livekit_ffi 0.13.2 
livekit_ffi: ^0.13.2 copied to clipboard
livekit.ioDart bindings for the LiveKit Rust SDK core, delivered as a prebuilt native library via Native Assets.
Changelog #
All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
[Unreleased] #
0.12.47 - 2026-02-10 #
0.12.45 - 2026-02-09 #
Fixed #
- fix the RPC race condition (#865)
Other #
- update proto & fix CI (#871)
- add can_manage_agent_session permission (#870)
- Use workspace dependencies & settings (#856)
- Upgrade protocol to v1.44.0 (#857)
- Use dedicated audio_runtime with high priority for audio capture (#854)
- Expose participant's permission to ffi layer (#824)
- Add a request_async_id to the async requests (#842)
- Use the correct download url in webrtc-sys build. (#825)
0.12.44 - 2026-01-15 #
0.12.41 - 2025-12-04 #
0.12.40 - 2025-11-20 #
0.12.39 - 2025-10-27 #
0.12.37 - 2025-10-22 #
0.12.36 - 2025-10-13 #
0.12.33 - 2025-09-09 #
0.12.31 - 2025-07-31 #
0.12.27 - 2025-06-17 #
0.12.16 - 2025-03-06 #
Fixed #
- Fixed metric report issue on audio filter where room_id is sometimes empty
0.12.15 - 2025-03-05 #
Fixed #
- Ensure RTC session continues even if audio filter initialization fails
0.12.7 - 2025-01-17 #
Added #
- Add DataStream.Trailer support
- Add an API to set buffer_amount_low_threshold for DataChannel
- Update RoomInfo to contain buffer_amount_low_threshold for DataChannel
0.12.6 - 2025-01-07 #
Fixed #
- Automatically close audio/video stream handles when the associated room is closed
0.13.2 - 2026-10-23 #
Fixes #
- fix: bump libwebrtc to webrtc-a545de4 to drop roots that trust the compromised eMudhra intermediates. The new build refreshes the TURN/TLS anchors compiled into
rtc_base/ssl_roots.h, which no longer includeAAA Certificate Services, the root that cross-signs the eight eMudhra intermediate CAs whose keys were disclosed on 2026-10-20. - Add a
kindtoCaptureErrorso clients can tell whether a capture failed in its source, was rejected by the RTC source, or ended because the pump panicked. - Fix the VAAPI H.264 encoder leaking VA parameter buffers on every frame; they are now destroyed after vaEndPicture().
- Publish the Swift package and the
LiveKitFFIpodspec with URLs of the renamedlivekit/livekit-ffi-xcframeworkhosting repo. The podspec summary now says LiveKit FFI instead of LiveKit UniFFI. - Stop the NVIDIA and Jetson encoders from reporting an already delivered frame as an encode error, which aborted the process in
SimulcastEncoderAdapterwhen dynacast paused a simulcast layer or a track was unpublished mid-encode.
0.13.1 - 2026-10-20 #
Fixes #
- Update
rustls-webpkito 0.103.16 andwebpki-roots/webpki-root-certsto 1.0.10 in prebuilt FFI binaries, which stop trusting the compromised eMudhra intermediate CAs.
0.13.0 - 2026-10-20 #
Breaking Changes #
Merge livekit-uniffi into livekit-ffi.
There is now one crate, one version and one release tag. The livekit-uniffi
crate, its changelog and its livekit-uniffi/v* tags are gone; its history is
preserved in git.
The crate carries two mutually exclusive surfaces, selected by feature:
room-apis(default) is the existing protobuf / C-ABI surface, withlivekitand libwebrtc. Everything that builtlivekit-ffibefore builds it unchanged.core-modulesis the formerlivekit-uniffisurface — access tokens, log forwarding, data tracks, data streams v2 — and depends on none oflivekit,libwebrtc,soxr-sysorimgproc. What previously builtlivekit-uniffinow builds-p livekit-ffi --no-default-features --features core-modules.
[package.metadata.platform-features] in livekit-ffi/Cargo.toml records
which surface each downstream package build gets, and CI resolves its build
flags from it.
The merge itself requires no downstream SDK changes: the UniFFI namespace is
pinned per surface and every published package name is preserved. The one
difference is the native library the crate builds, liblivekit_uniffi →
liblivekit_ffi, which is loaded by name from the generated bindings and the
Dart build hook and needs no consumer change. The package renames are a
separate change, released alongside this one.
The UniFFI node package (@livekit/uniffi) is not built for now and will be
reintroduced separately; node bindings ship from livekit-ffi-node-bindings.
Features #
Rename the published UniFFI packages to an FFI identity.
