Show HN: Alignmenter – Measure brand voice and consistency across model versions (www.alignmenter.com)

Alignmenter is an open-source tool for automated testing of AI chatbots that measures brand voice, safety, and behavioral consistency across model versions. It lets teams run the same dataset against GPT-4, Claude, custom GPTs or local models, compare results side-by-side, and produce repeatable, shareable HTML reports. The goal is to catch tone/personality regressions, safety regressions, and intra-session inconsistency that standard correctness tests miss—so you can validate releases, integrate checks into CI, and audit changes before users notice them. Technically, Alignmenter combines metric-based scoring with optional LLM judges and keyword filters. Brand scoring blends style_sim (0.6), traits (0.25) and lexicon (0.15); safety uses min(1 − violation_rate, judge_score); consistency is 1 − normalized_variance(embeddings). It runs entirely locally (data never uploaded), supports budget limits for costly AI reviewers, exports audit trails, and syncs Custom GPT instructions. Because it’s open source and free for commercial use, teams can extend scoring, add providers, and embed it in pipelines for regression detection, compliance reporting, and research into model alignment and persona drift.