@wix/evalforge-evaluator 0.183.0 → 0.184.0

package/build/index.js

@@ -8670,9 +8670,12 @@ function processMessages(timestampedMessages, startTime, endTime) {
       if (!step.toolCalls) continue;
       for (const tc of step.toolCalls) {
         if (tc.toolUseId && toolResultErrors.has(tc.toolUseId)) {
-          step.hasToolError = true;
-          step.toolErrorContent = toolResultErrors.get(tc.toolUseId);
-          break;
+          tc.isError = true;
+          tc.errorContent = toolResultErrors.get(tc.toolUseId);
+          if (!step.hasToolError) {
+            step.hasToolError = true;
+            step.toolErrorContent = tc.errorContent;
+          }
         }
       }
     }
@@ -8776,8 +8779,8 @@ function buildLLMTraceFromSteps(steps, totalDurationMs, usage, model) {
         },
         costUsd: stepCost / totalSubSteps,
         outputPreview: step.thinking?.slice(0, 200),
-        success: isSuccess,
-        error: errorMsg
+        success: true,
+        error: void 0
       });
     }
     if (toolCallCount > 0) {
@@ -8787,6 +8790,8 @@ function buildLLMTraceFromSteps(steps, totalDurationMs, usage, model) {
         const toolBudgetSteps = toolSubSteps + textSubSteps;
         const toolFraction = toolBudgetSteps > 0 ? 1 / toolBudgetSteps : 1;
         const remainingFraction = (totalSubSteps - thinkingSubSteps) / totalSubSteps;
+        const toolSuccess = !tc.isError;
+        const toolError = tc.isError ? tc.errorContent ?? "Tool call failed" : void 0;
         subSteps.push({
           id: (0, import_crypto2.randomUUID)(),
           stepNumber: 0,
@@ -8811,8 +8816,8 @@ function buildLLMTraceFromSteps(steps, totalDurationMs, usage, model) {
           toolName: tc.toolName,
           toolArguments: JSON.stringify(tc.args),
           outputPreview: tcIdx === 0 && !hasText ? (step.text || step.thinking)?.slice(0, 200) : void 0,
-          success: isSuccess,
-          error: errorMsg
+          success: toolSuccess,
+          error: toolError
         });
       }
     }
@@ -8833,8 +8838,8 @@ function buildLLMTraceFromSteps(steps, totalDurationMs, usage, model) {
         },
         costUsd: stepCost - subSteps.reduce((s, ss) => s + ss.costUsd, 0),
         outputPreview: step.text?.slice(0, 200),
-        success: isSuccess,
-        error: errorMsg
+        success: true,
+        error: void 0
       });
     }
     if (subSteps.length === 0) {