[Fix] WatsonX Tests Failing on CI Due to Missing Env Vars

[yuneng-berri]

Summary

Failure Path (Before Fix)

WatsonX tests in tests/llm_translation/test_watsonx.py fail on CI with assert mock_post.call_count == 0 == 1. The tests are fully mocked but the WatsonX provider requires WATSONX_URL and WATSONX_PROJECT_ID env vars to pass validation before reaching the mocked HTTP client. Without them, a validation error is raised and silently swallowed by the fixture's except block.

Fix

Added an autouse fixture that sets dummy WATSONX_URL and WATSONX_PROJECT_ID via monkeypatch, and clears WATSONX_ZENAPIKEY/WATSONX_TOKEN to prevent env leakage from .env files loaded by dotenv.load_dotenv(). No real API keys or CI/CD env var changes needed.

Testing

All 11 tests in tests/llm_translation/test_watsonx.py pass locally.

Type

🐛 Bug Fix
✅ Test

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
litellm	Ready	Preview, Comment	Mar 30, 2026 8:28pm

[CLAassistant]

All committers have signed the CLA.

[codspeed-hq]

Merging this PR will not alter performance

✅ 16 untouched benchmarks

---

_{Comparing litellm_/agitated-montalcini (123052e) with main (1757e1d)}

[greptile-apps]

Greptile Summary

This PR fixes WatsonX tests in tests/llm_translation/test_watsonx.py that were failing on CI because the provider's validation requires WATSONX_URL and WATSONX_PROJECT_ID env vars to be present before it reaches the mocked HTTP client. Without them, a validation error was silently swallowed inside the fixture's except block, leaving mock_post.call_count == 0 and causing the == 1 assertions to fail.

The fix adds a single autouse=True pytest fixture (watsonx_env_vars) that:

• Injects dummy WATSONX_URL and WATSONX_PROJECT_ID via monkeypatch (automatically torn down after each test)
• Clears WATSONX_ZENAPIKEY and WATSONX_TOKEN to prevent real credentials from leaking in via .env files loaded by dotenv

The approach is clean: individual tests that need to override these values (e.g., test_watsonx_token_in_env_var re-sets the token vars; test_watsonx_chat_completions_endpoint_space_id deletes WATSONX_PROJECT_ID) do so via their own monkeypatch calls which correctly override the autouse fixture's values within the same test scope. No real network calls are introduced and no test assertions are weakened.

Confidence Score: 5/5

• Safe to merge — adds a minimal, correctly-scoped autouse fixture with no assertion changes, no real network calls, and proper monkeypatch teardown.
• The change is a single, well-targeted autouse fixture. All existing test assertions remain intact. Individual tests that override the fixture's env vars do so correctly via their own monkeypatch calls. No P0 or P1 issues found.
• No files require special attention.

Important Files Changed

Filename	Overview
tests/llm_translation/test_watsonx.py	Adds an autouse pytest fixture that injects dummy WATSONX_URL/WATSONX_PROJECT_ID env vars and clears WATSONX_ZENAPIKEY/WATSONX_TOKEN to allow fully-mocked tests to reach the mocked HTTP client on CI without real credentials.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[pytest test starts] --> B[autouse fixture: watsonx_env_vars]
    B --> B1[monkeypatch.setenv WATSONX_URL]
    B1 --> B2[monkeypatch.setenv WATSONX_PROJECT_ID]
    B2 --> B3[monkeypatch.delenv WATSONX_ZENAPIKEY]
    B3 --> B4[monkeypatch.delenv WATSONX_TOKEN]
    B4 --> C{Test type}
    C -- token env var test --> D[test sets WATSONX_ZENAPIKEY or WATSONX_TOKEN]
    C -- space_id test --> E[test deletes WATSONX_PROJECT_ID, sets WATSONX_SPACE_ID]
    C -- other tests --> F[proceed with WATSONX_URL + WATSONX_PROJECT_ID set]
    D --> G[WatsonX provider passes validation]
    E --> G
    F --> G
    G --> H[mocked HTTP client is reached]
    H --> I[mock_post.call_count == 1 assertion passes]

Loading

_{Reviews (1): Last reviewed commit: "[Fix] WatsonX tests failing on CI due to..." | Re-trigger Greptile}