test(agent): codex TTFB fixture starts from a clean HERMES_CODEX_TTFB_* env

The #64507 test already assumed 'no override' without clearing the shell; fold
the delenv into _make_codex_agent (as the reasoning-effort sibling does) and
drop the per-test helper. The explicit-cap test pins reasoning off too, so the
effort floor can never mask a cap regression.
This commit is contained in:
kshitijk4poor
2026-09-19 00:41:18 +05:30
committed by kshitij
parent e3bde9d9c5
commit b1968795b2

View File

@@ -41,6 +41,11 @@ def _make_codex_agent(
monkeypatch.setenv("HERMES_HOME", str(tmp_path))
(tmp_path / ".env").write_text("", encoding="utf-8")
(tmp_path / "config.yaml").write_text("{}\n", encoding="utf-8")
# Every test here reasons about the built-in TTFB defaults; a developer shell override
# must not leak in (tests that need an override setenv it after this).
for name in ("HERMES_CODEX_TTFB_TIMEOUT_SECONDS", "HERMES_CODEX_TTFB_MAX_SECONDS",
"HERMES_CODEX_TTFB_DISABLE_ABOVE_TOKENS", "HERMES_CODEX_TTFB_STRICT"):
monkeypatch.delenv(name, raising=False)
from run_agent import AIAgent
agent = AIAgent(
@@ -595,16 +600,6 @@ def test_large_codex_request_hard_ceiling_reclaims_silent_stall(tmp_path, monkey
stop["flag"] = True
def _clear_ttfb_env(monkeypatch):
for name in (
"HERMES_CODEX_TTFB_TIMEOUT_SECONDS",
"HERMES_CODEX_TTFB_MAX_SECONDS",
"HERMES_CODEX_TTFB_DISABLE_ABOVE_TOKENS",
"HERMES_CODEX_TTFB_STRICT",
):
monkeypatch.delenv(name, raising=False)
def test_large_request_keeps_scaled_ttfb_instead_of_recapping(tmp_path, monkeypatch):
"""#91621 regression: with no TTFB env overrides, a >100k-token openai-codex
request scales the no-byte cutoff up to the 180s idle default — the cap must
@@ -612,7 +607,6 @@ def test_large_request_keeps_scaled_ttfb_instead_of_recapping(tmp_path, monkeypa
from agent import chat_completion_helpers as h
agent = _make_codex_agent(tmp_path, monkeypatch)
_clear_ttfb_env(monkeypatch)
agent.reasoning_config = {"enabled": False} # no effort floor: isolate the cap interaction
huge_input = "x" * 440_000 # ~110k estimated tokens → largest idle bucket
@@ -629,7 +623,7 @@ def test_explicit_ttfb_max_seconds_still_caps(tmp_path, monkeypatch):
from agent import chat_completion_helpers as h
agent = _make_codex_agent(tmp_path, monkeypatch)
_clear_ttfb_env(monkeypatch)
agent.reasoning_config = {"enabled": False}
monkeypatch.setenv("HERMES_CODEX_TTFB_MAX_SECONDS", "90")
huge_input = "x" * 440_000