Following the livekit-uniffi merge, the packages built from
livekit-ffi/core-modules are named after the crate that now produces them. No
compatibility aliases are published.
| was | now | |
|---|---|---|
| Swift package / product | LiveKitUniFFI |
LiveKitFFI |
| Swift FFI module | RustLiveKitUniFFI |
RustLiveKitFFI |
| Maven artifact | io.livekit:livekit-uniffi-android |
io.livekit:livekit-ffi-android |
| Kotlin package | io.livekit.uniffi |
io.livekit.ffi |
| pub.dev package | livekit_uniffi |
livekit_ffi |
| UniFFI namespace | livekit_uniffi |
livekit_ffi |
Breaking for the Swift, Android and Flutter SDKs. Imports, dependency coordinates and the generated module names all change together.
The Swift xcframework continues to be hosted at
livekit/livekit-uniffi-xcframework; only the package and product names inside
it change. Versions continue from the merged crate, so no published version
goes backwards within a name.
Fixes #
- Outgoing data stream chunk and trailer packets now carry the stream's destination identities. Before, only the header was targeted, so the content of a stream sent to specific participants was delivered to everyone in the room.
- Stop the Node bindings'
FfiHandlefinalizer from throwing when garbage collection runs afterlivekitDispose(), and treat handles already released by shutdown as disposed. - Synthesize index-based identifiers on platforms without hardware audio GUIDs and add index-based selection to PlatformAudio.
- Prefix every ffmpeg symbol bundled in the linux
libwebrtc.awithLK_, so an application that also links its own FFmpeg, shared or static, no longer calls libwebrtc's copy or fails to link. - Embed per-slice
.dSYMbundles in the SwiftRustLiveKitFFI.xcframeworkagain so Xcode archives and crash symbolication can resolve Rust frames (livekit/client-sdk-swift#1149). - Update
zipto 9.0.4, as older releases have security issues and were temporarily yanked (zip-rs/zip2#1075).
livekit-protocol exposes the conversation schema of a thread's database, as protocol's Go package
does: agent::THREAD_SCHEMA_VERSION and agent::THREAD_SCHEMA_V1 (the SQL); the crates that
depend on it follow its bump.
Stop failing an RPC call whose reply arrives just after its recipient leaves the room.
Purging a call's pending ack and response the instant its recipient disconnects assumed that
"participant left" implies "no reply is coming". It does not: the disconnect is announced on the
signalling connection while the ack and response travel on the data channel, and nothing orders
the two. A recipient that leaves because of the RPC it just answered — lk.sip.EndCall, where
the SIP participant drops out of the room as soon as it has ended the call — produces that order
routinely, so the reply was discarded ("Response received for unexpected RPC request") and the
caller was handed RecipientDisconnected for a call that in fact succeeded.
A disconnect now shortens the wait instead of ending it: the pending entries stay registered for up to 2 more seconds, so a reply already in flight still resolves the caller, and a recipient that really did leave without replying still fails fast rather than running out the 15s response timeout.
0.12.83 - 2026-10-12 #
Features #
- Add a capture source that ingests encoded video from a GStreamer pipeline.
Fixes #
- In-flight incoming data streams now error with
AbnormalEndwhen the room closes. Before, the stream readers ended cleanly, soread_all()returned truncated content as if it were complete. - Keep pre-encoded video decodable when WebRTC drops frames: the passthrough encoder disables WebRTC's frame dropper, and after a frame is dropped before it, holds delta frames and requests a keyframe until the next keyframe arrives.
- Add
PeerConnection::set_bitrateto set the send bitrate limits of a peer connection. - Public API tokens minted from an API key and secret now grant every thread action (list, create, read, write, delete) across the project. Before, they carried only
all, so creating or reading a thread was refused. - Re-apply dynacast publishing layers after every publisher answer, so that layers paused by dynacast stay paused after a renegotiation.
Room::closenow leaves the roomDisconnectedand emitsRoomEvent::Disconnectedbefore it returns, andDisconnectedis final: a reconnect still finishing in the background (a queued resume, or the republish after a full reconnect) can no longer bring a closed room back toConnectedor emitReconnectedafterDisconnected. Before, the room's event task could drop the engine'sDisconnectedevent when the close signal won the race, so a closed room kept reportingConnected(orReconnecting) about half the time.
VideoGrants gains the agent grant and Claims the kind claim (with
AccessToken::with_kind), which the Go, Python and JS SDKs already carry. An
agent worker's token is VideoGrants { agent: true } and a simulated job's
participant token is kind: "agent"; neither could be minted from Rust before.
livekit-uniffi exposes both: TokenOptions.kind, Claims.kind, and agent
on its VideoGrants record.
Breaking: the four grants the server infers when absent -- can_publish,
can_subscribe, can_publish_data, can_update_own_metadata -- are now
Option<bool>, as in the Go and JS SDKs. None leaves the decision to the
server, and the new getters (can_publish(), can_subscribe(),
can_publish_data(), can_update_own_metadata()) read a token the way the
server does, can_publish_data falling back to can_publish included. Code
that set these fields writes Some(..); code that read them uses the getters.
