Skip to content

Releases

RelaySwarm v0.4.4 (experimental)

Pre-release

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 23:05
fdebc64

Experimental peer-assisted HLS release.

Source: fdebc648, with a tree identical to reviewed head 85bbefa. All 14 exact-head PR checks passed: e2e and realtime transport.

  • Already-advertised framing-2 schema-3 holders can contribute bounded verified pieces even when a slow whole-object estimate would reject the entire segment.
  • Waiting for not-yet-held recorded main media now requires a live same-key FETCH lease. A contributor behind the player's VOD frontier no longer causes a speculative wait merely by serving or ranking ahead. Live, unknown and alternate-audio contexts retain their bounded coordination for synchronous release.
  • Sustainable deadlines, measured origin reserves, the 300 ms estimator floor and integrity verification remain intact. Legacy commitments retain whole-object projection; wire formats and committed WASM bytes are unchanged.

Twenty focused real-WASM regressions and the final full WASM loader sweep (311 passed, one expected JavaScript-prototype-only skip) passed, alongside syntax and documentation checks. The initial candidate also passed 370 serial JavaScript loader checks. Its parallel real-clock failures and initial live-broadcast policy failure are recorded in status evidence; the final scoped candidate restores live coordination.

The unchanged real-time broadcast browser fixture passed on an M4: 3,419,156 verified peer bytes over six consumer loads, peer playback at both renditions, and stable-identity producer restart. The consumer and origin-only baseline reported zero stalls; all four viewers reported zero fatal errors. This is a loopback broadcast fixture, separate from physical-device playback.

With separate website clock/first-frame fixes, the published physical Pixel/Vanadium + M4 original-4K fixture passed a bounded 51.889-second normal-cache window: 6,979,584 verified peer bytes received, 8,372,224 uploaded media bytes recorded, both 2160p, and zero reported stalls or dropped frames. This used owner-supplied expiring TLS relay profiles. Earlier longer trials retained clock-seek/recording-loop interruptions. A published cold-follower control received 245,760 peer bytes and recorded 1.64 seconds of waiting over 101 seconds with an established TLS link. Sustained continuity and useful cold-network contribution remain open.

Download checksums and source metadata accompany the JavaScript/WASM assets. The npm registry remains unpublished. See installation instructions.

Apple/Android packaging, verification and attachment passed. All 13 assets were anonymously downloaded and verified against their checksum manifests, exact-source metadata and SwiftPM checksum. The extra tag-time matrix had one existing short real-clock macOS loader failure on attempt 1. Its single bounded macOS retry and dependent live-relay check passed: tag workflow attempt 2 is green with 16 successful jobs and one optional soak skipped. The tag and assets remained immutable.

v0.4.3 — experimental sustainable peer delivery

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 21:13
6f63fb2

Experimental RelaySwarm 0.4.3 adds opt-in sustainable peer waits and range-sized WASM upload inputs.

  • Positive waits and extensions fit the fragment/part duration after measured HTTP recovery; the cached-peer race still starts HTTP immediately.
  • A 6 MiB object in 16 bounded requests copies 6 MiB into WASM instead of 96 MiB, with identical verified output. This is input-copy accounting, not a device throughput claim.
  • The optional range-upload ABI preserves the wire protocol, range validation, cancellation and older-core compatibility.

Validation includes 42 actual-WASM/protocol checks, 13 focused Rust range tests, 291 forced-core loader checks (one expected skip), Rust format/Clippy/workspace tests and hosted checks. Physical iPhone/iPad, Pixel/Vanadium peer continuity, varied NAT and sustained distributed net benefit remain open. No service is deployed by this release.

Assets contain the JavaScript package, standalone browser module, paired WASM, build metadata and SHA256SUMS. The npm registry is unpublished. Native packages are attached by their separate release workflow when those checks pass.

v0.4.2 — experimental admission renewal

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 14:04
2c2708c

Experimental library prerelease from 2c2708c9e11daf8e85bc059682d3eda09a0c4ffb, following immutable v0.4.1. Includes PR #25 and PR #26.

Uploads that arrive during a peer's changed-ticket admission check now join the existing bounded online decision before spending allowance or sending media. Previously a check using the remembered old ticket could close the live connection. Rejection, timeout, stop and channel closure still prevent delivery; subsequent uploads check online afresh.

peerConnectionDetails() exposes accepted peer identities and connection references so hosts can distinguish retention from replacement. Local native real-time build output is excluded from the JavaScript tarball.

