@agentica/benchmark 0.8.2 → 0.8.3-dev.20250227

package/src/internal/AgenticaSelectBenchmarkReporter.ts

@@ -1,212 +1,210 @@
-import { IAgenticaTokenUsage } from "@agentica/core";
-
-import { IAgenticaSelectBenchmarkEvent } from "../structures/IAgenticaSelectBenchmarkEvent";
-import { IAgenticaSelectBenchmarkResult } from "../structures/IAgenticaSelectBenchmarkResult";
-import { MathUtil } from "../utils/MathUtil";
-import { AgenticaBenchmarkUtil } from "./AgenticaBenchmarkUtil";
-
-/**
- * @internal
- */
-export namespace AgenticaSelectBenchmarkReporter {
-  export const markdown = (
-    result: IAgenticaSelectBenchmarkResult,
-  ): Record<string, string> =>
-    Object.fromEntries([
-      ["./README.md", writeIndex(result)],
-      ...result.experiments
-        .map((exp) => [
-          [`./${exp.scenario.name}/README.md`, writeExperimentIndex(exp)],
-          ...exp.events.map((event, i) => [
-            `./${exp.scenario.name}/${i + 1}.${event.type}.md`,
-            writeExperimentEvent(event, i),
-          ]),
-        ])
-        .flat(),
-    ]);
-
-  const writeIndex = (result: IAgenticaSelectBenchmarkResult): string => {
-    const events: IAgenticaSelectBenchmarkEvent[] = result.experiments
-      .map((r) => r.events)
-      .flat();
-    const average: number =
-      events
-        .map((e) => e.completed_at.getTime() - e.started_at.getTime())
-        .reduce((a, b) => a + b, 0) / events.length;
-    const aggregate: IAgenticaTokenUsage.IComponent<"aggregate"> =
-      result.usage.aggregate;
-    return [
-      "# LLM Function Selection Benchmark",
-      "## Summary",
-      `  - Aggregation:`,
-      `    - Scenarios: #${result.experiments.length.toLocaleString()}`,
-      `    - Trial: ${events.length}`,
-      `    - Success: ${events.filter((e) => e.type === "success").length}`,
-      `    - Failure: ${events.filter((e) => e.type === "failure").length}`,
-      `    - Average Time: ${MathUtil.round(average).toLocaleString()} ms`,
-      `  - Token Usage`,
-      `    - Total: ${aggregate.total.toLocaleString()}`,
-      `    - Input`,
-      `      - Total: ${aggregate.input.total.toLocaleString()}`,
-      `      - Cached: ${aggregate.input.cached.toLocaleString()}`,
-      `    - Output:`,
-      `      - Total: ${aggregate.output.total.toLocaleString()}`,
-      `      - Accepted Prediction: ${aggregate.output.accepted_prediction.toLocaleString()}`,
-      `      - Reasoning: ${aggregate.output.reasoning.toLocaleString()}`,
-      `      - Rejected Prediction: ${aggregate.output.rejected_prediction.toLocaleString()}`,
-      "",
-      "## Experiments",
-      " Name | Status | Time/Avg  ",
-      ":-----|:-------|----------:",
-      ...result.experiments.map((exp) =>
-        [
-          `[${exp.scenario.name}](./${exp.scenario.name}/README.md)`,
-          (() => {
-            const success: number = Math.floor(
-              (exp.events.filter((e) => e.type === "success").length /
-                exp.events.length) *
-                10,
-            );
-            return (
-              new Array(success).fill("■").join("") +
-              new Array(10 - success).fill("□").join("")
-            );
-          })(),
-          MathUtil.round(
-            exp.events
-              .map(
-                (event) =>
-                  event.completed_at.getTime() - event.started_at.getTime(),
-              )
-              .reduce((a, b) => a + b, 0) / exp.events.length,
-          ).toLocaleString() + " ms",
-        ].join(" | "),
-      ),
-    ].join("\n");
-  };
-
-  const writeExperimentIndex = (
-    exp: IAgenticaSelectBenchmarkResult.IExperiment,
-  ): string => {
-    const aggregate: IAgenticaTokenUsage.IComponent<"aggregate"> =
-      exp.usage.aggregate;
-    return [
-      `# ${exp.scenario.name}`,
-      "## Summary",
-      "  - Aggregation:",
-      `    - Trial: ${exp.events.length}`,
-      `    - Success: ${exp.events.filter((e) => e.type === "success").length}`,