The same fields are optional on the livekit-uniffi record, and
livekit-api re-exports the crate as livekit_api::access_token, so both
carry the change.
Nothing at its default is written into the token any more: unset claims and
grants are omitted, as the server's own omitempty grants are. Verification
of existing tokens is unchanged.
0.12.82 - 2026-10-07 #
Fixes #
- Scale the
x-google-start-bitratehint by connection setup time: the 1 Mbps camera cap now applies to connections that set up within 1.5 s and ramps linearly down to 300 kbps at 3.5 s or slower; screen share stays uncapped. - Increase the local and remote data track event buffers so a burst of track lifecycle events cannot fill the channel and deadlock room disconnect.
- Manage VAAPI displays across capability probes and encoder lifecycles so driver resources and threads are released without invalidating active encoders.
- Release local published tracks when a server-initiated room closure has already closed the peer connection.
- RoomClient::send_data now returns a Send future by dropping its RNG before awaiting.
- Never wait on the peer when closing a signal stream: a half-open socket (no FIN/RST) left the reader parked in
recv()forever, so the resume never dialled andRoom::closehung.SignalStream::closenow aborts the parked reader and bounds the writer's Close flush.
0.12.81 - 2026-09-24 #
Fixes #
- Avoid panics when malformed RTC error headers contain non-ASCII text.
- Java version in libwebrtc was bumped by Google, downgrade it again for Unity 2022 build compatibility - #1456 (@MaxHeimbrock)
- refactor(signaling): explicit signal lifecycle state machine - #1402 (@lukasIO)
- Upgrade to prost 0.14 across the whole project - #1447 (@1egoman)
Export SoxResampler through UniFFI
SoxResampler is now a UniFFI object, so foreign hosts can construct one and
drive push / flush directly, receiving the resampled samples by value.
The FFI request surface is unchanged: NewSoxResampler, PushSoxResampler and
FlushSoxResampler still hand back a pointer into the resampler's own buffer,
readable until the next call on that resampler. Both paths are covered by new
tests that drive them with the same input and compare the results.
0.12.80 - 2026-09-22 #
Features #
- Add H.264 and H.265 access unit parsing.
- Add the scaffolding for exposing video capture over FFI.
- Add a capture source that renders a wall clock on the GPU.
- Add a capture source that renders a test pattern on the GPU.
Fixes #
- Add a workspace-owned command for generating C++ bindings from
livekit-ffiwith the pinned LiveKit UniFFI C++ generator. - Correct the AppKit framework name so macOS linking works with case-sensitive SDK filesystems.
- Cancel and join the room SID notification task on disconnect so closing a room before its SID arrives does not retain the room and its WebRTC state.
- Cleanup unused dependencies
- Fix reliable data channel replay: keep the full retry buffer across resumes, drop duplicate reliable packets, and ignore replayed chunks on uncompressed streams instead of failing with
MissedChunk. - Stop answering a server-initiated Leave (room deleted, duplicate identity) with a client Leave. The server has already ended the session and is closing the signalling socket, so the reply only ever produced the warning "dropping pass-through signal — no stream available" on every such disconnect.
Fix the RTC engine leaking when a room is dropped without being closed.
The engine's event task held a strong reference to the engine, while the signal that stops
that task lives inside the engine itself, so the two kept each other alive. An engine
dropped without an explicit close() released nothing: the session, both peer connections,
the WebRTC runtime, and the signal client with its open websocket all stayed resident for
the lifetime of the process, and the server kept its half of the session because the socket
was never closed. The task now holds a weak reference and stops on its own once the engine
is gone.
EncryptionError::Failed and DecryptionError::Failed carry a reason string and are no longer flat_error,
so a foreign EncryptionProvider or DecryptionProvider returning an error no longer aborts the process with
"Can't lift flat errors" -- a failed data track decrypt (no E2EE manager, key mismatch, corrupt frame) now
drops the frame and leaves the room connected.
Report the reconnect reason to the server when resuming.
Resumes previously sent no reason, so server-side telemetry could not attribute why Rust
clients reconnect — every resume looked like RR_UNKNOWN. The engine now records what caused
the episode (signal disconnected, publisher failed, subscriber failed) and reports it on each
resume attempt. The v0 signalling path was also missing the reconnect_reason query parameter
entirely, so it would not have been reported even if a reason had been supplied.
Fix resume reporting success for a PeerConnection that had not recovered.
A resume decided recovery from PeerConnectionState, which keeps reading Connected for tens
of seconds after the far end goes away. A resume could therefore emit Resumed — and so
RoomEvent::Reconnected with ConnectionState::Connected — for a session whose subscriber
transport was dead, leaving applications with no signal that they had stopped receiving media.
A resume now requires each transport to have entered Connected since the resume began, or to
have held it throughout, rather than trusting the state it currently reports.
