simon
ff911cf6f7
Token Plan 迁移 / 配置热加载:
- configs/llm_models.yaml: 各场景切到 Token Plan(deepseek-v4.1-flash / qwen3.6-flash)
- 新增 configs/runtime_env.py: .env 按 (mtime_ns, size) 热加载并同步 os.environ,
统一 env_get 取值;llm / embedding / vectorstore / mcp / pipeline 改用 env_get
- configs/loader.py / scripts/run_scheduler.py 等配套调整
- 新增 tests/test_hot_reload.py
日报 AI 摘要为空修复(2026-09-25):
- 根因: 推理模型的 reasoning token 与正文共用 max_tokens, 预算 1500 被"思考"
占满 -> text_tokens=0 / finish_reason=length, 摘要静默为空且不重试
- daily_report 场景新增 max_tokens(默认 4000, YAML 保存即热生效);
LLMConfig 支持可选 max_tokens; 分块预算 800 -> 2000
- _llm_call 拆出 _call_once, 正文为空时自动加倍预算重试(上限 16000),
用尽才降级返回空串; 网络异常重试语义不变
- docs/user-guide.md 新增 FAQ; continuation.md 记录本次排查
- 已重跑 2026-09-25 日报(report_id=357)补回 466 字摘要
测试: 相关用例 56 passed(test_hot_reload 12 passed);
ruff 无新增问题; 3 个 crawler 既有失败与本改动无关
218 lines
8.4 KiB
Python
218 lines
8.4 KiB
Python
"""配置热加载测试：改 .env / llm_models.yaml 后无需重启即生效。
|
||
|
||
覆盖三层：
|
||
- configs/loader.py YAML 场景（(mtime, size) 失效缓存）
|
||
- configs/runtime_env .env（热更新 / 删除语义 / 外部显式覆盖优先）
|
||
- scripts/run_scheduler 定时任务热同步（改 SCHEDULE_TIMES 等）
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import os
|
||
from pathlib import Path
|
||
|
||
import pytest
|
||
|
||
from configs import loader, runtime_env
|
||
|
||
|
||
@pytest.fixture(autouse=True)
|
||
def _clean_caches() -> None:
|
||
"""每个用例前后都清干净，避免托管的环境变量污染其它测试。"""
|
||
loader.clear_cache()
|
||
runtime_env.reset_cache()
|
||
yield
|
||
loader.clear_cache()
|
||
runtime_env.reset_cache()
|
||
|
||
|
||
def _use_env_file(tmp_path: Path, monkeypatch: pytest.MonkeyPatch, content: str) -> Path:
|
||
"""把 runtime_env 指向临时 .env，返回其路径。"""
|
||
env = tmp_path / ".env"
|
||
env.write_text(content, encoding="utf-8")
|
||
monkeypatch.setenv(runtime_env.ENV_FILE_OVERRIDE, str(env))
|
||
runtime_env.reset_cache()
|
||
return env
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# YAML 场景热加载
|
||
# --------------------------------------------------------------------------- #
|
||
|
||
def test_yaml_scene_hot_reload(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""改 llm_models.yaml 后，下一次 load_scene_config 立即读到新值。"""
|
||
cfg = tmp_path / "llm_models.yaml"
|
||
cfg.write_text(
|
||
"scenes:\n daily_report:\n provider: qwen\n model: qwen3.6-flash\n",
|
||
encoding="utf-8",
|
||
)
|
||
monkeypatch.setenv(loader.MODELS_CONFIG_OVERRIDE, str(cfg))
|
||
|
||
assert loader.load_scene_config("daily_report")["model"] == "qwen3.6-flash"
|
||
|
||
# 内容长度不同，保证 (mtime, size) 签名一定变化
|
||
cfg.write_text(
|
||
"scenes:\n daily_report:\n provider: qwen\n model: deepseek-v4.1-flash\n",
|
||
encoding="utf-8",
|
||
)
|
||
# 不重启、不 clear_cache 也应生效
|
||
assert loader.load_scene_config("daily_report")["model"] == "deepseek-v4.1-flash"
|
||
|
||
|
||
def test_yaml_scene_removed_falls_back(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""场景被删掉后返回空 dict（由调用方回退 .env）。"""
|
||
cfg = tmp_path / "llm_models.yaml"
|
||
cfg.write_text("scenes:\n stock_report:\n model: x\n", encoding="utf-8")
