about the project

Open numbers. Open runtime next.

Voice OSS makes the economics of AI calling inspectable today while documenting the open frameworks and infrastructure paths that can make the runtime more ownable tomorrow.

Built by Himanshu Maurya

Himanshu Maurya, also known as codingstark, creates OSS.codes and builds practical open-source tools for developers.

Voice OSS is the public research and economics layer for his work on open Voice AI. It combines source-linked pricing, reproducible calculators, framework discovery, and honest ownership trade-offs rather than presenting one vendor's rate as the whole cost.

Himanshu is also working toward an open-source Voice AI calling stack designed around Cloudflare's edge platform. The direction explores realtime media, Workers, Durable Objects, WebSockets, and open agent orchestration; it is active build work, not a claim that the full calling platform is already released.

Follow codingstark on GitHub →

Why an edge-native call stack is interesting

Voice is a realtime distributed-systems problem: media, model streams, call state, events, recordings, tools, and failures all need coordination.

mediaCloudflare RealtimeWebRTC and SFU building blocks
computeWorkersedge request and tool execution
stateDurable Objectscall coordination and WebSockets
principleopen adaptersmodels and carriers stay replaceable

Inspect the primitives

These links describe the Cloudflare capabilities informing the architecture direction.