dsh-polonium

Network dependency is often the first bottleneck in a model-inference pipeline. Intranet deployments, offline experiments, and edge-side reasoning all share one hard requirement: it must work without an internet connection. dsh-polonium is a pure Node.js inference plugin built for exactly that scenario, keeping the entire inference loop inside a local Node process with zero outbound requests and zero remote service dependencies.

Its core strength is complete runtime self-containment: no remote model downloads, no cloud API calls, no extra network policy to carve out. Every inference step executes within the Node.js runtime, so the plugin drops cleanly into CI pipelines, local dev tooling, or embedded gateways without introducing any new network dependency.

Who is it for? If you are building inference integrations in air-gapped or offline environments, want an inference module that carries no additional network dependencies of its own, or need a lightweight way to invoke model capabilities inside the Node.js ecosystem, dsh-polonium offers a clean starting point. The MIT license makes it safe to adopt in both commercial products and internal tooling.