.length; i++) {
       const expectedItem = expected[i];
@@ -13064,22 +13002,25 @@ var ToolTrajectoryEvaluator = class {
         if (consumed.has(j)) continue;
         const actualCall = toolCalls[j];
         if (actualCall.name === expectedTool && argsMatch(expectedItem.args, actualCall.args, mode)) {
-          hits.push(`Found ${expectedTool} at position ${j}`);
+          assertions.push({ text: `Found ${expectedTool} at position ${j}`, passed: true });
           consumed.add(j);
           found = true;
           break;
         }
       }
       if (!found) {
-        misses.push(`Expected ${expectedTool} not found in actual trajectory`);
+        assertions.push({
+          text: `Expected ${expectedTool} not found in actual trajectory`,
+          passed: false
+        });
       }
     }
-    const score = expected.length > 0 ? hits.length / expected.length : 1;
+    const passedCount = assertions.filter((a) => a.passed).length;
+    const score = expected.length > 0 ? passedCount / expected.length : 1;
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: expected.length
     };
   }
@@ -13095,16 +13036,19 @@ var ToolTrajectoryEvaluator = class {
         return {
           score: 1,
           verdict: "pass",
-          hits: ["No tool calls and no expected tools"],
-          misses: [],
+          assertions: [{ text: "No tool calls and no expected tools", passed: true }],
           expectedAspectCount: 0
         };
       }
       return {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`${toolCalls.length} unexpected tool call(s) with empty allowed list`],
+        assertions: [
+          {
+            text: `${toolCalls.length} unexpected tool call(s) with empty allowed list`,
+            passed: false
+          }
+        ],
         expectedAspectCount: toolCalls.length
       };
     }
@@ -13112,13 +13056,11 @@ var ToolTrajectoryEvaluator = class {
       return {
         score: 1,
         verdict: "pass",
-        hits: ["No actual tool calls (trivially a subset)"],
-        misses: [],
+        assertions: [{ text: "No actual tool calls (trivially a subset)", passed: true }],
         expectedAspectCount: 0
       };
     }
-    const hits = [];
-    const misses = [];
+    const assertions = [];
     for (let i = 0; i < toolCalls.length; i++) {
       const actualCall = toolCalls[i];
       let allowed = false;
@@ -13130,17 +13072,23 @@ var ToolTrajectoryEvaluator = class {
         }
       }
       if (allowed) {
-        hits.push(`Position ${i}: ${actualCall.name} is in allowed set`);
+        assertions.push({
+          text: `Position ${i}: ${actualCall.name} is in allowed set`,
+          passed: true
+        });
       } else {
-        misses.push(`Position ${i}: ${actualCall.name} is not in allowed set`);
+        assertions.push({
+          text: `Position ${i}: ${actualCall.name} is not in allowed set`,
+          passed: false
+        });
       }
     }
-    const score = toolCalls.length > 0 ? hits.length / toolCalls.length : 1;
+    const passedCount = assertions.filter((a) => a.passed).length;
+    const score = toolCalls.length > 0 ? passedCount / toolCalls.length : 1;
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits,
-      misses,
+      assertions,
       expectedAspectCount: toolCalls.length
     };
   }
@@ -13151,8 +13099,12 @@ function runContainsAssertion(output, value) {
   const passed = output.includes(value);
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains "${value}"`] : [],
-    misses: passed ? [] : [`Output does not contain "${value}"`]
+    assertions: [
+      {
+        text: passed ? `Output contains "${value}"` : `Output does not contain "${value}"`,
+        passed
+      }
+    ]
   };
 }
 function runContainsAnyAssertion(output, values) {
@@ -13160,8 +13112,12 @@ function runContainsAnyAssertion(output, values) {
   const passed = matched.length > 0;
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains "${matched[0]}"`] : [],
-    misses: passed ? [] : [`Output does not contain any of: ${values.map((v) => `"${v}"`).join(", ")}`]
+    assertions: [
+      {
+        text: passed ? `Output contains "${matched[0]}"` : `Output does not contain any of: ${values.map((v) => `"${v}"`).join(", ")}`,