-      `    - Failure: ${exp.events.filter((e) => e.type === "failure").length}`,
-      `    - Average Time: ${MathUtil.round(
-        exp.events
-          .map(
-            (event) =>
-              event.completed_at.getTime() - event.started_at.getTime(),
-          )
-          .reduce((a, b) => a + b, 0) / exp.events.length,
-      ).toLocaleString()} ms`,
-      `  - Token Usage`,
-      `    - Total: ${aggregate.total.toLocaleString()}`,
-      `    - Input`,
-      `      - Total: ${aggregate.input.total.toLocaleString()}`,
-      `      - Cached: ${aggregate.input.cached.toLocaleString()}`,
-      `    - Output:`,
-      `      - Total: ${aggregate.output.total.toLocaleString()}`,
-      `      - Accepted Prediction: ${aggregate.output.accepted_prediction.toLocaleString()}`,
-      `      - Reasoning: ${aggregate.output.reasoning.toLocaleString()}`,
-      `      - Rejected Prediction: ${aggregate.output.rejected_prediction.toLocaleString()}`,
-      "",
-      "## Events",
-      " No | Type | Time",
-      "---:|:-----|----:",
-      ...exp.events.map((e, i) =>
-        [
-          `[${i + 1}.](./${i + 1}.${e.type}.md)`,
-          e.type,
-          MathUtil.round(e.completed_at.getTime() - e.started_at.getTime()) +
-            " ms",
-        ].join(" | "),
-      ),
-      "",
-      "## Scenario",
-      "### User Prompt",
-      exp.scenario.text,
-      "",
-      "### Expected",
-      "```json",
-      JSON.stringify(
-        AgenticaBenchmarkUtil.expectedToJson(exp.scenario.expected),
-        null,
-        2,
-      ),
-      "```",
-    ].join("\n");
-  };
-
-  const writeExperimentEvent = (
-    event: IAgenticaSelectBenchmarkEvent,
-    index: number,
-  ): string => {
-    return [
-      `# ${index + 1}. ${event.type}`,
-      `## Summary`,
-      `  - Name: ${event.scenario.name}`,
-      `  - Type: ${event.type}`,
-      `  - Time: ${(event.completed_at.getTime() - event.started_at.getTime()).toLocaleString()} ms`,
-      ...(event.type !== "error"
-        ? [
-            "  - Token Usage",
-            `    - Total: ${event.usage.aggregate.toLocaleString()}`,
-            `    - Prompt`,
-            `      - Total: ${event.usage.aggregate.input.total.toLocaleString()}`,
-            `      - Cached: ${event.usage.aggregate.input.cached.toLocaleString()}`,
-            `    - Completion:`,
-            `      - Total: ${event.usage.aggregate.output.total.toLocaleString()}`,
-            `      - Reasoning: ${event.usage.aggregate.output.reasoning.toLocaleString()}`,
-            `      - Accepted Prediction: ${event.usage.aggregate.output.accepted_prediction.toLocaleString()}`,
-            `      - Rejected Prediction: ${event.usage.aggregate.output.rejected_prediction.toLocaleString()}`,
-          ]
-        : []),
-      "",
-      "## Scenario",
-      "### User Prompt",
-      event.scenario.text,
-      "",
-      "### Expected",
-      "```json",
-      JSON.stringify(
-        AgenticaBenchmarkUtil.expectedToJson(event.scenario.expected),
-        null,
-        2,
-      ),
-      "```",
-      "",
-      ...(event.type === "success" || event.type === "failure"
-        ? [
-            "## Result",
-            ...event.selected.map((s) =>
-              [
-                `### ${s.name}`,
-                `  - Controller: \`${s.controller.name}\``,
-                `  - Function: \`${s.function.name}\``,
-                `  - Reason: ${s.reason}`,
-                "",
-                ...(s.function.description ? [s.function.description, ""] : []),
-              ].join("\n"),
-            ),
-          ]
-        : []),
-      ...(event.type === "error"
-        ? [
-            "## Error",
-            "```json",
-            AgenticaBenchmarkUtil.errorToJson(
-              JSON.stringify(event.error, null, 2),
-            ),
-            "```",
-            "",
-          ]
-        : []),
-    ].join("\n");
-  };
-}