VideoGrants gains the agent grant and Claims the kind claim (with
AccessToken::with_kind), which the Go, Python and JS SDKs already carry. An
agent worker's token is VideoGrants { agent: true } and a simulated job's
participant token is kind: "agent"; neither could be minted from Rust before.
livekit-uniffi exposes both: TokenOptions.kind, Claims.kind, and agent
on its VideoGrants record.
Breaking: the four grants the server infers when absent -- can_publish,
can_subscribe, can_publish_data, can_update_own_metadata -- are now
Option<bool>, as in the Go and JS SDKs. None leaves the decision to the
server, and the new getters (can_publish(), can_subscribe(),
can_publish_data(), can_update_own_metadata()) read a token the way the
server does, can_publish_data falling back to can_publish included. Code
that set these fields writes Some(..); code that read them uses the getters.
The same fields are optional on the livekit-uniffi record, and
livekit-api re-exports the crate as livekit_api::access_token, so both
carry the change.
Nothing at its default is written into the token any more: unset claims and
grants are omitted, as the server's own omitempty grants are. Verification
of existing tokens is unchanged.
Write the x-google-start-bitrate hint once per publisher connection, and exempt screen share from the 1 Mbps cap.
libwebrtc reads this fmtp parameter per m-section but applies it to the shared Call (WebRtcVideoSendChannel::ApplyChangedParams -> SetSdpBitrateParameters), where RtpBitrateConfigurator holds one config for the whole peer connection. It retains start_bitrate_bps and re-applies it on network route changes (RtpTransportControllerSend::OnNetworkRouteChanged), so a WiFi-to-cellular handover re-seeds the estimator from the original hint with no renegotiation. Rewriting the value on later offers was therefore at best a no-op and at worst a restart of a converged bandwidth estimator; it is now written only on the first offer that carries local video, and only once that offer is accepted locally. A full reconnect builds a new peer connection and seeds the new estimator again. The initial offer sent with the JoinRequest in single PC mode never carries the hint: it is created before any track is published, so no target bitrate exists yet.
Screen share is no longer capped at 1 Mbps, matching client-sdk-js and client-sdk-android: unlike camera content, a screen share is published at a high bitrate so text stays legible, and a conservative start costs more than a brief overshoot.
0.12.79 - 2026-09-10 #
Fixes #
- Add FFI support for publishing pre-encoded video frames and consuming encoder feedback.
0.12.78 - 2026-09-09 #
Fixes #
- Load the Jetson MMAPI encoder's runtime libraries (libnvbufsurface, libv4l2/libnvv4l2) lazily via dlopen instead of linking them, so an aarch64 binary built with Jetson support also loads on non-Jetson ARM systems and falls back to other encoders there.
- log warning if signal messages get dropped during reconnect - #1391 (@lukasIO)
Own the add_ice_candidate completion state
PeerConnection::add_ice_candidate captured ctx and on_complete by reference in the
completion lambda it hands to libwebrtc. That completion runs asynchronously on the signaling
thread and is deferred behind the operations chain whenever it is busy, for example while a
SetRemoteDescription is in flight, so it could execute after the calling frame had returned
and dereference freed stack memory (a crash on the signaling thread on the first ICE candidate
in practice). The lambda now owns its state through a shared_ptr.
Moves the RPC implementation into a new livekit-rpc crate, alongside the existing
livekit-data-stream and livekit-datatrack crates.
Breaking: the livekit::rpc module is gone. The RPC types it held are unchanged and
still re-exported from livekit::participant and the prelude, so most code needs no edit;
code that spelled the module out (use livekit::rpc::RpcError;) should import from
livekit::participant or the prelude instead. RpcClientManager, RpcServerManager and
HandleRequestOptions remain reachable under livekit::participant but are now
#[doc(hidden)]: they are internal SDK API and were never usable without the (private)
transport trait.
Within the livekit crate itself, RPC types are now imported from livekit-rpc directly
rather than through those re-exports.
The new crate does not depend on libwebrtc, so its unit tests run without building WebRTC.
The transport seam that made this possible was already in place; the only change to it is
that RpcTransport::publish_data now returns a message-only RpcTransportError instead of
livekit::RoomError, mirroring livekit_data_stream::api::SendError.
Also fixes four latent bugs found while moving the code:
- An RPC call to a participant who disconnects mid-call now fails promptly with
RecipientDisconnected. Pending calls were never purged on disconnect, so the caller waited out its full response timeout (15s by default) and gotResponseTimeoutinstead. - A server reporting a version that is not valid semver no longer panics the calling task. An unparseable version is no longer treated as evidence that the server is too old.
- A v1
RpcResponsecarrying a compressed payload, or no value at all, now fails with anApplicationErrorinstead of resolving the caller with an empty successful response. - Removed an unguarded
unwrapwhen building a v1 response packet, by giving the function a signature that cannot represent the invalid state.
Also drops the semver dependency from livekit, which was only used by the RPC client.