+        passed
+      }
+    ]
   };
 }
 function runContainsAllAssertion(output, values) {
@@ -13169,16 +13125,24 @@ function runContainsAllAssertion(output, values) {
   const passed = missing.length === 0;
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains all ${values.length} expected strings`] : [],
-    misses: passed ? [] : [`Output missing: ${missing.map((v) => `"${v}"`).join(", ")}`]
+    assertions: [
+      {
+        text: passed ? `Output contains all ${values.length} expected strings` : `Output missing: ${missing.map((v) => `"${v}"`).join(", ")}`,
+        passed
+      }
+    ]
   };
 }
 function runIcontainsAssertion(output, value) {
   const passed = output.toLowerCase().includes(value.toLowerCase());
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains "${value}" (case-insensitive)`] : [],
-    misses: passed ? [] : [`Output does not contain "${value}" (case-insensitive)`]
+    assertions: [
+      {
+        text: passed ? `Output contains "${value}" (case-insensitive)` : `Output does not contain "${value}" (case-insensitive)`,
+        passed
+      }
+    ]
   };
 }
 function runIcontainsAnyAssertion(output, values) {
@@ -13187,9 +13151,11 @@ function runIcontainsAnyAssertion(output, values) {
   const passed = matched.length > 0;
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains "${matched[0]}" (case-insensitive)`] : [],
-    misses: passed ? [] : [
-      `Output does not contain any of: ${values.map((v) => `"${v}"`).join(", ")} (case-insensitive)`
+    assertions: [
+      {
+        text: passed ? `Output contains "${matched[0]}" (case-insensitive)` : `Output does not contain any of: ${values.map((v) => `"${v}"`).join(", ")} (case-insensitive)`,
+        passed
+      }
     ]
   };
 }
@@ -13199,24 +13165,36 @@ function runIcontainsAllAssertion(output, values) {
   const passed = missing.length === 0;
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output contains all ${values.length} expected strings (case-insensitive)`] : [],
-    misses: passed ? [] : [`Output missing (case-insensitive): ${missing.map((v) => `"${v}"`).join(", ")}`]
+    assertions: [
+      {
+        text: passed ? `Output contains all ${values.length} expected strings (case-insensitive)` : `Output missing (case-insensitive): ${missing.map((v) => `"${v}"`).join(", ")}`,
+        passed
+      }
+    ]
   };
 }
 function runStartsWithAssertion(output, value) {
   const passed = output.trim().startsWith(value.trim());
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output starts with "${value}"`] : [],
-    misses: passed ? [] : [`Output does not start with "${value}"`]
+    assertions: [
+      {
+        text: passed ? `Output starts with "${value}"` : `Output does not start with "${value}"`,
+        passed
+      }
+    ]
   };
 }
 function runEndsWithAssertion(output, value) {
   const passed = output.trim().endsWith(value.trim());
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output ends with "${value}"`] : [],
-    misses: passed ? [] : [`Output does not end with "${value}"`]
+    assertions: [
+      {
+        text: passed ? `Output ends with "${value}"` : `Output does not end with "${value}"`,
+        passed
+      }
+    ]
   };
 }
 function runRegexAssertion(output, pattern, flags) {
@@ -13225,8 +13203,12 @@ function runRegexAssertion(output, pattern, flags) {
   const flagsLabel = flags ? ` (flags: ${flags})` : "";
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output matches pattern /${pattern}/${flags ?? ""}${flagsLabel}`] : [],
-    misses: passed ? [] : [`Output does not match pattern /${pattern}/${flags ?? ""}${flagsLabel}`]
+    assertions: [
+      {
+        text: passed ? `Output matches pattern /${pattern}/${flags ?? ""}${flagsLabel}` : `Output does not match pattern /${pattern}/${flags ?? ""}${flagsLabel}`,
+        passed
+      }
+    ]
   };
 }
 function runIsJsonAssertion(output) {
@@ -13238,16 +13220,24 @@ function runIsJsonAssertion(output) {
   }
   return {
     score: passed ? 1 : 0,
-    hits: passed ? ["Output is valid JSON"] : [],
-    misses: passed ? [] : ["Output is not valid JSON"]
+    assertions: [
+      {
