aurelio-labs/semantic-router is a Python library that compares the meaning of large language model (LLM) and agent requests against routes to choose a model or tool path. It is distributed under the MIT license, and the repository's most recent release is v1.0.2.
What this library does is decide which path to send a request down before calling a model. It is a decision layer that picks a destination based only on the meaning of the request, for services where several models and tools are attached. The README introduces two approaches: comparing the meaning of a request with candidate routes in vector space, and routing by querying a small decision model once. Either way, it provides an adjustable confidence as an actual probability value for each result. Route definitions, local FastEmbed embeddings, how to pass requests through the IndexedRouter, and the procedure for calibration and applying thresholds are all laid out in the README. The repository was created on October 30, 2023 and has gathered 3,960 GitHub stars.
That said, there is no release note in the available material, so it is impossible to know what v1.0.2 changed. There are also no comparative figures for routing accuracy or response latency. The README calling itself 'blazing fast' is self-description, not a measurement. The repository description also mentions multimodal data processing, but what the README covers is semantic routing. The material likewise does not answer whether semantic vector comparison translates into real routing performance. The structure is simple and the license is MIT, so it is worth trying, but the adoption decision is better made after tuning the threshold on a few actual requests yourself.