ukeeper / ukeeper-readability / 33687908792

72%

master: 78%

Pull #32

github

fix: share one OpenAI key between evaluation and summaries

Stacking the two branches left two key options whose env vars collided: the
top-level OPENAI_API_KEY and the group's API_KEY under env-namespace OPENAI
resolve to the same name, so setting it for summaries also switched on paid
auto-evaluation. Drop the group's key, use the single top-level one for both
features, and give evaluation its own opt-out to match disable-summaries.

Also cover GET /api/metrics, which had no test, and correct CLAUDE.md on the
summaries store: its tests need ENABLE_MONGO_TESTS and a local mongo rather
than testcontainers, and the Summaries and OpenAIClient interfaces and their
mocks were missing from the architecture notes.

215 of 433 new or added lines in 7 files covered. (49.65%)

1163 of 1622 relevant lines covered (71.7%)

8.33 hits per line