+import { IAgenticaTokenUsage } from "@agentica/core";
+
+import { IAgenticaSelectBenchmarkEvent } from "../structures/IAgenticaSelectBenchmarkEvent";
+import { IAgenticaSelectBenchmarkResult } from "../structures/IAgenticaSelectBenchmarkResult";
+import { MathUtil } from "../utils/MathUtil";
+import { AgenticaBenchmarkUtil } from "./AgenticaBenchmarkUtil";
+
+/**
+ * @internal
+ */
+export namespace AgenticaSelectBenchmarkReporter {
+  export const markdown = (
+    result: IAgenticaSelectBenchmarkResult,
+  ): Record<string, string> =>
+    Object.fromEntries([
+      ["./README.md", writeIndex(result)],
+      ...result.experiments
+        .map((exp) => [
+          [`./${exp.scenario.name}/README.md`, writeExperimentIndex(exp)],
+          ...exp.events.map((event, i) => [
+            `./${exp.scenario.name}/${i + 1}.${event.type}.md`,
+            writeExperimentEvent(event, i),
+          ]),
+        ])
+        .flat(),
+    ]);
+
+  const writeIndex = (result: IAgenticaSelectBenchmarkResult): string => {
+    const events: IAgenticaSelectBenchmarkEvent[] = result.experiments
+      .map((r) => r.events)
+      .flat();
+    const average: number =
+      events
+        .map((e) => e.completed_at.getTime() - e.started_at.getTime())
+        .reduce((a, b) => a + b, 0) / events.length;
+    const aggregate: IAgenticaTokenUsage.IComponent = result.usage.aggregate;
+    return [
+      "# LLM Function Selection Benchmark",
+      "## Summary",
+      `  - Aggregation:`,
+      `    - Scenarios: #${result.experiments.length.toLocaleString()}`,
+      `    - Trial: ${events.length}`,
+      `    - Success: ${events.filter((e) => e.type === "success").length}`,
+      `    - Failure: ${events.filter((e) => e.type === "failure").length}`,
+      `    - Average Time: ${MathUtil.round(average).toLocaleString()} ms`,
+      `  - Token Usage`,
+      `    - Total: ${aggregate.total.toLocaleString()}`,
+      `    - Input`,
+      `      - Total: ${aggregate.input.total.toLocaleString()}`,
+      `      - Cached: ${aggregate.input.cached.toLocaleString()}`,
+      `    - Output:`,
+      `      - Total: ${aggregate.output.total.toLocaleString()}`,
+      `      - Accepted Prediction: ${aggregate.output.accepted_prediction.toLocaleString()}`,
+      `      - Reasoning: ${aggregate.output.reasoning.toLocaleString()}`,
+      `      - Rejected Prediction: ${aggregate.output.rejected_prediction.toLocaleString()}`,
+      "",
+      "## Experiments",
+      " Name | Status | Time/Avg  ",
+      ":-----|:-------|----------:",
+      ...result.experiments.map((exp) =>
+        [
+          `[${exp.scenario.name}](./${exp.scenario.name}/README.md)`,
+          (() => {
+            const success: number = Math.floor(
+              (exp.events.filter((e) => e.type === "success").length /
+                exp.events.length) *
+                10,
+            );
+            return (
+              new Array(success).fill("■").join("") +
+              new Array(10 - success).fill("□").join("")
+            );
+          })(),
+          MathUtil.round(
+            exp.events
+              .map(
+                (event) =>
+                  event.completed_at.getTime() - event.started_at.getTime(),
+              )
+              .reduce((a, b) => a + b, 0) / exp.events.length,
+          ).toLocaleString() + " ms",
+        ].join(" | "),
+      ),
+    ].join("\n");
+  };
+
+  const writeExperimentIndex = (
+    exp: IAgenticaSelectBenchmarkResult.IExperiment,
+  ): string => {
+    const aggregate: IAgenticaTokenUsage.IComponent = exp.usage.aggregate;
+    return [
+      `# ${exp.scenario.name}`,
+      "## Summary",
+      "  - Aggregation:",
+      `    - Trial: ${exp.events.length}`,
+      `    - Success: ${exp.events.filter((e) => e.type === "success").length}`,
+      `    - Failure: ${exp.events.filter((e) => e.type === "failure").length}`,
+      `    - Average Time: ${MathUtil.round(
+        exp.events
