Create the LiveKit room through the RoomService API before returning a join token, on join, lobby admission, roomkit dial-in and connection tests.
Rooms are only created when missing: LiveKit's CreateRoom overwrites the metadata of an existing room, which would clear the recording state.
This lets LiveKit run with room.auto_create: false, so Meet can share an instance with services that need it to stay secure.
Refs #1209
Add a documentation site built with Zensical: getting started,
self-hosting with compose and Kubernetes, configuration, user guide,
contributing and reference pages, and every environment variable on
one page, grouped by feature.
Co-authored-by: davd-gzl <60177543+davd-gzl@users.noreply.github.com>
The ProConnect sign-in link was an empty <a> named only by an
aria-label, its wording living in a CSS background image. RGAA
test 6.2.1 asks for a label between <a> and </a>.
Move the label into visually hidden text inside the link and drop
the aria-label, so the accessible name comes from the content.
Fixes#1898
Only cards with expandable details had an h3, so the titles of the
other cards were missing from the heading structure.
Wrap the row of non-expandable cards in an h3 too. No visual change.
RGAA 9.1
Fixes#1902
The global focus rule overrode the trigger's inset outline offset,
pushing the ring outside the card where overflow: hidden clipped it.
Force the inset offset so the ring stays visible.
RGAA 10.7
Fixes#1905
Most people using La Suite Meet never land on GitHub, and those
who do had no prompt to star the project. Stars help other people
and public organisations discover it.
Add a single line at the end of the README introduction, kept
discreet to avoid stacking a second callout next to the Matrix one.
Below 800px the settings tabs only render their icon, which left
each tab, and the tab panel it labels, without an accessible name.
Screen reader users could not tell which settings section they
were opening.
Set aria-label on each tab with its existing translated label.
On wide screens it matches the visible text, so nothing changes.
Fixes#1855
Inactive dots of the home intro slider were 1.57:1 on white, below
the 3:1 RGAA 3.3 asks for, so low-vision users could not tell how
many slides exist.
They are now hollow: a primary.100 fill with a primary.500 outline
at 4.22:1 on white. This passes without darkening the dots so much
that they blend with the active one. Dots grow from 6px to 8px to
keep the outline legible.
Fixes#1851
The step cards of each connection test section are a series of
similar items but were rendered as plain divs, so assistive
technologies could not announce them as a list or give the item
count.
Render each series as a ul with one li per card. The heading and
accordion stay inside the li. No visual change.
Fixes#1903
Selecting a language in the settings modal translates the whole
interface immediately. Nothing announced this beforehand, so a
screen reader user could end up in a language they don't read,
with the dialog title already announced in it.
Add a visible description under the "Langue de l'application"
field, linked to the select trigger with aria-describedby, stating
that the change applies to the whole interface on selection.
This satisfies RGAA 7.4 (WCAG 3.2.2 On Input) without adding an
extra confirmation step.
Fixes#1859
Zero counters, pending and skipped labels and step summaries used
greys below the 4.5:1 minimum of RGAA 3.2. They now use greyscale.600,
which passes on white, hovered and expanded rows.
Fixes#1897
Language names are listed in their own language ("English",
"Deutsch", "Nederlands"...), but nothing told assistive technologies
so. Screen readers pronounced them with the rules of the interface
language, which can make them hard to recognise.
Select items accept an optional `lang`, rendered on a span around
the label. useLanguageLabels sets it for every language it returns.
RGAA 8.7
Fixes#1861
Text colors were set without a matching background on the document.
With custom browser or system colors, the user's background could
end up behind our text color and make it unreadable.
Set both the default background and text colors on body, so they
are always defined as a pair.
RGAA 10.5
Fixes#1863
In the transcript panel, the button showing the meeting language
opens the transcription settings, but its accessible name was only
the language itself, e.g. "Français (fr)". Screen reader users had no
way to know it changes the meeting language.
Add an aria-label that keeps the language and states the action. The
visible label is unchanged. Translated in all six locales.
RGAA 7.1
Fixes#1868
The side panel already exposes its title as a level 1 heading. The
transcript and screen recording panels rendered their intro line as a
second h1, so screen reader users navigating by headings could not
tell the panel title from its subtitle.
Render both intro headings as h2 while keeping the h1 visual style,
so the hierarchy is correct and nothing changes on screen.
RGAA 9.1
Fixes#1882
A raised hand matters as much as someone speaking, but it went
unnoticed when the tile was scrolled out of the carousel. It is now
brought to the top like a new speaker and stays visible while raised.