The native Chrome/WASM renewal fixture retains the exact accepted connection, identities, loader, cached media, upload consent and cumulative spend through three ticket rotations on each side. It verifies four 256 KiB peer loads; later online revocation sends zero peer bytes and falls back to HTTP. Its receipt is retained in CI. These are synthetic-byte loopback protocol checks.

Downloads include the installable JavaScript package, standalone browser bundle and matching WASM, Apple XCFramework/Swift bindings and Android AAR, with three checksum manifests and metadata identifying this tagged source. The npm registry remains unpublished. RelaySwarmKit v0.6.0 retains its published v0.3.0 core pin; these assets do not upgrade Kit or add a native WebRTC application.

The retained R70 distributed qualifier remains RED; its far-end neighbour replacement diagnosis is distinct from this admission overlap fix. Distributed retention, TURN renewal, physical-device/public-NAT operation, sustained playback and net savings remain open. No public service or pilot is deployed by this release.

Validation: all 14 pre-merge PR checks passed on the same source tree. Both merged-source native real-time jobs passed. Fifteen tag jobs passed, including native archive build, independent verification and attachment. The tag browser job remains RED after two attempts at the existing cached-race assertion normal HTTP starts before asynchronous segment-key hashing; subsequent renewal/broadcast steps in that job were skipped. A local diagnostic with unchanged assertions passed, but the cause of the hosted failures is not established. This is not an all-green tag run. See the tag workflow and the passing PR workflow.

All 13 public assets were anonymously downloaded and checked against GitHub digests and all three checksum manifests; build metadata identifies this tagged commit. Fresh tarball installation passed 39 protocol test cases, including 140 shared vectors, and CLI help. The published standalone browser bundle and WASM passed the exact-connection renewal/revocation fixture on an unchanged rerun; its first local attempt fell back to HTTP on one expected peer load. Original failures are retained, and no assets were replaced. Physical-device, public-NAT and distributed acceptance remain open.

v0.4.1 — experimental cached-peer race

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 12:57
1dc2ae6

Experimental library prerelease from 1dc2ae664de46e358e463a1561541ea339d7dc6e, following immutable v0.4.0. Includes PR #22 and the release/CI follow-through in PR #24.

When no safe peer wait remains, opt-in zeroBudgetPeerRace: true starts normal whole-segment HTTP immediately and races only an already advertised cached schema-3 object. Only a complete verified winner reaches the player. Stopping the swarm leaves pending HTTP running; loader cancellation stops both arms. The default stays false. Progressive media bypasses the race, and nonempty HTTP progress locks delivery to HTTP.

Multi-holder origin tails retain their measured per-load recovery reserve and absolute deadline. Bounded, URL-free counters distinguish range failures, timeouts, aborts and wrong-sized bodies. Accepted-contributor holder recovery is shared by ordinary peer-first loads and cached races; unsuccessful or origin-only work receives no peer credit.

Downloads include an installable package tarball, standalone browser bundle, matching WASM, Apple core XCFramework/Swift bindings and Android AAR, with three checksum manifests and build metadata carrying this tagged source. The native Chrome cached-race receipt is now retained in CI and proves both winners, one whole HTTP request per load and reuse of the same connection with synthetic loopback data.

This is not field qualification. Physical devices, public NAT, sustained quality and net hosting benefit remain open. Accepted peer bytes alone do not prove net savings because a losing HTTP request may already have transferred bytes. No public service or pilot was deployed. The npm registry remains unpublished. RelaySwarmKit v0.6.0 retains its published v0.3.0 core pin; these standalone native core assets do not upgrade Kit or add a native WebRTC application.

Validation: all 14 PR checks, 14 tag-pipeline jobs and both merge-source realtime jobs passed. The final hosted relay check and attachment job stayed unassigned for over twenty minutes, so their queued remainder was cancelled. The same audit and public-relay journeys passed locally (zero vulnerabilities; both hash-verified journeys), and the CI-built, CI-verified native archives were uploaded without replacing assets. The tag workflow records that cancelled remainder; it is not an all-green tag run.

v0.4.0 — experimental library integration

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 11:20
3cafb5b

RelaySwarm v0.4.0 integrates holder recovery with the current Rust/WASM core and the verified-delivery, range-cancellation and relay-startup fixes from the pushed 3abac38 snapshot of PR #22. This is an experimental library release. It deploys no public service or pilot.