+          .map(
+            (event) =>
+              event.completed_at.getTime() - event.started_at.getTime(),
+          )
+          .reduce((a, b) => a + b, 0) / exp.events.length,
+      ).toLocaleString()} ms`,
+      `  - Token Usage`,
+      `    - Total: ${aggregate.total.toLocaleString()}`,
+      `    - Input`,
+      `      - Total: ${aggregate.input.total.toLocaleString()}`,
+      `      - Cached: ${aggregate.input.cached.toLocaleString()}`,
+      `    - Output:`,
+      `      - Total: ${aggregate.output.total.toLocaleString()}`,
+      `      - Accepted Prediction: ${aggregate.output.accepted_prediction.toLocaleString()}`,
+      `      - Reasoning: ${aggregate.output.reasoning.toLocaleString()}`,
+      `      - Rejected Prediction: ${aggregate.output.rejected_prediction.toLocaleString()}`,
+      "",
+      "## Events",
+      " No | Type | Time",
+      "---:|:-----|----:",
+      ...exp.events.map((e, i) =>
+        [
+          `[${i + 1}.](./${i + 1}.${e.type}.md)`,
+          e.type,
+          MathUtil.round(e.completed_at.getTime() - e.started_at.getTime()) +
+            " ms",
+        ].join(" | "),
+      ),
+      "",
+      "## Scenario",
+      "### User Prompt",
+      exp.scenario.text,
+      "",
+      "### Expected",
+      "```json",
+      JSON.stringify(
+        AgenticaBenchmarkUtil.expectedToJson(exp.scenario.expected),
+        null,
+        2,
+      ),
+      "```",
+    ].join("\n");
+  };
+
+  const writeExperimentEvent = (
+    event: IAgenticaSelectBenchmarkEvent,
+    index: number,
+  ): string => {
+    return [
+      `# ${index + 1}. ${event.type}`,
+      `## Summary`,
+      `  - Name: ${event.scenario.name}`,
+      `  - Type: ${event.type}`,
+      `  - Time: ${(event.completed_at.getTime() - event.started_at.getTime()).toLocaleString()} ms`,
+      ...(event.type !== "error"
+        ? [
+            "  - Token Usage",
+            `    - Total: ${event.usage.aggregate.toLocaleString()}`,
+            `    - Prompt`,
+            `      - Total: ${event.usage.aggregate.input.total.toLocaleString()}`,
+            `      - Cached: ${event.usage.aggregate.input.cached.toLocaleString()}`,
+            `    - Completion:`,
+            `      - Total: ${event.usage.aggregate.output.total.toLocaleString()}`,
+            `      - Reasoning: ${event.usage.aggregate.output.reasoning.toLocaleString()}`,
+            `      - Accepted Prediction: ${event.usage.aggregate.output.accepted_prediction.toLocaleString()}`,
+            `      - Rejected Prediction: ${event.usage.aggregate.output.rejected_prediction.toLocaleString()}`,
+          ]
+        : []),
+      "",
+      "## Scenario",
+      "### User Prompt",
+      event.scenario.text,
+      "",
+      "### Expected",
+      "```json",
+      JSON.stringify(
+        AgenticaBenchmarkUtil.expectedToJson(event.scenario.expected),
+        null,
+        2,
+      ),
+      "```",
+      "",
+      ...(event.type === "success" || event.type === "failure"
+        ? [
+            "## Result",
+            ...event.selected.map((s) =>
+              [
+                `### ${s.name}`,
+                `  - Controller: \`${s.controller.name}\``,
+                `  - Function: \`${s.function.name}\``,
+                `  - Reason: ${s.reason}`,
+                "",
+                ...(s.function.description ? [s.function.description, ""] : []),
+              ].join("\n"),
+            ),
+          ]
+        : []),
+      ...(event.type === "error"
+        ? [
+            "## Error",
+            "```json",
+            AgenticaBenchmarkUtil.errorToJson(
+              JSON.stringify(event.error, null, 2),
+            ),
+            "```",
+            "",
+          ]
+        : []),
+    ].join("\n");
+  };
+}

package/src/structures/IAgenticaBenchmarkExpected.ts

@@ -1,58 +1,58 @@
-import { IAgenticaOperation } from "@agentica/core";
-
-/**
- * Expected operation determinant.