In the focus layout, when someone off screen started speaking, their
tile replaced whichever tile was evicted, often at the bottom of the
column. A screen share that started did the same. Users had to scan
the whole column to find who was talking, and the spot changed every
time.
New speakers now always appear at the top, right below the local
camera and the screen shares, and the other tiles slide down by one.
The most relevant tiles are where the eye already is, and the column
moves in a predictable way.
The ordering still relies on LiveKit's sort and stable update; only
the position of the incoming tile changes. This is the enhancement
#1805 was after.
Regression: in the focus layout, a participant who started speaking
while their tile was outside the visible part of the carousel was no
longer brought into view.
Cause: 81f23aeb moved useTracks in StageLayout to `updateOnlyOn: []`
so the whole stage stops re-rendering on every ActiveSpeakersChanged.
useVisualStableUpdate only promotes speakers when its host renders, so
it needs something to wake it up. The grid layout got that with
useSpeakerPromotionTrigger in the same commit; the carousel did not,
and nothing else re-renders it on speaker events.
Fix: the carousel now uses the same tripwire. Unlike the grid, which
passes its current page, the carousel renders every tile in a scroll
container, so only the first maxVisibleTiles are passed as visible.
The tripwire also takes the full list of tiles the layout shows and
ignores speakers who have none. In the focus layout the pinned camera
is rendered by the focus view, not the carousel: without this, each
speaker event from that participant would force a render that cannot
promote anyone. The grid passes all its tracks, so its behaviour is
unchanged.
Known limit, not addressed here: when a single tile fits, it stays on
the local camera, which useVisualStableUpdate sorts first, so a remote
speaker cannot be promoted and the tripwire keeps firing for them.
This supersedes #1805, which will be closed. On top of this fix, it
pinned the active speaker to a fixed slot next to the local camera,
first with a 2s hold, a reserved slot and an eligibility check
(867a14d0), then simplified to the last remote speaker with no timer
(378e3d01). That is an enhancement rather than part of the fix, and it
is left out here to keep the regression fix small. It can come back in
a dedicated PR; it would also cover the single-tile case above.
Since routes are lazy loaded, the LiveKit stylesheet only ships
with the room chunk. It was the only rule applying `--fonts-sans`
to the document, so pages outside a room fell back to Panda's
preflight system font stack, ignoring custom CSS themes (e.g.
Marianne in the DINUM image) and accessibility fonts.
Point Panda's `--global-font-body` at `--fonts-sans` so the
document font follows the token regardless of which chunks are
loaded.
A stored null was read as disabled by CanMuteParticipant, since
`.get(key, True)` only defaults when the key is missing, while the
frontend reads null as enabled. Only an explicit `false` now
disables participant muting.
Fixes#1942
Supersedes #1944
The Marianne subsets were generated with a French-only whitelist, so
most Portuguese accented letters were missing. Words like "não",
"ação" or "país" fell back to the system font for single letters,
which looked broken in the middle of Marianne text.
Regenerate the 12 files from the full Marianne sources tracked in
numerique-gouv/lasuite-landingpage (src/fonts/sources), using the
glyphhanger 5.0.0 command documented there, with these characters
added to the whitelist:
- letters: Á Â Ã Í Ó Õ Ú Ü á ã í ó õ ú ü
- ordinal indicators: º ª
- punctuation: – — ‘
The previous files came from an older whitelist, so this also brings
in the characters added upstream since then (ä ł ń $ ~). No glyph
was removed. Each file grows by roughly 1 KB.
Add Portuguese (pt / pt-BR) localization across the frontend UI
(all 11 i18next namespaces) and backend Django catalog (pt_BR).
Contributed by the Alan Turing Laboratory (ATLab) at the Federal
University of Ceara (UFC), Brazil.
Meeting links inserted into events by the Outlook calendar add-in carry
tracking query parameters, e.g. `?from=outlook-addin` or `utm_*`.
Pasting such a link into the "Join a meeting" dialog failed: the
validation regex was anchored on the room id, so any trailing query
string made the link invalid, and the submit handler only stripped the
origin, so the query string would have ended up in the room id.
Parse the input with the URL API in a new `extractRoomId` helper, used
by both validation and submit. It accepts a bare room id or a link on
the current origin (trailing slash allowed), reads the room id from the
pathname only, and returns the query string and fragment separately so
they are kept when redirecting to the room (utm tracking is preserved).
The nginxinc/nginx-unprivileged image is no longer rebuilt and
ships libxslt 1.1.43-r3, affected by CVE-2026-125350,
CVE-2026-87636 and CVE-2026-9947.