  • Whole-object holders are charged only for judged failures. Verified delivery clears holder history, a spent replacement can re-arm after twenty seconds, and an untested holder gets one bounded trial. Eligible rotation victims are ordered by contribution tier. Rust, WASM and native discovery/selection carry the matching state and diagnostics.
  • peerConnections() returns accepted active native connections. onPeerDelivery attributes only first-verified peer bytes in an accepted object. Control traffic, corrupt/duplicate/incomplete media and all-origin assembly receive no peer credit. All-origin assembly does not become a peer timing/ABR sample or renew a peer replacement.
  • The piece scheduler preserves verified progress when origin has reserved the tail. Reliable channels can drain a cancelled response and carry the next request. Reconnecting pools start on their first healthy relay and replay current presence to later relays while retaining bounded socket/retry ownership.
  • Includes the experimental native real-time source integration delivered after v0.3.0, with guarded backend preparation and bounded TURN teardown. The native core archives are transport-independent; they do not provide the separate WebRTC transport.

Validation includes 315 JavaScript tests and 254 core-backed browser tests with one deliberate skip; Rust regressions and the shared replay; 140 verifier vectors, 48 link traces, six selection sequences and twelve discovery sequences through generated Swift bindings, with Kotlin checked in CI. Actual Chrome reliable channels survived cancellation of a 6 MiB ranged response, the drain deadline and a subsequent request. CI retains that receipt and covers media playback/recovery, broadcasting, Swift/browser interoperability, platform packaging and native realtime transport. These are local/CI results.

Download SHA256SUMS for the JavaScript tarball, standalone bundle, WASM module and JavaScript metadata. Native downloads have separate SHA256SUMS-apple and SHA256SUMS-android manifests. All three metadata files identify the tagged source. Use JavaScript and WASM files from the same release, and the matching native binary and generated wrapper together. The npm registry remains unpublished. Existing v0.3.0 assets and tags are unchanged.

Package version 0.4.0, session wire version 3, metrics version 2, commitment schemas 1/2/3 and object framing 1/2 are distinct. Kit v0.6.0 remains checksum-pinned to its published v0.3.0 core; the standalone v0.4.0 native downloads contain their matching updated bindings.

Historical R70 at 20693ea passed 15 of 16 gates: in-place renewal failed and high-quality peer delivery margin was thin. This integrated source has not rerun that distributed qualifier. Physical devices, public NAT, sustained quality and net hosting benefit remain unqualified. Applications choose media, relays, admission, trust and consent. Newer work on PR #22 after 3abac38 is excluded from this release snapshot and remains in its original branch/worktree.

Integration: #10
Included source snapshot: 3abac38

PR checks: https://github.com/forgesworn/relayswarm/actions/runs/40086279758 and https://github.com/forgesworn/relayswarm/actions/runs/40086279670

v0.3.0 — experimental shared core

Pre-release

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 24 Oct 07:59

Experimental RelaySwarm v0.3.0: the opt-in Rust/WASM core, schema 3 piece commitments and multi-holder transfer path, UniFFI Swift/Kotlin facade, advisory peer supply hints and rendition-switch reporting. Built from a3b45cac19dcfbf4e4794b971ba6978852754d2c after PRs #14–#20 merged. This remains a proof of concept; field acceptance is open.

JavaScript downloads include an installable relayswarm-0.3.0.tgz, a standalone relayswarm-hls-swarm-0.3.0.mjs, the optional relayswarm-core-0.3.0.wasm, build metadata and SHA256SUMS. The host supplies hls.js. The npm registry remains unpublished.

Native downloads include the arm64 Apple XCFramework, its actual SwiftPM checksum, generated Swift bindings with the unchanged local-path package, and an Android AAR for arm64-v8a, armeabi-v7a and x86_64. Platform checksum manifests and build metadata bind these assets to the tagged source. See the packaging guide for minimum versions and dependencies. RelaySwarmKit does not yet use this core.