0.12.77 - 2026-09-08 #
Features #
- Removes livekit-runtime and converts this package to be tokio only again - #1375 (@1egoman)
Fixes #
- Add data streams v2 to exposed uniffi interface - #1286 (@1egoman)
- Add the
PASSTHROUGHencoding preset and remove the unusedUpdateEgressRequestfrom the generated protocol - Fix pre-encoded frame segfault on macOS
- Handle capture of dmabuf using existing capture path
- Add
self_test_http_get/self_test_ws_echo/has_http_client/has_ws_clientUniFFI exports so foreign hosts can exercise the transport seam end-to-end.
Make AdmProxy worker-thread-affine: all platform ADM access now happens on the WebRTC worker thread, matching the ADM threading contract.
- The platform ADM is now created lazily on the first PlatformAudio acquire on all platforms, so apps that never use platform audio never construct it.
- Fixes Android platform recording delivering no audio: the audio transport was never registered on the lazily created ADM.
- Fixes a shutdown race by keeping the runtime threads alive as long as Rust can reach the audio device controller.
- Adds a
platform_audioexample exercising the PlatformAudio API and the worker-thread marshaling.
Add agent guidance for detecting and preventing memory-lifecycle regressions in
Rust, FFI, and native WebRTC code.
Close peer connections before awaiting signal teardown
SessionInner::close released the peer connections only after two awaits that can block
indefinitely, so cancelling close() — for example by wrapping it in a timeout — left the
transports open and their ICE UDP sockets bound for the lifetime of the process. Long-lived
clients eventually exhausted their file descriptors. The transports are now closed before
the first await, which makes the teardown safe to cancel.
Moves the internal region-discovery cache into a new livekit-region crate. No
public API or behaviour change.
Moves the signalling client into a new livekit-signaling crate. livekit-api
re-exports it under the historical livekit_api::signal_client path, now marked
deprecated: it is internal SDK API, and dependents should use livekit-signaling
directly. livekit-api no longer depends on livekit-net.
Also drops two dependencies that were declared but never used: scopeguard and
bytes.
Fix CUDA and FFI resource cleanup during SDK shutdown.
NVIDIA encoder and decoder factories now share a reference-counted CUDA context
and destroy it when the final factory is dropped. FFI shutdown now releases
leftover handles one at a time so nested drop_handle calls do not re-enter
DashMap::clear(). Adds regression coverage for FFI-handle, watcher, and
configuration cleanup during disposal.
Expose network_type on IceCandidateStats
Chromium's local RTCIceCandidateStats carries a non-standard networkType field (WiFi,
cellular, ethernet, etc.), but IceCandidateStats had no place to put it, so it was silently
dropped during get_stats() deserialization. Adds network_type: Option<String> to the struct;
non-breaking since it already derives #[serde(default)].
Fix room-session and data-channel leaks across connect/disconnect cycles.
The E2EE manager callback now captures RoomSession weakly so the session can
drop after disconnect. Data-channel observer callbacks are cleared during RTC
teardown so the observer/callback cycle cannot keep peer connections alive.
Adds regression coverage for room-session destruction and data-channel callback
cleanup.
Fix native video-source lifecycle and NVENC initialization failure handling.
The raw-video keepalive task now uses a weak liveness check and defers its
black I420 buffer allocation until source liveness is confirmed, so dropping
an unused source releases its resources. nvEncInitializeEncoder failures now
propagate instead of leaving the encoder half-initialized.
0.12.76 - 2026-08-25 #
Fixes #
- differentiate signal connection errors correctly from timeouts - #1234 (@lukasIO)
- Automatically retry webrtc build downloads
- feat: upgrade libwebrtc to m150. - #1284 (@cloudwebrtc)
- fix(uniffi): register the Bytes custom type once, in livekit-common - #1343 (@pblazej)
- Strip DWARF debug information from release iOS FFI archives.
- yuv-sys: fail with an actionable error when the libyuv submodule is not initialized
fix: bump libwebrtc to webrtc-b9233c3-2 so TURN/TLS can use the OS trust store
WebRTC validates TURN/TLS against a small set of anchors compiled into
rtc_base/ssl_roots.h, generated in 2023 from Google's own PKI list. It has no
Amazon, Starfield Services or ISRG roots, so a relay-only connection to a TURN
server fronted by AWS ACM or Let's Encrypt times out with unknown_ca even
though the host OS trusts the chain.
This build picks up webrtc-sdk/webrtc#277, which falls back to the operating
system's trust store when the built-in anchors yield no path. It sits in
rtc_base below every SDK, so the C++ API that webrtc-sys binds is covered
without any Rust-side change.
feat: enable WARP (SPED + SNAP) by default, gated by the server
WARP is now always enabled on the client and negotiated with the SFU: SPED
(DTLS-in-STUN) via the WebRTC-IceHandshakeDtls field trial, and SNAP
(SCTP-INIT-in-SDP) via the RtcConfiguration.enable_sctp_snap field. When the
server does not enable WARP it is not advertised and the connection falls back to
plain DTLS/SCTP, so there is no client-side toggle.
