fix: fail loud on diag model fingerprint drift vs .env SEARCH_LLM_MODEL

Codex CHANGES_REQUESTED 复审:
- Critical:diag_fingerprint 用 config.diag.model,但 Phase 1 诊断 LLM 从 .env
  SEARCH_LLM_MODEL 构建,两者不一致会让信号以错误模型指纹落库,破坏可复现/
  resume/隔离。build_diagnosis_deps 新增 expected_model 参数,Phase 1 执行前
  fail loud 校验 config.model == settings.search_llm_model(附两值)。
- Minor:config/video_split.yaml diag.model 注释由 JUDGE_LLM_MODEL 更正为
  SEARCH_LLM_MODEL,与实现对齐。
- 补两个单测:模型不一致 fail loud + 缺凭证 fail loud。

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
2026-07-15 13:20:51 -04:00
parent 2844732126
commit 02b8145b7f
3 changed files with 55 additions and 4 deletions
+36
View File
@@ -161,6 +161,42 @@ def test_dry_run_computes_fingerprint_without_llm(monkeypatch, tmp_path, capsys)
assert not (tmp_path / "_dry_run_signals.db").exists()
def test_build_diagnosis_deps_model_mismatch_fails_loud(monkeypatch, tmp_path):
"""config.diag.model 与 .env SEARCH_LLM_MODEL 不一致 → fail loud(指纹漂移防护)。"""
class _FakeSettings:
search_llm_model = "actual-model-in-env"
search_llm_base_url = "https://api.example"
search_llm_api_key = "sk-xxx"
monkeypatch.setattr(cli, "_DiagLLMSettings", lambda: _FakeSettings())
with pytest.raises(SystemExit) as exc:
cli.build_diagnosis_deps(
harness_db=tmp_path / "h.db",
concurrency=2,
expected_model="deepseek-v4-pro", # 与 env 不一致
)
# 报错须同时暴露两个值,便于人对齐
msg = str(exc.value)
assert "deepseek-v4-pro" in msg
assert "actual-model-in-env" in msg
def test_build_diagnosis_deps_missing_credentials_fails_loud(monkeypatch, tmp_path):
""".env 缺 search LLM 凭证 → fail loud(先于模型一致性校验)。"""
class _EmptySettings:
search_llm_model = ""
search_llm_base_url = ""
search_llm_api_key = ""
monkeypatch.setattr(cli, "_DiagLLMSettings", lambda: _EmptySettings())
with pytest.raises(SystemExit):
cli.build_diagnosis_deps(
harness_db=tmp_path / "h.db", concurrency=2, expected_model="deepseek-v4-pro"
)
def test_git_short_sha_nonempty():
"""仓库内 git_short_sha 返回非空短 SHA。"""
sha = cli.git_short_sha()