Install the JavaScript package:

npm install https://github.com/forgesworn/relayswarm/releases/download/v0.3.0/relayswarm-0.3.0.tgz

See Getting started, Integration, the changelog and dated evidence.

The core, multi-holder path and advisory hint require explicit opt-in. The hint reports known HAVE supply and never forces player quality. Scheduler defaults remain 32 KiB/s and at most three holders. Signature verification remains in the core following the measured comparison. The original WebRTC CLI still signs plaintext SDP; browser-loader signalling and native dial authorisation use NIP-44.

Loopback, CI, simulator/emulator and package checks do not establish physical-device, public-NAT, sustained field, battery or net hosting-cost acceptance. This release does not deploy or qualify the separate controlled pilot.

Validation: all 12 required PR checks passed on the release commit. Local macOS arm64 checks passed 234 JavaScript loader tests and 174 core-backed tests (one deliberate skip), all 140 verifier vectors through a fresh 361-member tarball install, native CLI help and Chrome 155 standalone-bundle/WASM hint replay. All 16 tag workflow jobs passed, including native builds, combined verification and release attachment; each platform's metadata identifies the exact tagged source.

RelaySwarm v0.2.1: experimental holder-discovery prerelease

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 21 Oct 14:46
1d54538

Experimental RelaySwarm prerelease, built from 1d54538c56179387b7541924e068d0d9c41d18e6 (PR #9). It stops a linked holder that is never ahead from suppressing neighbour discovery, and adds session-local holder labels to the loader's local peer diagnostics. Labels are counters, not keys; they are never sent on the wire and are not in the opted-in numeric report. Wire version 3, commitment schemas 1 and 2 and metrics version 2 are unchanged. The full list is in the changelog.

Downloads:

  • relayswarm-0.2.1.tgz: installable package containing the modules, demo, broadcaster, documentation and recorded fixtures.
  • relayswarm-hls-swarm-0.2.1.mjs: standalone browser ESM bundle with its runtime dependencies included. The host supplies hls.js.
  • SHA256SUMS: checksums for both assets.

Install the library:

npm install https://github.com/forgesworn/relayswarm/releases/download/v0.2.1/relayswarm-0.2.1.tgz

Run the demo from source (Node 22.13+ and ffmpeg for synthetic media):

git clone --branch v0.2.1 --depth 1 https://github.com/forgesworn/relayswarm.git
cd relayswarm
npm ci
npm run demo

See GETTING-STARTED.md, BROADCASTING.md and INTEGRATION.md.

Validation: all nine required CI jobs passed on this commit and on its PR head. A fresh package consumption of this commit passed with 287 package members, and its built demo client is byte-identical to the client served by the pilot. On the same loader source, deployed to the public pilot, the fourteen-role short qualifier (R66) passed all 16 gates for the first time, including both-quality peer media for every receiver, with relay-only TURN over TLS/TCP. Desktop installed-browser journeys passed for Firefox 152.0.6 and Safari 26.6.2 on macOS arm64; this is desktop scope only, not physical devices, hard NAT or a separate network. At publication the pilot runs 140c256, whose runtime differs from this release only in the package version fields.

This is a prerelease and field acceptance remains open, by owner decision. The sustained two-hour run with churn and natural credential expiry is RED (its latest attempt stopped on a harness race after 640 seconds of measurement). The 3 MB both-quality shared-allowance lifecycle is RED: high-quality transfers over the relay path did not complete before the documented reservation charging used up the 3 MB cap. Desktop Safari 17.6 is RED. Physical devices, difficult networks and hard NAT are not established, and the marginal hosting cost is not determined. The npm registry remains unpublished; this release supplies GitHub assets and a pinned source tag.

Current documentation, including corrections after this release, starts at the documentation index, installation guide and release evidence page. The release tag and assets retain their original contents.

RelaySwarm v0.2.0: experimental broadcaster pilot prerelease

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 20 Oct 23:37
bf43940

Experimental RelaySwarm prerelease, built from bf4394031809abb37b6a534d6be86aae3a3c556b (PR #8). It adds the supervised broadcaster and production feature work to v0.1.0: a live H.264/AAC broadcast bridge with two HLS renditions, source recovery and optional RTMPS forwarding; private persistent publisher identity; cumulative viewer upload allowances with an optional shared-account ledger; authenticated operator status and policy, opted-in numeric reporting, bounded admission and crash supervision; session wire version 3 with signed media commitments; and bounded peer waits, rendition-neighbour discovery and clock-skew diagnostics in the loader. The full list is in the changelog.

