feat: Phase 4 — Experiment 自动归档 + 异步 Job(状态机 / SSE / 一键复跑)

- 数据表:job / experiment(spec/result JSON 存档、code_version),Alembic 迁移 53113c80257f
- Job:queued→running→(success|failed) 状态机,BackgroundTasks 本地执行 + 失败兜底标记;结果与 Experiment 关联
- Experiment:每次研究成功自动归档(含 git commit 与收益摘要),支持一键复跑(同 spec 重建 Job)
- API:POST /api/jobs、GET /api/jobs/{id}(内嵌结果)、SSE /api/jobs/{id}/events、/api/experiments 列表/详情/rerun
- 前端:新增「实验」页(列表 / 详情 / 复跑 + Job 轮询);导航更新
- 端到端验证:真实 20 股 job 提交→后台执行→success→EXP 归档(-12.41%);executor 成功/失败路径单测
- 测试 73 passed(新增 5 项 Job/Experiment)/ ruff clean / 前端 tsc + build 通过
This commit is contained in:
Simon
2026-09-06 17:18:46 +08:00
parent 92627f5b6b
commit 0ea229d766
14 changed files with 938 additions and 2 deletions
@@ -0,0 +1,146 @@
"""Job 执行编排与 Experiment 归档(Phase 4)。
execute_job 在后台运行:状态机 queued→running→(success|failed),
成功时自动把 spec/result 存为 Experiment(含代码版本),实现「研究可复现」(AGENT.md §21)。
独立 Session 生命周期(不依赖请求 scope),可被 FastAPI BackgroundTasks 或测试直接调用。
"""
from __future__ import annotations
import json
import subprocess
import uuid
from collections.abc import Callable
from datetime import datetime
from pathlib import Path
from app.domain.entities.research import (
BacktestResult,
ExperimentRecord,
FactorTestReport,
JobStatus,
ResearchSpec,
)
from app.quant.service import ResearchService
PROJECT_ROOT = Path(__file__).resolve().parents[3]
def new_id(prefix: str) -> str:
return f"{prefix}-{uuid.uuid4().hex[:8].upper()}"
def _git_short_rev() -> str | None:
try:
out = subprocess.run(
["git", "rev-parse", "--short", "HEAD"],
cwd=PROJECT_ROOT,
capture_output=True,
text=True,
timeout=5,
check=False,
)
return out.stdout.strip() or None
except Exception: # noqa: BLE001
return None
def _summary_text(kind: str, result: BacktestResult | FactorTestReport) -> str | None:
if kind == "backtest" and isinstance(result, BacktestResult):
s = result.summary
return f"总收益 {s.total_return_pct:.2f}% · 年化 {s.annual_return_pct:.2f}% · 回撤 {s.max_drawdown_pct:.2f}%"
if isinstance(result, FactorTestReport):
return f"IC {result.ic_mean:.4f} · RankIC {result.rank_ic_mean:.4f} · 样本 {result.sample_days} 日"
return None
def _execute_inner(
job_id: str,
*,
session_factory: Callable,
job_repo_factory: Callable,
experiment_repo_factory: Callable,
stock_repo_factory: Callable,
daily_repo_factory: Callable,
engine,
) -> None:
with session_factory() as session:
job_repo = job_repo_factory(session)
experiment_repo = experiment_repo_factory(session)
job = job_repo.get(job_id)
if job is None:
return
job.status = JobStatus.RUNNING
job.started_at = datetime.now()
job_repo.update(job)
session.commit()
try:
spec = ResearchSpec.model_validate_json(job.spec_json)
service = ResearchService(
stock_repo_factory(session), daily_repo_factory(session), engine
)
if spec.type == "backtest":
result = service.run_backtest(spec)
else:
result = service.run_factor_test(spec)
result_json = json.dumps(result.model_dump(mode="json"), ensure_ascii=False)
experiment = ExperimentRecord(
id=new_id("EXP"),
kind=spec.type,
spec_json=job.spec_json,
result_json=result_json,
summary_text=_summary_text(spec.type, result),
code_version=_git_short_rev(),
job_id=job.id,
created_at=datetime.now(),
)
experiment_repo.save(experiment)
job.result_json = result_json
job.experiment_id = experiment.id
job.status = JobStatus.SUCCESS
job.error = None
except Exception as exc: # noqa: BLE001 —— 统一记为 failed 供前端展示
job.status = JobStatus.FAILED
job.error = f"{type(exc).__name__}: {exc}"
finally:
job.finished_at = datetime.now()
job_repo.update(job)
session.commit()
def execute_job(
job_id: str,
*,
session_factory: Callable,
job_repo_factory: Callable,
experiment_repo_factory: Callable,
stock_repo_factory: Callable,
daily_repo_factory: Callable,
engine,
) -> None:
"""入口包装:任何未预期异常都将 Job 标记 failed(防止卡在 queued/running)。"""
try:
_execute_inner(
job_id,
session_factory=session_factory,
job_repo_factory=job_repo_factory,
experiment_repo_factory=experiment_repo_factory,
stock_repo_factory=stock_repo_factory,
daily_repo_factory=daily_repo_factory,
engine=engine,
)
except Exception as exc: # noqa: BLE001
try:
with session_factory() as session:
repo = job_repo_factory(session)
job = repo.get(job_id)
if job is not None:
job.status = JobStatus.FAILED
job.error = f"内部错误: {type(exc).__name__}: {exc}"
job.finished_at = datetime.now()
repo.update(job)
session.commit()
except Exception: # noqa: BLE001
pass