mirror of
https://github.com/jxxghp/MoviePilot.git
synced 2026-10-29 11:27:43 +08:00
212 lines
8.7 KiB
Python
212 lines
8.7 KiB
Python
"""
|
||
校验 MoviePilot 与原生 Codex 的同条件评测报告并生成成对摘要。
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import json
|
||
from pathlib import Path
|
||
from typing import Any
|
||
|
||
MAX_REPORT_BYTES = 8 * 1024 * 1024
|
||
_MODEL_CONDITION_FIELDS = (
|
||
"requested_model", "reasoning_effort", "max_model_calls", "max_output_tokens",
|
||
"timeout_seconds", "harness_context_window", "auth_mode",
|
||
)
|
||
_EVIDENCE_KINDS = {
|
||
"moviepilot_live_model": "moviepilot",
|
||
"codex_native_controlled": "codex",
|
||
}
|
||
|
||
|
||
def _read_report(path: Path) -> dict[str, Any]:
|
||
"""
|
||
读取有限大小的 JSON 报告，拒绝目录、超限或非对象正文。
|
||
"""
|
||
if not path.is_file():
|
||
raise ValueError(f"评测报告不存在：{path}")
|
||
if path.stat().st_size > MAX_REPORT_BYTES:
|
||
raise ValueError("评测报告超过 8 MiB")
|
||
try:
|
||
payload = json.loads(path.read_text(encoding="utf-8"))
|
||
except (OSError, UnicodeError, ValueError) as error:
|
||
raise ValueError(f"评测报告不是有效 JSON：{path}") from error
|
||
if not isinstance(payload, dict):
|
||
raise ValueError(f"评测报告必须是 JSON 对象：{path}")
|
||
return payload
|
||
|
||
|
||
def _conditions(report: dict[str, Any]) -> dict[str, Any]:
|
||
"""
|
||
提取必须一致的模型条件，不把令牌或完整工具正文带入比较结果。
|
||
"""
|
||
model = report.get("model")
|
||
if not isinstance(model, dict):
|
||
raise ValueError("评测报告缺少模型条件")
|
||
return {field: model.get(field) for field in _MODEL_CONDITION_FIELDS}
|
||
|
||
|
||
def _metric(report: dict[str, Any], key: str) -> Any:
|
||
"""
|
||
读取报告中的公开轨迹指标，未知用量保持 None 而不是补零。
|
||
"""
|
||
if key == "total_tokens":
|
||
tokens = report.get("tokens")
|
||
return tokens.get("total_tokens") if isinstance(tokens, dict) else None
|
||
return report.get(key)
|
||
|
||
|
||
def _side(report: dict[str, Any], role: str) -> dict[str, Any]:
|
||
"""
|
||
生成不含最终正文、工具参数和凭据的单侧摘要。
|
||
"""
|
||
return {
|
||
"role": role,
|
||
"passed": report.get("passed"),
|
||
"violations": report.get("violations", []),
|
||
"intelligence_evaluated": report.get("intelligence_evaluated"),
|
||
"usage_complete": report.get("usage_complete"),
|
||
"model_calls": _metric(report, "model_calls"),
|
||
"completed_model_calls": _metric(report, "completed_model_calls"),
|
||
"failed_tool_calls": _metric(report, "failed_tool_calls"),
|
||
"tool_calls": _metric(report, "tool_calls"),
|
||
"duplicate_attempts": _metric(report, "duplicate_attempts"),
|
||
"side_effects": _metric(report, "side_effects"),
|
||
"elapsed_seconds": _metric(report, "elapsed_seconds"),
|
||
"total_tokens": _metric(report, "total_tokens"),
|
||
}
|
||
|
||
|
||
def _delta(moviepilot: dict[str, Any], codex: dict[str, Any], *, usage_comparable: bool = True) -> dict[str, Any]:
|
||
"""
|
||
计算两个方向的数值差；任一侧未知时保留 unknown。
|
||
"""
|
||
result: dict[str, Any] = {}
|
||
for key in ("model_calls", "completed_model_calls", "failed_tool_calls", "tool_calls",
|
||
"duplicate_attempts", "side_effects", "elapsed_seconds", "total_tokens"):
|
||
if key == "total_tokens" and not usage_comparable:
|
||
result[key] = None
|
||
continue
|
||
left, right = _metric(moviepilot, key), _metric(codex, key)
|
||
result[key] = None if not isinstance(left, (int, float)) or not isinstance(right, (int, float)) else left - right
|
||
return result
|
||
|
||
|
||
def _has_cancel_action(report: dict[str, Any]) -> bool:
|
||
"""
|
||
确认取消场景确实发出了子代理取消动作，而不是凭场景名放宽用量门禁。
|
||
"""
|
||
trace = report.get("agent_trace") if report.get("evidence_kind") == "moviepilot_live_model" else report.get("native_events")
|
||
if not isinstance(trace, list):
|
||
return False
|
||
for entry in trace:
|
||
if not isinstance(entry, dict):
|
||
continue
|
||
if entry.get("type") == "ai":
|
||
data = entry.get("data")
|
||
calls = data.get("tool_calls", []) if isinstance(data, dict) else []
|
||
if any(
|
||
isinstance(call, dict)
|
||
and str(call.get("name", "")).rsplit(".", 1)[-1] == "subagent_task"
|
||
