@agentv/core 3.6.0 → 3.8.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (14) hide show
  1. package/dist/{chunk-2IZOTQ25.js → chunk-3ZS3GCMI.js} +143 -3
  2. package/dist/chunk-3ZS3GCMI.js.map +1 -0
  3. package/dist/evaluation/validation/index.cjs +227 -39
  4. package/dist/evaluation/validation/index.cjs.map +1 -1
  5. package/dist/evaluation/validation/index.js +84 -5
  6. package/dist/evaluation/validation/index.js.map +1 -1
  7. package/dist/index.cjs +23 -22
  8. package/dist/index.cjs.map +1 -1
  9. package/dist/index.d.cts +6 -9
  10. package/dist/index.d.ts +6 -9
  11. package/dist/index.js +429 -562
  12. package/dist/index.js.map +1 -1
  13. package/package.json +1 -1
  14. package/dist/chunk-2IZOTQ25.js.map +0 -1
package/dist/index.d.cts CHANGED
@@ -251,10 +251,8 @@ interface TokenUsage {
251
251
  interface TraceSummary {
252
252
  /** Total number of events in trace */
253
253
  readonly eventCount: number;
254
- /** Unique tool names, sorted alphabetically */
255
- readonly toolNames: readonly string[];
256
254
  /** Map of tool name to call count */
257
- readonly toolCallsByName: Readonly<Record<string, number>>;
255
+ readonly toolCalls: Readonly<Record<string, number>>;
258
256
  /** Number of error events */
259
257
  readonly errorCount: number;
260
258
  /** Per-tool duration arrays in milliseconds (optional) */
@@ -1134,7 +1132,6 @@ interface EvaluationResult {
1134
1132
  readonly conversationId?: string;
1135
1133
  readonly score: number;
1136
1134
  readonly assertions: readonly AssertionEntry[];
1137
- readonly outputText: string;
1138
1135
  readonly target: string;
1139
1136
  /** Token usage metrics from provider (optional) */
1140
1137
  readonly tokenUsage?: TokenUsage;
@@ -1157,10 +1154,10 @@ interface EvaluationResult {
1157
1154
  readonly trace?: TraceSummary;
1158
1155
  /** Path to the temporary workspace directory (included on failure for debugging) */
1159
1156
  readonly workspacePath?: string;
1160
- /** Input messages or prompt string sent to the agent */
1161
- readonly input?: readonly Message[] | string;
1162
- /** Full output messages from agent execution (only included when --trace flag is set) */
1163
- readonly output?: readonly Message[];
1157
+ /** Input messages sent to the agent. Always Message[] for consistent shape with output. */
1158
+ readonly input?: readonly Message[];
1159
+ /** Output messages from agent execution. Always present — at minimum contains the final assistant message. */
1160
+ readonly output: readonly Message[];
1164
1161
  /** Captured output from workspace before_all script */
1165
1162
  readonly beforeAllOutput?: string;
1166
1163
  /** Captured output from workspace before_each script */
@@ -1202,7 +1199,7 @@ interface EvaluatorResult {
1202
1199
  readonly verdict?: EvaluationVerdict;
1203
1200
  readonly assertions: readonly AssertionEntry[];
1204
1201
  readonly rawRequest?: JsonObject;
1205
- readonly evaluatorProviderRequest?: JsonObject;
1202
+ readonly input?: JsonObject;
1206
1203
  readonly scores?: readonly EvaluatorResult[];
1207
1204
  /** Optional structured details from code graders (e.g., TP/TN/FP/FN counts). */
1208
1205
  readonly details?: JsonObject;
package/dist/index.d.ts CHANGED
@@ -251,10 +251,8 @@ interface TokenUsage {
251
251
  interface TraceSummary {
252
252
  /** Total number of events in trace */
253
253
  readonly eventCount: number;
254
- /** Unique tool names, sorted alphabetically */
255
- readonly toolNames: readonly string[];
256
254
  /** Map of tool name to call count */
257
- readonly toolCallsByName: Readonly<Record<string, number>>;
255
+ readonly toolCalls: Readonly<Record<string, number>>;
258
256
  /** Number of error events */
259
257
  readonly errorCount: number;
260
258
  /** Per-tool duration arrays in milliseconds (optional) */
@@ -1134,7 +1132,6 @@ interface EvaluationResult {
1134
1132
  readonly conversationId?: string;
1135
1133
  readonly score: number;
1136
1134
  readonly assertions: readonly AssertionEntry[];
1137
- readonly outputText: string;
1138
1135
  readonly target: string;
1139
1136
  /** Token usage metrics from provider (optional) */
1140
1137
  readonly tokenUsage?: TokenUsage;
@@ -1157,10 +1154,10 @@ interface EvaluationResult {
1157
1154
  readonly trace?: TraceSummary;
1158
1155
  /** Path to the temporary workspace directory (included on failure for debugging) */
1159
1156
  readonly workspacePath?: string;
1160
- /** Input messages or prompt string sent to the agent */
1161
- readonly input?: readonly Message[] | string;
1162
- /** Full output messages from agent execution (only included when --trace flag is set) */
1163
- readonly output?: readonly Message[];
1157
+ /** Input messages sent to the agent. Always Message[] for consistent shape with output. */
1158
+ readonly input?: readonly Message[];
1159
+ /** Output messages from agent execution. Always present — at minimum contains the final assistant message. */
1160
+ readonly output: readonly Message[];
1164
1161
  /** Captured output from workspace before_all script */
1165
1162
  readonly beforeAllOutput?: string;
1166
1163
  /** Captured output from workspace before_each script */
@@ -1202,7 +1199,7 @@ interface EvaluatorResult {
1202
1199
  readonly verdict?: EvaluationVerdict;
1203
1200
  readonly assertions: readonly AssertionEntry[];
1204
1201
  readonly rawRequest?: JsonObject;
1205
- readonly evaluatorProviderRequest?: JsonObject;
1202
+ readonly input?: JsonObject;
1206
1203
  readonly scores?: readonly EvaluatorResult[];
1207
1204
  /** Optional structured details from code graders (e.g., TP/TN/FP/FN counts). */
1208
1205
  readonly details?: JsonObject;