Use nginx/nginx-unprivileged in both frontend images. Its
packages are up to date, so drop the manual package upgrades.
The "Your meeting is ready" dialog is a react-aria Dialog rendered
without a ModalOverlay, so nothing handled Escape. It stays non-modal
on purpose, so the meeting remains usable while it is open.
Reuse useEscapeToClose, as the side panel does, so Escape only closes
the dialog when focus is inside it and does not take the key from the
chat, the side panel or menus.
The dialog opens on its own when joining, so there is no trigger to
restore focus to. Both the close button and Escape now move focus to
the first control bar button instead of letting it fall back to body.
Record signed URL protection, secret-aware validation errors and safer
logging and Sentry capture under Unreleased.
Document the retention of redacted frame locals and application DEBUG
diagnostics.
Redact signed URLs and remove untrusted exception messages, request
bodies and breadcrumb payloads from Sentry events and transactions.
Preserve stack-frame locals after the existing content scrubber runs.
Cover real HTTP failures, chained exceptions and DEBUG breadcrumbs.
Raise safe LLM errors for upstream failures and invalid structured
responses without retaining the original exception chain.
Keep generated section titles out of INFO progress logs and preserve
the existing DEBUG diagnostics.
Replace raw download, transcription, webhook and Docs exceptions with
safe messages while preserving retryable exception categories.
Retain document IDs and DEBUG responses. Give webhook regression tests
their own tenant settings instead of relying on development tenants.
Use cleaned validation errors for task execution, prerun signals,
transcription responses and API results. Share the same cleanup with
FastAPI validation responses while retaining ordinary input values.
Verify secret masking, preserved diagnostics and summary retries.
Validate signed URLs as secrets and reveal them explicitly for file
access, Celery messages, webhook delivery and authorized API responses.
Cover model representations, nested broker payloads and URL consumers.
Preserve Pydantic error categories and diagnostics while masking nested
SecretStr inputs and avoiding retained validator exception objects.
Support custom error codes and safe pickling for Celery.
Strip URL credentials, query strings and fragments from file-service
logs and analytics source metadata using one shared helper.
Cover malformed URLs, non-string inputs and signed source URLs.
A registered room's LiveKit room is named after its id, and a room slug
can never be a UUID. A lookup such as "<id>!" does not parse as a UUID,
so it was searched as a slug, found nothing and was treated as an
unregistered room. Its slug, normalized back to the bare id, then got a
token for the registered room without going through the lobby, whenever
unregistered rooms are allowed.
Unregistered rooms whose slug is a UUID now return a 404.
Registered rooms stopped returning is_administrable in v1.25.0, and
nothing reads it for unregistered rooms either: administrable rooms are
detected from their accesses.
The frontend ApiRoom type drops is_administrable as well, which the API
no longer returns.
The room query joins the room, creating it first when missing, and keeps
the response, LiveKit credentials included, in the room cache entry. It
joins again to renew the token once its data is 6 hours old.
Conference and Lobby component were simplified by regrouping custom
hooks, states/refs, memo/callback and mutations.
Room access tokens were returned by GET /rooms/{id}/, with the LiveKit
display name passed as a query parameter. We want to leverage Django's CSRF
checks and avoid displaying usernames in access logs.
Tokens are now only minted by POST /rooms/{id}/join/, which reads the
display name from the request body and the livekit information have been
removed where they were not required.
Joining an unregistered room now builds its livekit block with
generate_livekit_config, which mints the same token as the dict it
replaces.
The serializer and the lobby now share LobbyService.can_bypass_lobby to
decide who may enter directly to centralize the logic.
The trivy scan fails on HIGH vulnerability CVE-2026-119373
found in the Debian 13 base images. As it has no fixed version,
add an exception to trivyignore. Also add expiry date for all
exceptions.
Since LiveKit Egress v1.13, transcripts no longer go through Chrome:
audio-only egresses always use the SDK source, which only starts on the
first audio packet. A muted microphone sends none, so a transcript
started while every microphone is muted stays "starting" until someone
unmutes.
The hint asking participants to turn on their microphone only showed
when no microphone was published at all, as it counted muted ones. Only
count unmuted microphones for transcripts. Screen recordings still run
in Chrome and keep counting every track.
The backend now notifies participants with transcriptionInterrupted or
screenRecordingInterrupted when LiveKit Egress cuts a recording short but
keeps what it recorded so far.
