test(cost-calc): stop 182 global writes leaking out of the cost-calc suites

[yuneng-berri]

TLDR

Problem this solves:

• 58 tests set an env var and never put it back
• 18 more hand-roll copy-then-restore around two globals
• A hand-rolled restore is skipped when the test fails

How it solves it:

• One monkeypatch fixture the env-var tests ask for
• monkeypatch.setattr for the two config globals

User Flow

No end-user behavior changes. Both suites report the same counts, 89 and 174,
and a test that runs after them in the same worker now starts from the
environment it would have had on its own.

Relevant issues

Linear ticket

Pre-Submission checklist

• I have added meaningful tests
• The handful of test files covering my change pass locally
• My PR passes all required CI/CD checks (e.g., lint, schema.d.ts sync check, etc.)
• My PR's scope is as isolated as possible; it only solves 1 specific problem
• I have received a Greptile Confidence Score of at least 4/5 before requesting a maintainer review

Screenshots / Proof of Fix

Shared setup: run a suite, then read the process back in the same interpreter.
What is left over is what the next test in that worker inherits.

uv run python -c '
import os, litellm, pytest
before = {k: getattr(litellm, k) for k in ("cost_margin_config", "cost_discount_config")}
pytest.main(["tests/test_litellm/test_cost_calculator.py", "-q", "-p", "no:randomly"])
print("LITELLM_LOCAL_MODEL_COST_MAP after the run:", os.environ.get("LITELLM_LOCAL_MODEL_COST_MAP", "<unset>"))
for k, v in before.items():
    print(f"litellm.{k} after the run:", "restored" if getattr(litellm, k) == v else "LEAKED")
'

Before (ff02d5c)

the leak the suites leave behind

1. The command above

89 passed, 356 warnings in 0.47s
LITELLM_LOCAL_MODEL_COST_MAP after the run: True
litellm.cost_margin_config after the run: restored
litellm.cost_discount_config after the run: restored

2. The same command against tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py

174 passed, 2 warnings in 0.96s
LITELLM_LOCAL_MODEL_COST_MAP after the run: True

what the budget rules see

3. uv run python scripts/check_test_quality.py tests/test_litellm/test_cost_calculator.py tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py | grep -o 'TQ00[0-9]' | sort | uniq -c

   2 TQ003
  75 TQ004
 105 TQ005

After (eee8cb9)

the leak the suites leave behind

1. The same command

89 passed, 356 warnings in 0.44s
LITELLM_LOCAL_MODEL_COST_MAP after the run: <unset>
litellm.cost_margin_config after the run: restored
litellm.cost_discount_config after the run: restored

2. The same command against the other file

174 passed, 2 warnings in 0.96s
LITELLM_LOCAL_MODEL_COST_MAP after the run: <unset>

what the budget rules see

3. The same count

(no output: neither file has a TQ violation left)

4. uv run python scripts/test_quality_gate.py

OK: every TQ rule is within its test-suite ceiling (base origin/litellm_internal_staging)

Type

✅ Test

🧹 Refactoring

Caveats (if any)

• The two config globals already came back, on both sides
• The 117-entry conftest snapshot is what was catching them
• So that half is about not needing the snapshot, not a live leak
• Neutering the cost-map fixture leaves all 174 green either way
• TQ003 drops 3, one of which was stale before this branch
• TQ005's ceiling stays 4 under its count, exactly as on staging

Final Attestation

• The tests check the right things, including the edge cases, and regressions in the respective real-world customer use-cases are not possible after this PR

[greptile-apps]

Greptile Summary

The PR replaces hand-written environment and global-state restoration in the cost-calculation suites with pytest monkeypatch fixtures.

• Restores the local model-cost-map environment and module global after each affected test.
• Uses monkeypatch for cost margin and discount configuration changes.
• Tightens the test-quality budgets to reflect the removed direct mutations.

Confidence Score: 5/5

The PR appears safe to merge.

No blocking failure remains.

Important Files Changed

Filename	Overview
test-quality-budget.json	Lowers the TQ003, TQ004, and TQ005 ceilings to match the removed test anti-patterns.
tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py	Centralizes local model-cost-map setup and cleanup in a monkeypatch-backed fixture.
tests/test_litellm/test_cost_calculator.py	Replaces direct environment and LiteLLM configuration mutations with automatically restored monkeypatch operations.

_{Reviews (2): Last reviewed commit: "test(cost-calc): stop 182 global writes ..." | Re-trigger Greptile}

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[yuneng-berri]

@greptileai