Downloads:

  • relayswarm-0.2.0.tgz: installable package containing the modules, demo, broadcaster, documentation and recorded fixtures.
  • relayswarm-hls-swarm-0.2.0.mjs: standalone browser ESM bundle with its runtime dependencies included. The host supplies hls.js.
  • SHA256SUMS: checksums for both assets.

Install the library:

npm install https://github.com/forgesworn/relayswarm/releases/download/v0.2.0/relayswarm-0.2.0.tgz

Run the demo from source (Node 22.13+ and ffmpeg for synthetic media):

git clone --branch v0.2.0 --depth 1 https://github.com/forgesworn/relayswarm.git
cd relayswarm
npm ci
npm run demo

See GETTING-STARTED.md, BROADCASTING.md and INTEGRATION.md.

Validation: all nine required CI jobs passed on this commit. The public pilot runs 5f2ece0, whose nine required CI jobs, fresh package consumption, guarded deployment, reporting lifecycle (P05) and parent-crash recovery (P07) passed. This release adds the later loader clock-skew diagnostics, which are not deployed. RelaySwarmKit v0.5.0 interoperability is measured on macOS arm64 loopback.

This is a prerelease and field acceptance remains open. The fourteen-role field qualifier is RED: its latest run (R65) completed the 300-second measurement and passed 14 of 16 gates, but only one of four receivers obtained high-quality peer media and one contributor uploaded nothing during measurement. Sustained two-hour and churn operation, natural credential expiry, physical devices and Safari, difficult networks and hard NAT, measured wire and cost benefit, and downloaded-release acceptance are not established. The npm registry remains unpublished; this release supplies GitHub assets and a pinned source tag.

Current documentation, including corrections after this release, starts at the documentation index, installation guide and release evidence page. The release tag and assets retain their original contents.

RelaySwarm v0.1.0 — experimental HLS field-test release

Choose a tag to compare

@TheCryptoDonkey TheCryptoDonkey released this 12 Oct 10:19
3e480e1

First experimental RelaySwarm GitHub release, built from 3e480e19dfe4d675f14bcb2968f3dc4926fbca25 (PR #6). It includes browser shadow and bounded peer-first HLS delivery, a Node publisher with signed playlist-bound segment commitments, the native HyperDHT proof of concept, and a publisher/player demo with explicit upload consent, limits, an origin-only kill switch and local receipts.

Downloads:

  • relayswarm-0.1.0.tgz: installable package containing the modules, demo, documentation and recorded fixtures.
  • relayswarm-hls-swarm-0.1.0.mjs: standalone browser ESM bundle with its runtime dependencies included. The host supplies hls.js.
  • SHA256SUMS: checksums for both assets.

Install the library:

npm install https://github.com/forgesworn/relayswarm/releases/download/v0.1.0/relayswarm-0.1.0.tgz

Run the demo from source (Node 22+ and ffmpeg for synthetic media):

git clone --branch v0.1.0 --depth 1 https://github.com/forgesworn/relayswarm.git
cd relayswarm
npm ci
npm run demo

Open http://localhost:8787. See FIELD-TEST.md for existing media, HTTPS access, consent and receipt scope, and INTEGRATION.md for the host-page contract.

Validation: all seven main CI jobs passed. The twelve-viewer loopback soak recorded zero playback stalls or fatal errors, 68.7% of fragment loads supplied by peers, corruption rejection and bounded sampled resources. Two viewers used origin throughout; these are media-load results, not net network-data savings. The packaged public modules were also checked in a clean production consumer, and the browser asset was smoke-tested in Chrome.

This is a proof of concept and a prerelease. Physical phones, separate-network/NAT trials, sustained operation, fMP4, broad adaptive bitrate behaviour and independent-client interoperability remain unverified. The npm registry remains unpublished; this release supplies GitHub assets and a pinned source tag.

Current documentation, including corrections after this release, starts at the documentation index, installation guide and release evidence page. AI index and full Markdown context are maintained on main. The release tag and assets retain their original contents.