Sterister
3243647aa1
publish-all.ts now does a tsc → dist/ build per package before pack, then rewrites package.json's main/types/exports to point at the built artefacts and ensures `files: ["dist"]` so the tarball ships only the built code. The in-repo package.json is restored in the finally block so dev/typecheck keep working without a build pass. Why: strict-mode consumers (Cyndr) were forced to recompile Shade source under their own tsconfig and tripped on internal `process.env.X` accesses and implicit-any parameters. Shipping pre-built `.js` + `.d.ts` makes the strictness contract live entirely inside Shade.
@shade/transport-bridge
Transport-agnostic delivery for Shade: WS → SSE → long-poll, in priority
order, behind a single IncomingMessage interface.
import {
FallbackBridgeTransport,
WsBridge,
SseBridge,
LongPollBridge,
} from '@shade/transport-bridge';
const auth = { crypto, signingPrivateKey, address: 'bob' };
const bridge = new FallbackBridgeTransport([
new WsBridge({ baseUrl, auth }),
new SseBridge({ baseUrl, auth }),
new LongPollBridge({ baseUrl, auth }),
]);
await bridge.connect({
onMessage: (msg) => {
// msg: { from: string; bytes: Uint8Array; receivedAt: number; msgId?: string }
},
});
console.log(bridge.activeKind); // "ws" | "sse" | "long-poll"
Pair with createBridgeRoutes in @shade/inbox-server to expose the
matching /v1/bridge/{stream,poll,ws} endpoints. Full design + threat
model in docs/transport.md.
What it solves
Browser extensions, strict corporate proxies, and edge runtimes routinely block long-lived WebSockets. Apps that already use the Shade inbox shouldn't have to write three custom delivery paths to handle the realistic mix of hostile networks they ship into. This package is the canonical answer.
Status
V3.7. Stable wire format, additive change to @shade/inbox-server. See
CHANGELOG.