BREAKING CHANGE: libwebrtc::RtcConfiguration is now #[non_exhaustive] and has a
new enable_sctp_snap field. Construct it from RtcConfiguration::default() and set
the fields you need instead of a struct literal.
Moves access-token generation and verification into a new livekit-token crate.
livekit_api::access_token::* continues to resolve to the same types via a
re-export, so no consumer changes are needed.
Also fixes the services-tokio and services-async features, which used the
access-token types without declaring the access-token feature. Building with
--no-default-features --features services-tokio previously failed to compile.
0.12.75 - 2026-08-10 #
0.12.74 - 2026-08-03 #
Fixes #
- Add a unified
EgressClient::start_egressthat calls the v2Egress.StartEgressRPC with aStartEgressRequest, alongside the existing per-type helpers. - Fix a publisher-transport deadlock during renegotiation. When another negotiation was requested while an offer was awaiting its answer,
set_remote_descriptionre-enteredcreate_and_send_offerwhile holding the transport's non-reentrant inner mutex, permanently wedging publishing. - Move the concept of "internal" data streams into
livekitcrate fromlivekit-data-stream- #1304 (@1egoman)
Fix H.264, H.265, and AV1 NVENC sessions so live bitrate and frame rate updates
reconfigure the hardware without restarting the encoder.
0.12.73 - 2026-07-29 #
Fixes #
- allow for audio filters to be registered after initial room connection - #1273 (@lukasIO)
- Caching of tokio backend reqwest http client - #1285 (@MaxHeimbrock)
- Add data streams v2 - #1192 (@1egoman)
- Ensure participant disconnects are synthesized after connection resume - #1250 (@lukasIO)
0.12.72 - 2026-07-27 #
Fixes #
- Address typo in parsing rpc server version - #1268 (@1egoman)
- Data tracks schema metadata support.
- Emit black keepalive frames from NativeVideoSource instead of uninitialized memory. webrtc::I420Buffer::Create leaves the pixel planes uninitialized, so the pre-capture keepalive frames could leak recycled heap contents (often fragments of earlier frames from the same process) to subscribers as the first keyframes - #1271 (@eh-steve)
- ensure failing audio filter init doesn't degrade audio quality - #1270 (@lukasIO)
- Add NVIDIA NVENC AV1 encoding when the GPU reports AV1 encode support.
Route LiveKit signalling through a pluggable transport (new livekit-net crate).
The signalling WebSocket and the two pre-connect HTTP GETs (validate, region discovery) now go through pluggable transport traits (WsClient for the WebSocket, HttpClient for request/response) resolved from a process-global registry with independent slots — a consumer can bring only HTTP, or only WebSocket. The new livekit-net crate owns the WebSocket/HTTP/TLS stack behind those traits and ships native (tokio / async-std) backends. Native builds are unchanged in behavior.
Breaking (livekit-api, and livekit via EngineError::Signal):
SignalError::WsErroris removed —tungsteniteis no longer part of the public API. A failed WebSocket handshake now surfaces its HTTP status asSignalError::Client/Server; transport connection and close failures surface as the newSignalError::Connection(String)/SignalError::Closedvariants (previously all collapsed intoTimeout).SignalErroris now#[non_exhaustive], and gains aSignalError::TransportNotConfiguredvariant — returned when no transport is registered (host/foreign builds must calllivekit_net::set_ws_client/set_http_clientbefore connecting). This is a permanent configuration error; callers must not retry.- The signalling WebSocket/HTTP/TLS crates are no longer transitive dependencies of
livekit-api; TLS features delegate tolivekit-net. Existingsignal-client-tokio/-async/-dispatcherand TLS feature names are unchanged.
0.12.71 - 2026-07-17 #
Fixes #
- Emit room EOS when the underlying LiveKit room event channel closes after a server-initiated disconnect, and ignore duplicate disconnect events during teardown.
- Don't log an expected publisher data channel close as unexpected - #1224 (@longcw)
0.12.70 - 2026-07-14 #
Fixes #
- refactor: extract data-stream logic and shared types into new
livekit-commonandlivekit-data-streamcrates (public API unchanged; types are re-exported fromlivekit) - Use concrete type for data track manager output events
0.12.69 - 2026-07-09 #
Fixes #
- feat: auto failover APIs with LK Cloud - #1196 (@davidzhao)
- Fix for dynacast error - #1213 (@MaxHeimbrock)
- Fix malformed RTC error handling
- Handle data track SID reassignment
- introduce LiveKitAPI construct, added smoke tests - #1220 (@davidzhao)
- Turn single peerconnection off by default - #1206 (@cnderrauber)
0.12.68 - 2026-06-30 #
Features #
- Add
user_datasupport to frame metadata, allowing arbitrary application-supplied bytes to be attached to a video frame via thePTF_USER_DATApacket trailer feature.
