Export nixosModules.default so the service (systemd unit, service user,
sops secret) is defined in the flake, not re-declared in each host config.
Consume with imports = [ inputs.n8n-openai-adapter.nixosModules.default ]
+ services.n8n-openai-adapter = { enable = true; domain = ...; port = ...; }.
Exposes self-hosted n8n chat agents behind /v1/chat/completions and
/v1/models. Model -> n8n webhook routing via a GenServer registry, so
multiple agents map to multiple models. Plug + Bandit, req for the
n8n webhook call, Bearer auth (ADAPTER_API_KEY). Ships a flake.nix
(beamPackages.mixRelease) so it can be consumed as a NixOS flake input.