|
||
monkeypatch.setenv(loader.MODELS_CONFIG_OVERRIDE, str(cfg))
|
||
assert loader.load_scene_config("stock_report")["model"] == "x"
|
||
|
||
cfg.write_text("scenes: {}\n", encoding="utf-8")
|
||
assert loader.load_scene_config("stock_report") == {}
|
||
|
||
|
||
def test_yaml_missing_file_returns_empty(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
monkeypatch.setenv(loader.MODELS_CONFIG_OVERRIDE, str(tmp_path / "nope.yaml"))
|
||
assert loader.load_scene_config("daily_report") == {}
|
||
assert loader.load_defaults() == {}
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# .env 热加载
|
||
# --------------------------------------------------------------------------- #
|
||
|
||
def test_env_hot_reload(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""改 .env 后 env_get 立即读到新值。"""
|
||
env = _use_env_file(tmp_path, monkeypatch, "HOT_RELOAD_KEY=one\n")
|
||
monkeypatch.delenv("HOT_RELOAD_KEY", raising=False)
|
||
|
||
assert runtime_env.env_get("HOT_RELOAD_KEY") == "one"
|
||
|
||
env.write_text("HOT_RELOAD_KEY=two-longer-value\n", encoding="utf-8")
|
||
assert runtime_env.env_get("HOT_RELOAD_KEY") == "two-longer-value"
|
||
|
||
|
||
def test_env_removed_key_is_unset(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""从 .env 删掉的托管键要同步从 os.environ 移除。"""
|
||
env = _use_env_file(tmp_path, monkeypatch, "HOT_REMOVE_KEY=abc\n")
|
||
monkeypatch.delenv("HOT_REMOVE_KEY", raising=False)
|
||
assert runtime_env.env_get("HOT_REMOVE_KEY") == "abc"
|
||
assert os.environ["HOT_REMOVE_KEY"] == "abc"
|
||
|
||
env.write_text("OTHER_KEY=1\n", encoding="utf-8")
|
||
assert runtime_env.env_get("HOT_REMOVE_KEY") is None
|
||
assert "HOT_REMOVE_KEY" not in os.environ
|
||
|
||
|
||
def test_env_empty_value_is_unset(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
_use_env_file(tmp_path, monkeypatch, "HOT_EMPTY_KEY=\n")
|
||
monkeypatch.delenv("HOT_EMPTY_KEY", raising=False)
|
||
assert runtime_env.env_get("HOT_EMPTY_KEY", "fallback") == "fallback"
|
||
|
||
|
||
def test_explicit_process_env_wins(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""进程环境显式设置且与文件不同 → 不被 .env 覆盖（一次性覆盖语义）。"""
|
||
env = _use_env_file(tmp_path, monkeypatch, "HOT_OVERRIDE_KEY=from-file\n")
|
||
monkeypatch.setenv("HOT_OVERRIDE_KEY", "from-shell")
|
||
|
||
assert runtime_env.env_get("HOT_OVERRIDE_KEY") == "from-shell"
|
||
|
||
env.write_text("HOT_OVERRIDE_KEY=from-file-changed\n", encoding="utf-8")
|
||
assert runtime_env.env_get("HOT_OVERRIDE_KEY") == "from-shell"
|
||
|
||
|
||
def test_llm_config_follows_env_hot_reload(
|
||
tmp_path: Path, monkeypatch: pytest.MonkeyPatch
|
||
) -> None:
|
||
"""端到端：改 .env 里的 QWEN_MODEL，load_llm_config 立即用新模型。"""
|
||
from llm.client import load_llm_config
|
||
|
||
contents = (
|
||
"LLM_PROVIDER=qwen\n"
|
||
"QWEN_API_KEY=sk-test\n"
|
||
"QWEN_BASE_URL=https://token-plan.example/compatible-mode/v1\n"
|
||
"QWEN_MODEL=qwen3.6-flash\n"
|
||
)
|
||
env = _use_env_file(tmp_path, monkeypatch, contents)
|
||