Since LiveKit Egress v1.14, an egress cut short still uploads what it
recorded, but Meet either throws it away or saves it silently:
- a transcript that loses its room connection, once the SDK gave up
reconnecting, ends FAILED with "recording ended early: connection to
room failed": it was flagged as failed and was never transcribed
- an egress drained because its host ran out of CPU ends COMPLETE with
the end reason "CPU exhausted": the truncated file was saved without
telling anyone. If the drain outlasts its grace period, the egress gets
killed and ends FAILED instead, with the same end reason, and with or
without its file uploaded.
A new INTERRUPTED worker event is created :
- the recording is saved and sent to external services as usual, and the
participants of an active recording are told that only part of it was
kept.
- A FAILED egress only counts as interrupted when one of its files
has a location, which egress sets once the upload succeeded. Other
failures stay failures, even with an uploaded file.
When a LiveKit node crashes or Redis stalls, every client loses its signal
connection at the same time. LiveKit's default policy has been jittered
since 2.23.0, but it retries within 500 ms, then again within 450 ms, and
stops after about 45 s of delays, which can be too short for an outage.
Add a reconnection policy with a random exponential backoff. Before its
first attempt, a client waits a random delay up to 1 s by default. That
ceiling doubles after every failed attempt, capped at 10 s, and later
delays fall between half the ceiling and the ceiling. Without that floor,
retries happen at the same time when the service comes back
A client gives up after 3 minutes by default, unless its media still
flows. A client that drops again soon after reconnecting resumes the
backoff one step above where it stopped. The PostHog telemetry now reports how
long a reconnection lasted and when the policy gave up.
When a LiveKit node crashes or Redis stalls, every client loses its signal
connection at the same time and tries to reconnect at once. The frontend
is about to spread these reconnections with a random exponential backoff.
Expose its parameters under livekit.reconnect in the config endpoint to
let each deployment tune them without rebuilding the frontend.
LIVEKIT_RECONNECT_MAX_DELAY_MS defaults to 10 s: with delays drawn between
half of it and all of it, a client retries every 7.5 s on average, close
to the 7 s of LiveKit's default policy in a long outage.
LIVEKIT_RECONNECT_BACKOFF_ENABLED switches back to LiveKit's default
policy.
When a room ended for a reason Conference did not handle, the user was
left on a dead call screen. This happens when the SDK gives up
reconnecting and when the join itself fails. Since livekit-client 2.22.1
it also happens when the WebSocket upgrade times out during the join,
which 2.21.0 reported as a cancellation and sent to the "you left the
meeting" page.
Send these users to the feedback page with a heading of its own: the
connection was lost if the room had connected, the meeting could not be
joined otherwise. Both keep the buttons to rejoin or go home. A join that
never succeeded leaves no call to rate, so the rating is hidden.
A full reconnection removes every remote participant until it succeeds,
which the idle-disconnect timer took for being alone in the room. Only
run that timer while the room is connected, so a long reconnection does
not end with the user sent away.
Release images were building the JS bundles under QEMU for arm64, which
took most of the 10 to 17 minutes spent on each frontend image on CI.
Since the output is mostly static files, we can use the build platform
for the multi stage build steps of frontend images and share them
between architectures.
Room URLs were appearing in search engine results.
The indexing is now restricted to allow only the landing page.
- Emit a separate `home.html` during the Vite build without a `noindex` tag
- Configure Nginx to serve `home.html` exclusively at `/`
- Serve all other routes with a `noindex` meta tag and an
`X-Robots-Tag: noindex` header
LaSuite no longer uses Crisp for support.
Users now reach the team through the La Suite support form. It opens
from the in-room "Contact support" menu item and from the main page
as well
The widget is configured with the widget_script_url, widget_api_url and
widget_channel keys that uses shared values with other LaSuite projects
It won't show if the script URL are not configured.
Fix three vulnerabilities in djangorestframework
- CVE-2026-114136: a compressing codec in the Content-Type charset
lets an unauthenticated body bypass Django's size limit,
causing memory-amplification denial of service
- CVE-2026-114137: TokenProxy.delete() reports success without deleting
the token
- CVE-2026-114138: crafted Accept headers make content negotiation burn
seconds of CPU per request, exhausting workers
Bump the nginx-unprivileged base image to 1.30.6-alpine3.24, which
fixes a heap overflow when using alias with rewrite, along with
CVE-2026-93479, CVE-2026-95101 and CVE-2026-97629.
https://nginx.org/en/security_advisories.html
CVE-2026-111497 is an out-of-bounds write in the brotli decoder,
reachable by decompressing crafted brotli data.
