With stateful, provider-independent session lineage and model-name-based routing across way too many heterogeneous LLM providers.
The same route can be shared and revoked over end-to-end HTTPS with another machine, a friend or your phone.
Hell, you can even make an LLM mesh with one Swobu swobing remote Swobus. Anyway.
One <10 MB Go binary. TLS terminates locally - relay is a dumb pipe.
In other words, I’ve tried to factorize LLM traffic along orthogonal client, backend and network dimensions.
Hope I find some product-market fit, because I don’t want to go back to job-market fit.