Improve initial video quality by setting x-google-start-bitrate SDP hint for all video codecs (VP8, VP9, AV1, H264, H265) and defaulting to MaintainResolution degradation preference.
This addresses the issue where video starts blurry for several seconds before improving, by:
- Telling WebRTC's bandwidth estimator to start at 70% of target bitrate instead of ramping up from ~300kbps
- Preferring frame drops over resolution reduction when bandwidth is constrained
The DegradationPreference option is now exposed via FFI for Python, C++, Unity, and Node SDKs.
Add MaintainFramerateAndResolution to DegradationPreference enum to align with WebRTC M144.
MAINTAIN_FRAMERATE_AND_RESOLUTIONis now the recommended value (replaces deprecatedDISABLED)DISABLEDis deprecated but still supported for backwards compatibility- Both values map to the same behavior: maintain framerate and resolution, dropping frames if needed
Fixes #
- Fix AV1 subscriber decode when packet trailers are enabled.
- Improve log messages around plugin loading - #1186 (@lukasIO)
0.12.67 - 2026-06-24 #
Fixes #
- Increase room event ready timeout
- harden reconnect behaviour - #1148 (@lukasIO)
0.12.66 - 2026-06-23 #
0.12.65 - 2026-06-19 #
Fixes #
- fix: escalate to full reconnect if connection failed during a resume - #1175 (@davidzhao)
0.12.64 - 2026-06-17 #
Fixes #
- Add
LK_DISABLE_NVDECto bypass NVIDIA NVDEC decoder registration when the environment variable is set. - return DeviceNotFound when device is not there for set_recording_devi… - #1155 (@xianshijing-lk)
Add dynacast support - #1003 (@chenosaurus, @stephen-derosa)
This includes a minor breaking change for libwebrtc: RtpParameters now
contains additional RTP sender state that must be preserved when round-tripping
through set_parameters().
0.12.63 - 2026-06-09 #
Fixes #
- Reject oversized data messages before they break the data channel.
- Upgrade dashmap to v6
- Add per-publication video encoder backend selection. Add a video encoder backend availability query. Remove
LIVEKIT_PREFERRED_HW_ENCODERin favor of per-publication backend selection.
0.12.62 - 2026-06-03 #
Fixes #
- Add rpc max_round_trip_latency and move to builder pattern - #1127 (@1egoman)
- [allow(dead_code)] for dead function in room module - #1128 (@stephen-derosa)
- Send publisher offer with join request to accelerate connection - #996 (@cnderrauber)
0.12.60 - 2026-05-21 #
Features #
- Introduce pipeline options for remote data tracks, support multiple in-flight frames.
feat: add Android application context initialization for PlatformAudio support.
Android requires ContextUtils.initialize(applicationContext) before WebRTC audio components can be created. This change:
- Adds
livekit_ffi_initialize_android_context()C FFI function for Unity and other FFI consumers - Uses
CreateAndroidAudioDeviceModule()instead of genericCreateAudioDeviceModule()on Android - Handles empty device GUIDs on Android (falls back to index 0)
- Documents Android-specific limitations: single default device, no app-level device selection
Platform notes:
- Android device enumeration returns only one "default" device with empty name/GUID
- Audio routing (speaker/earpiece/Bluetooth) is controlled by Android's AudioManager, not WebRTC
Fixes #
- Filter internal data streams out of livekit-ffi interface - #1112 (@1egoman)
0.12.59 - 2026-05-20 #
Fixes #
- Bugfix: Always emit Disconnected on engine close - #1096 (@MaxHeimbrock)
- (WIP) FFI room event ready signal after initial connection - #1068 (@ladvoc, @stephen-derosa)
- Support for large RPC messages using data streams - #1013 (@1egoman)
0.12.58 - 2026-05-18 #
Features #
- FFI logging improvements
Make sample_rate and num_channels optional in NewAudioSourceRequest.
These fields are ignored for AudioSourcePlatform (ADM uses hardware native settings) and for AudioSourceNative fast path (queue_size_ms=0, frame values used directly). Defaults to 48000 Hz and 1 channel when not specified.
Fixes #
- fix: don't fire local_track_subscribed during reconnect - #1099 (@davidzhao)
- Fix LocalTrackPublished handle leak - #1065 (@MaxHeimbrock)
- Return EOS event from data track stream read request
0.12.57 - 2026-05-14 #
Fixes #
- feat: add scalability mode for AV1/VP9. - #1076 (@cloudwebrtc)
- Add
LIVEKIT_PREFERRED_HW_ENCODERto prefernvencorvaapihardware video encoding when both are available. - Reword audio filter logs to be less confusing - #1092 (@1egoman)
Get WebRTC ADM into Rust - #1037 (@xianshijing-lk)
This PR introduces platform audio device management via WebRTC's Audio Device Module (ADM).
