LLM Request Router (LLMhop)

A lightweight proxy that routes requests to multiple local LLM servers (vLLM, llama.cpp, sglang) based on the model field. Solves the problem of managing multiple LLM backends in production environments with authentication and easy deployment via NixOS.

The surge in teams running multiple local LLM backends simultaneously — driven by cost control, data privacy requirements, and model specialization — makes intelligent routing infrastructure genuinely necessary right now, not theoretical. No clear incumbent owns this specific niche, though Litellm handles some overlapping routing concerns and would be the first comparison any prospect makes. The $2k–10k/mo band is realistic only if sold as a managed or hosted layer, because self-hosted devtools at this level of specificity rarely convert to recurring revenue without a strong services wrapper or enterprise contract. The biggest risk is that the target user — someone sophisticated enough to run vLLM and sglang in production — is also sophisticated enough to write this themselves in an afternoon, making willingness-to-pay extremely low.