- backend: upgrade Brotli to 1.2.1, used by urllib3 and aiohttp to
decode "Content-Encoding: br" responses
- docker: require brotli-libs >= 1.2.1-r0 in the backend, summary and
frontend images, the nginx base image ships 1.2.0-r1
- frontend: build the Outlook add-in on Node 22.23.4, Node 20 is EOL
and will not get the fix
https://www.strix.ai/cve/CVE-2026-111497
Eight eMudhra intermediate CAs were compromised due to weak private
keys, and the emSign Root CA - G3 / ECC Root CA - G3 roots have been
removed from Mozilla and Chrome trust stores.
Update uv, and bump certifi to 2026.10.19 and require
ca-certificates >= 20261020-r0 in Alpine/Debian images
Node.js has not shipped a fixed bundled root store yet.
The mic shortcut handler treated the value returned by toggle() as a
"did change" flag and skipped the announcement when it was false. In
the room, toggle comes from LiveKit's useTrackToggle, which resolves to
the new enabled state, so turning the mic off returned false and the
screen reader never heard "Microphone turned off".
Make every toggle given to ToggleDevice resolve to the enabled state it
ends in: the camera wrapper now returns LiveKit's result, and the join
screen's toggleTrack, moved to a tested helper, resolves to the initial
state when muting or unmuting fails. ToggleDevice announces only when
that state matches the requested one.
Each request opened and closed its own Postgres connection which added a
big overhead. The goal is to add Django's native psycopg pool behind a new
DB_PSYCOPG_POOL_ENABLED settings with explicit min/max size and keepalive.
Each pool lives in one process, and gunicorn's sync workers serve a single
request at a time which means that a given worker won't use more than one
connection. The default pool minimum size is set to 1, which keeps always
one connection open and the maximum size to 3 as headroom for code touching
the database from an extra thread.
Pooled connections can sit idle for hours, so TCP keepalives are enabled
to keep load balancers and NAT from dropping them silently, and each
connection is health checked on checkout so a broken one is replaced
instead of failing the request.
A timed-out call raised a bare TimeoutError past every
service's TwirpError handler, so callers that recover from
a failed LiveKit call, the permission checks and the
webhook handlers among them, failed instead.
The SDK passes timeout=None to aiohttp on every call, so a
LiveKit that stops answering held a worker until the kernel
gave up. A client middleware now bounds each request, its
answer included, and the client closes the session it is
given, which fixes#1648.
Each recording start writes a Recording row and spins up a recording
worker, so a user or buggy client hammering the endpoint is costly even
when the calls end in a conflict or a worker failure.
Throttle it per authenticated user, mirroring room creation: a short-term
limit of 10/minute absorbs bursts and a daily cap of 100/day bounds
sustained use. Both are configurable via START_RECORDING_THROTTLE_RATES
and START_RECORDING_DAILY_THROTTLE_RATES.
Only starting a recording is throttled, as it is the costly action.
Stopping a recording only releases resources, so it is left unthrottled
and a throttled user can always stop an ongoing recording.
Register MenshenAuthentication as the menshen_token auth method to record
room operations made with a token exchanged through Menshen as an authentication
acting for its user
core/auditing.py both declared the actions that emitters import and
registered the models and authentication classes: it made it impossible
for authentication class to emit an audit event without an import cycle.
Keep the actions and transit_code_target in core/auditing.py and move the
model and authentication method registrations to core/audit_registry.py, the
module the audit app now autodiscovers.
Trace the login flow of the embedded iframe frontend flow.
An application minting a transit code for its delegated user should
logs the transit_code issuance and it's exchange for a user access
token. The transit_code can be identified and with a
SHA-256 to audit eventual replay attacks.
Emit audit events across the recording lifecycle: recording.start,
recording.stop and recording.delete from the API, recording.end when
LiveKit reports the egress ended, recording.transcript.request when the
backend sends a recording to the summary service, and
recording.transcript.report and recording.summary.report when that
service calls back.
recording.end has an unknown outcome when LiveKit reports an egress
status the backend does not map, and recording.transcript.request names
the summary service as service.target.
Small bugfix: AuditViewMixin keeps a copy of a deleted target, since deleting an
instance clears its primary key.
Every ModelAdmin now emits an ECS audit event when an object is created,
changed or deleted, and when a bulk action runs. Actions are templated
as admin.<target>.<verb>, after the model name. A signed-in account
without staff access reaching the admin is recorded as a denied
admin.access.