for key in ("LLM_PROVIDER", "QWEN_API_KEY", "QWEN_BASE_URL", "QWEN_MODEL"):
|
||
monkeypatch.delenv(key, raising=False)
|
||
|
||
cfg = load_llm_config(scene="")
|
||
assert cfg.provider == "qwen"
|
||
assert cfg.model == "qwen3.6-flash"
|
||
|
||
env.write_text(contents.replace("qwen3.6-flash", "deepseek-v4.1-flash"), encoding="utf-8")
|
||
assert load_llm_config(scene="").model == "deepseek-v4.1-flash"
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# 调度任务热同步
|
||
# --------------------------------------------------------------------------- #
|
||
|
||
def test_desired_jobs_tracks_env(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""SCHEDULE_TIMES / CNINFO_SCHEDULE_TIME / STOCK_REPORT_TIME 改完立即反映。"""
|
||
from scripts.run_scheduler import _desired_jobs
|
||
|
||
_use_env_file(
|
||
tmp_path,
|
||
monkeypatch,
|
||
"SCHEDULE_TIMES=09:15\nCNINFO_SCHEDULE_TIME=05:45\nSTOCK_REPORT_TIME=\n",
|
||
)
|
||
|
||
jobs = _desired_jobs()
|
||
assert set(jobs) == {"pipeline_0915", "pipeline_cninfo"}
|
||
# 当天唯一次数 → 附带日报步骤
|
||
assert jobs["pipeline_0915"]["steps"][-1] == "report"
|
||
assert jobs["pipeline_cninfo"]["hour"] == 5
|
||
assert jobs["pipeline_cninfo"]["minute"] == 45
|
||
|
||
|
||
def test_sync_jobs_adds_and_removes(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""_sync_jobs 只按需增删：加时间点、启停个股日报、幂等。"""
|
||
from apscheduler.schedulers.background import BackgroundScheduler
|
||
|
||
import scripts.run_scheduler as rs
|
||
|
||
env = _use_env_file(
|
||
tmp_path,
|
||
monkeypatch,
|
||
"SCHEDULE_TIMES=09:15\nCNINFO_SCHEDULE_TIME=05:45\nSTOCK_REPORT_TIME=\n",
|
||
)
|
||
rs._jobs_sig = None
|
||
|
||
sched = BackgroundScheduler()
|
||
assert rs._sync_jobs(sched) is True
|
||
assert {j.id for j in sched.get_jobs()} == {"pipeline_0915", "pipeline_cninfo"}
|
||
|
||
# 新增一个时间点 + 启用个股日报
|
||
env.write_text(
|
||
"SCHEDULE_TIMES=09:15,16:40\nCNINFO_SCHEDULE_TIME=05:45\nSTOCK_REPORT_TIME=07:30\n",
|
||
encoding="utf-8",
|
||
)
|
||
assert rs._sync_jobs(sched) is True
|
||
ids = {j.id for j in sched.get_jobs()}
|
||
assert ids == {"pipeline_0915", "pipeline_1640", "pipeline_cninfo", "stock_report"}
|
||
|
||
# 配置未变 → 不重复变更
|
||
assert rs._sync_jobs(sched) is False
|
||
|
||
|
||
def test_parse_hhmm_falls_back_on_bad_input() -> None:
|
||
from scripts.run_scheduler import _parse_hhmm
|
||
|
||
assert _parse_hhmm("06:30", (0, 0)) == (6, 30)
|
||
assert _parse_hhmm("99:99", (6, 30)) == (6, 30)
|
||
assert _parse_hhmm("garbage", (7, 0)) == (7, 0)
|
||
|
||
|
||
# --------------------------------------------------------------------------- #
|
||
# 其它模块复用热读取
|
||
# --------------------------------------------------------------------------- #
|
||
|
||
def test_reporter_days_back_is_lazy(tmp_path: Path, monkeypatch: pytest.MonkeyPatch) -> None:
|
||
"""日报的 STOCK_REPORT_DAYS 走懒读取，不再冻结在模块常量里。"""
|
||
from scheduler import reporter
|
||
|
||
_use_env_file(tmp_path, monkeypatch, "STOCK_REPORT_DAYS=15\n")
|
||
monkeypatch.delenv("STOCK_REPORT_DAYS", raising=False)
|
||
assert reporter._cninfo_days_back() == 15
|