+        text: passed ? "Output is valid JSON" : "Output is not valid JSON",
+        passed
+      }
+    ]
   };
 }
 function runEqualsAssertion(output, value) {
   const passed = output.trim() === value.trim();
   return {
     score: passed ? 1 : 0,
-    hits: passed ? [`Output equals "${value}"`] : [],
-    misses: passed ? [] : [`Output does not equal "${value}"`]
+    assertions: [
+      {
+        text: passed ? `Output equals "${value}"` : `Output does not equal "${value}"`,
+        passed
+      }
+    ]
   };
 }
 
@@ -13460,10 +13450,8 @@ var InlineAssertEvaluator = class {
     return {
       score,
       verdict: scoreToVerdict(score),
-      hits: score >= 0.8 ? [result.name] : [],
-      misses: score < 0.5 ? [result.name] : [],
+      assertions: [{ text: result.name, passed: score >= 0.5 }],
       expectedAspectCount: 1,
-      reasoning: void 0,
       details: result.metadata ? result.metadata : void 0
     };
   }
@@ -13656,9 +13644,7 @@ var containsFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? `Output contains "${c.value}"` : `Output does not contain "${c.value}"`,
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13670,9 +13656,7 @@ var regexFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? `Output matches pattern /${c.value}/${c.flags ?? ""}` : `Output does not match pattern /${c.value}/${c.flags ?? ""}`,
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13683,9 +13667,7 @@ var isJsonFactory = () => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? "Output is valid JSON" : "Output is not valid JSON",
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13697,9 +13679,7 @@ var equalsFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? `Output equals "${c.value}"` : `Output does not equal "${c.value}"`,
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13711,9 +13691,7 @@ var containsAnyFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13725,9 +13703,7 @@ var containsAllFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13739,9 +13715,7 @@ var icontainsFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13753,9 +13727,7 @@ var icontainsAnyFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13767,9 +13739,7 @@ var icontainsAllFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13781,9 +13751,7 @@ var startsWithFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -13795,9 +13763,7 @@ var endsWithFactory = (config) => {
     return {
       score: result.score,
       verdict: result.score === 1 ? "pass" : "fail",
-      hits: result.hits,
-      misses: result.misses,
-      reasoning: result.score === 1 ? result.hits[0] : result.misses[0],
+      assertions: result.assertions,
       expectedAspectCount: 1
     };
   });
@@ -14868,7 +14834,7 @@ async function runEvaluation(options) {
         if (!cliModel) {
           throw new Error('--grader-target "agentv" requires --model (e.g., "openai:gpt-5-mini")');
         }
-        const { AgentvProvider: AgentvProvider2 } = await import("./agentv-provider-HDSAUUEF.js");
+        const { AgentvProvider: AgentvProvider2 } = await import("./agentv-provider-NFFLXG5M.js");
         return new AgentvProvider2("agentv", { model: cliModel, temperature: 0 });
       }
       const overrideTarget = resolveTargetByName(cliGraderTarget);
@@ -15203,8 +15169,7 @@ async function runEvaluation(options) {
             testId: evalCase.id,
             dataset: evalCase.dataset,
             score: 0,
-            hits: [],
-            misses: [],
+            assertions: [],
             answer: "",
             target: target.name,
             error: `Suite budget exceeded ($${cumulativeBudgetCost.toFixed(4)} / $${totalBudgetUsd.toFixed(4)})`,
@@ -15240,8 +15205,7 @@ async function runEvaluation(options) {
             testId: evalCase.id,
             dataset: evalCase.dataset,
             score: 0,
-            hits: [],
-            misses: [],
+            assertions: [],
             answer: "",
             target: target.name,
             error: errorMsg,
@@ -16208,11 +16172,9 @@ async function evaluateCandidate(options) {
     dataset: evalCase.dataset,
     conversationId: evalCase.conversation_id,
     score: score.score,
