@twinlight/client

High-level TwinLight SDK for browser, Electron, Tauri, and custom UI hosts. It composes control-message builders, client-side batching, viewport ownership, asset registration, and capability gating around a native TwinLight renderer.

60-second browser attach

import { createTwinLightClient } from "@twinlight/client";

const client = await createTwinLightClient({
  endpoints: {
    control: "ws://127.0.0.1:61721",
    video: "ws://127.0.0.1:61720",
    signaling: "ws://127.0.0.1:61722",
  },
  connect: false,
});

client.render.setQuality("max_quality");
client.render.setResolution({ width: 1920, height: 1080 });
client.render.patchFeatures({
  shadow: "rt",
  ambientOcclusion: "rtao",
  reflection: "rt",
  gi: "rtgi",
  resolve: "dlss_sr",
  dlssQuality: "quality",
});
client.scene.load("raster_showcase");

Professional integration notes

• patchFeatures() always sends a complete render.set_features wire message. The backend does not accept a partial feature-delta object.
• client.batch() flushes individual control frames in one synchronous burst. It does not send scene.batch because the current native backend does not parse atomic batch envelopes.
• Capability fields reflect the current renderer capability schema. Unprobed effects such as volumes, clouds, tessellation, Gaussian splats, and point clouds are reported as source: "unknown" unless a future backend schema exposes them.
• Texture URI registration is JSON-only. Binary texture upload requires a binary-capable control socket and uses the native chunked upload protocol.

Not supported

• No CPU or JavaScript renderer fallback.
• No silent software encoder fallback.
• No fake WebGLRenderer compatibility.
• No capability claims for fields the native backend does not currently emit.