LangChain introduced AI Search in LangSmith, letting users describe a behavior in plain language — such as whether the user got frustrated or PII leaked — and get back the sessions where it happened. Per LangChain, structured filters narrow the candidates first, then a fast decision model reads each normalized "trajectory" and streams matches as they are found; SmithDB, the vendor's database built for trace data, builds those trajectories on arrival and normalizes formats from the OpenAI and Anthropic SDKs, LangChain, LangGraph, the Vercel AI SDK and OpenTelemetry, while removing repeated conversation history to cut context bloat. AI Search is in public beta on LangSmith Cloud in the GCP US, EU and APAC regions at no additional cost, and org admins must enable it in settings.