chore: import upstream snapshot with attribution

2026-07-13 12:32:26 +08:00
commit 1443d3fdf9
732 changed files with 196602 additions and 0 deletions
@@ -0,0 +1,129 @@
+from __future__ import annotations
+
+import argparse
+from datetime import UTC, datetime
+
+from langfuse import Langfuse
+
+
+PYTHON_TASK_ITEMS = [
+    {
+        "id": "py-task-001",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定 nums = [7, -2, 5, 11, -8, 4]，"
+            "只保留正数，平方后求和。请只输出最终整数。"
+        ),
+        "expected_output": "211",
+    },
+    {
+        "id": "py-task-002",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定文本 'alpha,beta;gamma delta\\nepsilon'，"
+            "按逗号、分号、空白和换行切分 token，统计 token 数。请只输出最终整数。"
+        ),
+        "expected_output": "5",
+    },
+    {
+        "id": "py-task-003",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：计算 1 到 200 中所有能被 7 整除但不能被 5 整除的整数个数。"
+            "请只输出最终整数。"
+        ),
+        "expected_output": "23",
+    },
+    {
+        "id": "py-task-004",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定 records = [('a', 3), ('b', 5), ('a', 4), ('c', 2), ('b', -1)]，"
+            "按 key 汇总数值，并按 key 字母序输出形如 a=7,b=4,c=2 的字符串。"
+        ),
+        "expected_output": "a=7,b=4,c=2",
+    },
+    {
+        "id": "py-task-005",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定矩阵 [[1, 2, 3], [4, 5, 6], [7, 8, 9]]，"
+            "计算主对角线之和减去副对角线之和。请只输出最终整数。"
+        ),
+        "expected_output": "0",
+    },
+    {
+        "id": "py-task-006",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定 s = 'Yuxi Agent Evaluation'，"
+            "忽略大小写统计元音字母 a/e/i/o/u 的总数。请只输出最终整数。"
+        ),
+        "expected_output": "10",
+    },
+    {
+        "id": "py-task-007",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定区间 [10, 50]，找出所有质数并求和。"
+            "请只输出最终整数。"
+        ),
+        "expected_output": "311",
+    },
+    {
+        "id": "py-task-008",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定 JSON 字符串 "
+            "'{\"items\":[{\"price\":12.5,\"qty\":2},{\"price\":3,\"qty\":5},{\"price\":8,\"qty\":1}]}'，"
+            "计算所有 price * qty 的总和。请只输出最终数字，保留一位小数。"
+        ),
+        "expected_output": "48.0",
+    },
+    {
+        "id": "py-task-009",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：给定 words = ['graph', 'rag', 'agent', 'tool', 'trace']，"
+            "按单词长度升序、长度相同时按字母序排序，并用 '-' 连接。"
+        ),
+        "expected_output": "rag-tool-agent-graph-trace",
+    },
+    {
+        "id": "py-task-010",
+        "input": (
+            "请用 Python 完成任务并给出最终答案：生成 Fibonacci 序列前 12 项（从 0, 1 开始），"
+            "取偶数项求和。请只输出最终整数。"
+        ),
+        "expected_output": "44",
+    },
+]
+
+
+def parse_args() -> argparse.Namespace:
+    parser = argparse.ArgumentParser(description="Upload a Python task dataset to Langfuse.")
+    parser.add_argument(
+        "--dataset-name",
+        default=f"yuxi-python-tasks-{datetime.now(UTC).strftime('%Y%m%dT%H%M%SZ')}",
+        help="Langfuse dataset name. Defaults to a timestamped name.",
+    )
+    return parser.parse_args()
+
+
+def main() -> None:
+    args = parse_args()
+    client = Langfuse()
+    client.create_dataset(
+        name=args.dataset_name,
+        description="Yuxi agent evaluation dataset: deterministic Python programming tasks.",
+        metadata={
+            "source": "scripts/eval/upload_langfuse_python_tasks_dataset.py",
+            "task_type": "python_programming",
+            "item_count": len(PYTHON_TASK_ITEMS),
+        },
+    )
+    for item in PYTHON_TASK_ITEMS:
+        client.create_dataset_item(
+            dataset_name=args.dataset_name,
+            id=item["id"],
+            input={"input": item["input"]},
+            expected_output=item["expected_output"],
+            metadata={"category": "python_programming", "source": "yuxi_eval_smoke"},
+        )
+    client.flush()
+    print(args.dataset_name)
+
+
+if __name__ == "__main__":
+    main()