and isinstance(call.get("args"), dict)
|
||
and str(call["args"].get("action", "")).strip().lower() == "cancel"
|
||
for call in calls
|
||
):
|
||
return True
|
||
if entry.get("type") not in {"item.started", "item.completed"}:
|
||
continue
|
||
item = entry.get("item")
|
||
if isinstance(item, dict) and str(item.get("tool", "")).rsplit(".", 1)[-1] in {
|
||
"close_agent", "interrupt_agent", "cancel",
|
||
}:
|
||
return True
|
||
return False
|
||
|
||
|
||
def _intentional_cancel_usage(report: dict[str, Any]) -> bool:
|
||
"""
|
||
确认取消场景的执行质量，并约束未完整用量只能来自一个取消请求。
|
||
"""
|
||
if report.get("scenario_id") != "subagent_cancel_recovery":
|
||
return False
|
||
calls = report.get("model_calls")
|
||
completed = report.get("completed_model_calls")
|
||
execution_completed = (
|
||
report.get("agent_execution_success") is True
|
||
if report.get("evidence_kind") == "moviepilot_live_model"
|
||
else report.get("native_turn_completed") is True and report.get("native_exit_code") == 0
|
||
)
|
||
return (
|
||
(
|
||
report.get("usage_complete") is True
|
||
or (type(calls) is int and type(completed) is int and calls - completed == 1)
|
||
)
|
||
and report.get("intelligence_evaluated") is True
|
||
and execution_completed
|
||
and report.get("runner_error_type") is None
|
||
and report.get("failed_tool_calls") == 0
|
||
and _has_cancel_action(report)
|
||
)
|
||
|
||
|
||
def compare_reports(moviepilot: dict[str, Any], codex: dict[str, Any]) -> dict[str, Any]:
|
||
"""
|
||
仅在同场景、同源码和同模型预算成立时生成 Codex 成对比较摘要。
|
||
"""
|
||
roles = {}
|
||
for report in (moviepilot, codex):
|
||
evidence_kind = report.get("evidence_kind")
|
||
role = _EVIDENCE_KINDS.get(evidence_kind)
|
||
if role is None:
|
||
raise ValueError("成对比较需要 MoviePilot live 与 Codex native 报告")
|
||
if role in roles:
|
||
raise ValueError("成对比较不能重复使用同一侧报告")
|
||
roles[role] = report
|
||
moviepilot, codex = roles["moviepilot"], roles["codex"]
|
||
|
||
for field in ("scenario_id", "scenario_sha256", "production_agent_sha256", "skills_sha256", "harness_sha256"):
|
||
if moviepilot.get(field) != codex.get(field):
|
||
raise ValueError(f"成对报告的 {field} 不一致")
|
||
moviepilot_conditions, codex_conditions = _conditions(moviepilot), _conditions(codex)
|
||
if moviepilot_conditions != codex_conditions:
|
||
raise ValueError("成对报告的模型、推理档位或预算不一致")
|
||
complete_usage = all(report.get("usage_complete") for report in (moviepilot, codex))
|
||
intentional_cancel_usage = (
|
||
moviepilot["scenario_id"] == "subagent_cancel_recovery"
|
||
and not complete_usage
|
||
and all(_intentional_cancel_usage(report) for report in (moviepilot, codex))
|
||
)
|
||
if not all(report.get("intelligence_evaluated") for report in (moviepilot, codex)):
|
||
raise ValueError("成对比较需要两侧都完成真实模型调用")
|
||
if not complete_usage and not intentional_cancel_usage:
|
||
raise ValueError("成对比较需要两侧都取得完整用量；主动取消场景只能按受限行为配对")
|
||
|
||
moviepilot_side, codex_side = _side(moviepilot, "moviepilot"), _side(codex, "codex")
|
||
return {
|
||
"schema_version": 1,
|
||
"codex_comparison": True,
|
||
"pair_valid": True,
|
||
"both_passed": bool(moviepilot.get("passed") and codex.get("passed")),
|
||
"scenario_id": moviepilot["scenario_id"],
|
||
"scenario_sha256": moviepilot["scenario_sha256"],
|
||
"production_agent_sha256": moviepilot["production_agent_sha256"],
|
||
"skills_sha256": moviepilot["skills_sha256"],
|
||
"harness_sha256": moviepilot["harness_sha256"],
|
||
"usage_comparable": complete_usage,
|
||
"usage_note": (
|
||
None if complete_usage else
|
||
"主动取消使至少一侧的一个模型请求未取得完整供应商用量；保留行为配对，token 与成本差不作比较"
|
||
),
|
||
"model": moviepilot_conditions,
|
||
"moviepilot": moviepilot_side,
|
||
"codex": codex_side,
|
||
"delta_moviepilot_minus_codex": _delta(moviepilot, codex, usage_comparable=complete_usage),
|
||
}
|
||
|
||
|
||
def compare_report_files(moviepilot_path: Path, codex_path: Path) -> dict[str, Any]:
|
||
"""
|
||
从两个路径读取报告并执行严格成对校验。
|
||
"""
|
||
return compare_reports(_read_report(moviepilot_path), _read_report(codex_path))
|