AiSanity: Monitors your AI for hallucinations. QA layer for AI apps

Hey Product Hunt! 👋 I’m Phuripat, a 19-year-old developer and student. A few months ago, I woke up to angry dms from users of my previous AI project. They said the app was broken. I checked my uptime monitor it was all normal. Status: 200 OK. I was confused. Immediately dive into the Vercel's logs and realized that the API was responding, but the GPT model had drifted. Instead of valid JSON, it was replying with "I'm sorry, as an AI language model..." My frontend crashed, but my uptime monitor didn't care. That’s when I realized: For AI apps, "200 OK" is a lie. I looked for tools to monitor "AI Quality," but everything was Enterprise-grade (Datadog/LangSmith) requiring heavy SDKs and expensive plans. I just wanted a simple ping that checks: "Is the AI actually smart right now?" So I built AiSanity. It’s an active QA for your AI. It sends a test prompt every few hours (saving your tokens!) and verifies the answer using a LLM-as-a-Judge model. - Checks for broken JSON schemas. - Catches hallucinations/drift. - Alerts you before your users do. I built this for Indie Hackers and Vibe Coders who need good enough QA without the enterprise bloat. I’d love to hear what you think! Does this solve a pain point for you?