STN-01 — AGENTS.md

Purpose

STN-01 is a proof of concept for small-scale, ephemeral live streaming over the open Internet.

BLUF: Treat the live stream as an ephemeral network event, not a file delivered live.

The initial target is one publisher broadcasting directly to approximately five viewers using WebRTC fan-out.

Principles

• FOSS and open standards first.
• No intermediary streaming platforms.
• No persistent media storage.
• No recording unless explicitly added later.
• Prefer direct publisher → viewer connections.
• Spend bandwidth before spending storage or infrastructure.
• Keep the implementation small enough to understand end-to-end.
• Add infrastructure only when a demonstrated constraint requires it.

POC Architecture

• Browser-based publisher and viewer where practical.
• WebRTC for audio/video transport.
• One RTCPeerConnection per viewer.
• Minimal HTTP/WebSocket signaling for SDP and ICE exchange.
• STUN for NAT discovery.
• Self-hosted TURN may exist only as a connectivity fallback.
• Media should not transit the signaling server.

Out of Scope

Do not introduce these unless the experiment explicitly advances to a later phase:

• HLS/DASH
• RTMP distribution
• CDNs
• SFUs/MCUs
• transcoding farms
• adaptive-bitrate ladders
• persistent stream archives
• user accounts
• databases
• analytics or telemetry
• audience-scale optimization

Design Rule

Before adding a component, answer:

1. What demonstrated problem does it solve?
2. Can a standard browser/network primitive solve it first?
3. Does it introduce storage, state, interception, or operational complexity?
4. Can it remain replaceable?

Build the smallest brick that proves live publisher → viewer transport first.