-    hits: score.hits,
-    misses: score.misses,
+    assertions: score.assertions,
     answer: candidate,
     target: target.name,
-    reasoning: score.reasoning,
     tokenUsage,
     costUsd,
     durationMs,
@@ -16386,9 +16348,7 @@ async function runEvaluatorList(options) {
         score: score2.score,
         weight,
         verdict: score2.verdict,
-        hits: score2.hits,
-        misses: score2.misses,
-        reasoning: score2.reasoning,
+        assertions: score2.assertions,
         evaluatorProviderRequest: score2.evaluatorRawRequest,
         details: score2.details,
         scores: mapChildResults(score2.scores),
@@ -16403,10 +16363,10 @@ async function runEvaluatorList(options) {
       const fallbackScore = {
         score: 0,
         verdict: "fail",
-        hits: [],
-        misses: [`Evaluator '${evaluatorConfig.name}' failed: ${message}`],
-        expectedAspectCount: 1,
-        reasoning: message
+        assertions: [
+          { text: `Evaluator '${evaluatorConfig.name}' failed: ${message}`, passed: false }
+        ],
+        expectedAspectCount: 1
       };
       const weight = evaluatorConfig.weight ?? 1;
       scored.push({
@@ -16422,9 +16382,12 @@ async function runEvaluatorList(options) {
         score: 0,
         weight,
         verdict: "fail",
-        hits: [],
-        misses: [`Evaluator '${evaluatorConfig.name ?? "unknown"}' failed: ${message}`],
-        reasoning: message,
+        assertions: [
+          {
+            text: `Evaluator '${evaluatorConfig.name ?? "unknown"}' failed: ${message}`,
+            passed: false
+          }
+        ],
         durationMs: endedAt.getTime() - startedAt.getTime(),
         startedAt: startedAt.toISOString(),
         endedAt: endedAt.toISOString()
@@ -16440,9 +16403,7 @@ async function runEvaluatorList(options) {
           ...scores[lastScoresIdx],
           score: negated.score,
           verdict: negated.verdict,
-          hits: [...negated.hits],
-          misses: [...negated.misses],
-          reasoning: negated.reasoning
+          assertions: [...negated.assertions]
         };
       }
     }
@@ -16457,21 +16418,13 @@ async function runEvaluatorList(options) {
   const aggregateScore = hasRequiredFailure ? 0 : scorable.length > 0 ? computeWeightedMean(
     scorable.map((entry) => ({ score: entry.score.score, weight: entry.weight }))
   ) : 0;
-  const hits = scored.flatMap((entry) => entry.score.hits);
-  const misses = scored.flatMap((entry) => entry.score.misses);
-  const expectedAspectCount = scored.reduce(
-    (total, entry) => total + (entry.score.expectedAspectCount ?? 0),
-    0
-  );
-  const reasoningParts = scored.map((entry) => entry.score.reasoning ? `${entry.name}: ${entry.score.reasoning}` : void 0).filter(isNonEmptyString);
-  const reasoning = reasoningParts.length > 0 ? reasoningParts.join(" | ") : void 0;
+  const assertions = scored.flatMap((entry) => entry.score.assertions);
+  const expectedAspectCount = assertions.length || 1;
   const score = {
     score: aggregateScore,
     verdict: scoreToVerdict(aggregateScore),
-    hits,
-    misses,
-    expectedAspectCount,
-    reasoning
+    assertions,
+    expectedAspectCount
   };
   return { score, scores };
 }
@@ -16575,8 +16528,7 @@ function buildErrorResult(evalCase, targetName, timestamp, error, promptInputs,
     dataset: evalCase.dataset,
     conversationId: evalCase.conversation_id,
     score: 0,
-    hits: [],
-    misses: [`Error: ${message}`],
+    assertions: [{ text: `Error: ${message}`, passed: false }],
     answer: `Error occurred: ${message}`,
     target: targetName,
     requests,
@@ -16686,9 +16638,7 @@ function mapChildResults(children) {
     score: child.score,
     weight: child.weight,
     verdict: child.verdict,
-    hits: child.hits,
-    misses: child.misses,
-    reasoning: child.reasoning,
+    assertions: child.assertions,
     evaluatorProviderRequest: child.evaluatorRawRequest,
     scores: mapChildResults(child.scores),
     details: child.details,
@@ -17653,7 +17603,6 @@ export {
   freeformEvaluationSchema,
   generateRubrics,
   getAgentvHome,
-  getHitCount,
   getOutputFilenames,
   getSubagentsRoot,
   getTraceStateRoot,