blog
Blog
Practical writing on LLM regression testing: how to build evals, gate pull requests on them, and keep quality from drifting.
- Anthropic SDK: "temperature: Extra inputs are not permitted"Why Anthropic SDK 1.x requests fail with temperature: Extra inputs are not permitted, how a live eval run caught it, and why we stayed on SDK 0.x.
- How to run LLM evals in GitHub ActionsHow to run LLM evals in GitHub Actions on every pull request: scaffold a suite, commit baselines, add the workflow, and fail the check on a regression.
- LLM regression testing: a practical guideA practical guide to LLM regression testing: golden datasets, evaluators, baselines and thresholds, CI gating, flaky evals, and tracking quality over time.
- LLM evaluation tools compared for CI (2026)LLM evaluation tools compared: promptfoo, DeepEval, Ragas, LangSmith, Braintrust, Langfuse and Drift-CI on licensing, CI gating, baselines and hosting.