- *
- * `IAgenticaBenchmarkExpected` is a type for determining what
- * operation is expected in the benchmarking.
- *
- * And `IAgenticaBenchmarkExpected` is an union type of 4 types,
- * especially designed for the detailed determination of the expected
- * operations.
- *
- * @author Samchon
- */
-export type IAgenticaBenchmarkExpected =
-  | IAgenticaBenchmarkExpected.IAllOf
-  | IAgenticaBenchmarkExpected.IAnyOf
-  | IAgenticaBenchmarkExpected.IArray
-  | IAgenticaBenchmarkExpected.IStandalone;
-export namespace IAgenticaBenchmarkExpected {
-  /**
-   * All of them must meet the condition, but sequence is not important.
-   */
-  export interface IAllOf {
-    type: "allOf";
-    allOf: Array<
-      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IAllOf>
-    >;
-  }
-
-  /**
-   * At least one of them must meet the condition.
-   */
-  export interface IAnyOf {
-    type: "anyOf";
-    anyOf: Array<
-      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IAnyOf>
-    >;
-  }
-
-  /**
-   * All of them must meet the condition, and sequence is important.
-   */
-  export interface IArray {
-    type: "array";
-    items: Array<
-      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IArray>
-    >;
-  }
-
-  /**
-   * Standalone operation.
-   */
-  export interface IStandalone {
-    type: "standalone";
-    operation: IAgenticaOperation;
-  }
-}
+import { IAgenticaOperation } from "@agentica/core";
+
+/**
+ * Expected operation determinant.
+ *
+ * `IAgenticaBenchmarkExpected` is a type for determining what
+ * operation is expected in the benchmarking.
+ *
+ * And `IAgenticaBenchmarkExpected` is an union type of 4 types,
+ * especially designed for the detailed determination of the expected
+ * operations.
+ *
+ * @author Samchon
+ */
+export type IAgenticaBenchmarkExpected =
+  | IAgenticaBenchmarkExpected.IAllOf
+  | IAgenticaBenchmarkExpected.IAnyOf
+  | IAgenticaBenchmarkExpected.IArray
+  | IAgenticaBenchmarkExpected.IStandalone;
+export namespace IAgenticaBenchmarkExpected {
+  /**
+   * All of them must meet the condition, but sequence is not important.
+   */
+  export interface IAllOf {
+    type: "allOf";
+    allOf: Array<
+      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IAllOf>
+    >;
+  }
+
+  /**
+   * At least one of them must meet the condition.
+   */
+  export interface IAnyOf {
+    type: "anyOf";
+    anyOf: Array<
+      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IAnyOf>
+    >;
+  }
+
+  /**
+   * All of them must meet the condition, and sequence is important.
+   */
+  export interface IArray {
+    type: "array";
+    items: Array<
+      Exclude<IAgenticaBenchmarkExpected, IAgenticaBenchmarkExpected.IArray>
+    >;
+  }
+
+  /**
+   * Standalone operation.
+   */
+  export interface IStandalone {
+    type: "standalone";
+    operation: IAgenticaOperation;
+  }
+}