Features
- ADM Proxy: New
AdmProxyclass that switches between Dummy ADM (synthetic mode) and Platform ADM (real audio I/O) - PlatformAudio API: High-level Rust API for microphone capture and speaker playout with AEC/AGC/NS
- Device enumeration: List and select recording/playout devices by index or GUID
- Mode switching: Seamlessly switch between synthetic mode (FFI callbacks) and platform mode (native speakers) while audio is active
- FFI platform audio support: Expose platform audio device enumeration and selection through
livekit-ffi - Audio processing: Configure echo cancellation, noise suppression, and auto gain control with platform-specific defaults (hardware on iOS, software elsewhere)
Audio Modes
| Mode | Recording | Playout | Use Case |
|---|---|---|---|
| Synthetic | NativeAudioSource | Dummy ADM + FFI | Unity audio, agents |
| Platform | Platform ADM mic | Platform ADM speakers | VoIP with AEC |
API
// Create PlatformAudio for microphone/speaker access
let audio = PlatformAudio::new()?;
// Enumerate and select devices
for i in 0..audio.recording_devices() as u16 {
println!("Mic {}: {}", i, audio.recording_device_name(i));
}
audio.set_recording_device(0)?;
// Create audio track for publishing
let track = LocalAudioTrack::create_audio_track("mic", audio.rtc_source());
0.12.56 - 2026-05-11 #
Fixes #
- fix: Sync inner.enabled state for E2EE manager. - #1073 (@cloudwebrtc)
- Upgrade protocol to v1.45.8
0.12.55 - 2026-05-11 #
Fixes #
- chore: add LocalTrackRepublished event for FFI clients - #1072 (@davidzhao)
0.12.54 - 2026-05-10 #
Features #
- Bump
rustls-webpkito 0.103.13, addressing GHSA-82j2-j2ch-gfr8 - Expose error message on EOS for data track subscriptions
Fixes #
- Fix missing
libwebrtc.jarfor Android builds, harden build scripts - fix: derive
simulcastedfrom non-deprecated TrackInfo fields - #1052 (@cloudwebrtc) - fix race in download_webrtc to reduce flaky build - #1047 (@hechen-eng)
- Improve WebRTC build scripts and add external_audio_source patch - #1053 (@xianshijing-lk)
- support SimulateScenario through FFI to improve testing - #1069 (@davidzhao)
- TEL-464: reduce redundant resampling in audio filter - #1019 (@hechen-eng)
0.12.53 - 2026-04-23 #
Features #
Add support for frame level packet trailer
##890 by @chenosaurus
- Add support to attach/parse frame level timestamps & frame ID to VideoTracks as a custom payload trailer.
- Breaking change in VideoFrame API, must include
frame_metadataor use VideoFrame::new().
Fixes #
- Add device-info crate and send device_info to telemetry - #982 (@maxheimbrock)
- Fix data track packet format issue breaking E2EE
- Fix unbound send queue that can cause latency in data track messages - #1032 (@chenosaurus)
- Fix for raw stream drop called from non tokio thread like Unity .NET GC - #1016 (@MaxHeimbrock)
0.12.52 - 2026-04-02 #
Features #
- Initial support for data tracks
Fixes #
use the bounded buffer for video stream
##956 by @xianshijing-lk
Before this PR, it uses an unbounded buffer for video stream, that will cause multiple problems: 1, video will be lagged behind if rendering is slow or just wake up from background 2, it will be out of sync with audio
This PRs provides options to set a bounded buffer for video stream, and use 1 buffer as the default option.
0.12.51 - 2026-03-31 #
Fixes #
- Expose participant active event, state, and joined at
- fix unity android build with "livekit" prefixed jni - #983 (@xianshijing-lk)
- Upgrade to thiserror 2
0.12.50 - 2026-03-22 #
Features #
E2EE: allow setting key_ring_size and key_derivation_algorithm, update webrtc to m144
##921 by @onestacked
This PR uses this webrtc-sdk PR to configure the KDF.
I've tested this with https://codeberg.org/esoteric_programmer/matrix-jukebox and it is compatible with Element Call.
Fixed: https://github.com/livekit/rust-sdks/issues/796
Fixes #
- Fix H.264 codec matching
add bounded buffer to audio_stream, and use 10 frames as the default
##945 by @xianshijing-lk
Send client os and os_version from rust
##952 by @MaxHeimbrock
Adds os_info crate as dependency and sends the data for client connections.
Metadata
Documentation
Publisher
Weekly Downloads
Metadata
Dart bindings for the LiveKit Rust SDK core, delivered as a prebuilt native library via Native Assets.
License
Apache-2.0 (license)
Dependencies
archive, code_assets, crypto, ffi, hooks, http