The wiring is a custom AdminSite installed through AdminConfig.default_site:
it mixes the auditing into every admin class at registration, including the
ones Django declares, so a new ModelAdmin is covered without doing anything.
It hangs off log_addition, log_change, log_deletions and get_actions, which
Django calls in every path, rather than off save_model. Deletions noted by
log_deletions are emitted from delete_model and delete_queryset, once their
outcome is known.
Changed field names are always reported; their values only for the
admin_values each model is registered with, and never for anything that
looks like a secret. An account acted upon is reported as user.target:
the target itself when it is a user, else the account its model
registers as user_target, as the user an access grants a role to. The
emitter derives it for every event, so user.provision names the user it
creates as well.
Describe the audit event shape, the catalogue, how views and other code
emit events, the environment variables configuring them, how the
project registers its actions, models and authentication classes.
Emit audit events on the most exposed authentication surface: every
request to the client-credentials token endpoint, issued or refused, as
application.token.issue + provisional user creation + room create,
update, retrieve and list through the delegated rooms API.
Events identify the application by client id, once its credentials are
verified, and the delegated user by id, OIDC sub and email domain
Add a core.audit package emitting one Elastic Common Schema JSON line per
security-relevant action on a dedicated "audit" logger. Only lasuite.*
and entity.target.raw.* extends on ECS core fields. Each event names
the data stream it belongs to (data_stream.*, event.dataset) and carries
a unique event.id.
The audit logs reads the actor, auth method, tenant and network fields from
the request it is given, else from the request being served, which
AuditLogMiddleware keeps in a context variable so that a service needs
no request argument.
A person is identified by primary key, OIDC sub when the account has one
and email domain, with its privileges as user.roles; an internal peer
calling the backend by service.origin.name. The tenant is the
application client id, else that domain. The client IP is DRF's throttle
ident, which now honours a NUM_PROXIES setting.
The middleware also replaces the inbound X-Request-ID with a fresh id,
unless REQUEST_ID_TRUST_HEADER says the ingress overwrites it, and echoes
it in the response and the gunicorn access log. Audit events carry it as
http.request.id, next to the client's user agent.
An action is defined with a dotted name with its default ECS category
and types, validated at import against the types ECS expects in that
category, so a call site only names what was attempted. A category or
types given to audit.log win over them, and a bare dotted name is
accepted too.
Meet declares all of the events core/auditing.py:
- audit.register lists the fields describing a model as a target, an
ECS entity.target, and its ECS entity type. Every target carries its
model name and primary key, so an unregistered one is still
identified, and a user target its OIDC sub.
- audit.register_auth_method names the lasuite.auth.method of a DRF
authentication class
AuditViewMixin audits every response of the CRUD actions a view maps in
audit_actions, and of the extra actions naming theirs with
@action(audit_action=...), from finalize_response: the outcome, reason
and status come from the response. An exception DRF does not handle is recorded
as an internal error before it propagates. The core.audit app also
records Django's login, failed login and logout signals.
Menshen is La Suite's OAuth 2.0 token exchange server (RFC 8693): another
service exchanges its user's access token for a token targeting Meet, then
calls the external API on behalf of that user. Tokens are validated by
introspection (RFC 7662). Menshen only reports a token as active when Meet
is among its audiences.
Add Menshen auth to the external API. Only activate it for room creation
scope (other scopes may be added later). If a user doesn't exist for the
provided sub, create it, unless the provided email is already associated
to a an existing account.
This PR does not implement fallback to email to automatically create
user when no sub or an invalid sub is passed and email is provided, when
OIDC_FALLBACK_TO_EMAIL_FOR_IDENTIFICATION=True.
Any authenticated participant can now start a recording on trusted
and public rooms. This is essential, requested by many users, and
should encourage people to authenticate on public rooms. It also
means the starter may hold no role on the room, so who gets access
to the recording has to be decided explicitly.
The part open to discussion is the recording access: who should be
owner/admin on the recording. This commit picks the simplest
design:
* The room's owners and administrators get an `OWNER`
`RecordingAccess`: they can retrieve, download, stop and delete
the recording.
* The starter gets a `MEMBER` `RecordingAccess`, unless they already
own or administer the room, in which case they get `OWNER`. A
member can retrieve and download the recording but cannot delete
it: deletion is irreversible and affects everyone with access, so
it stays with the people responsible for the room.
* Everyone with an access, members included, receives the
"recording ready" email.
* Room roles are copied when the recording starts: later changes
to the room's accesses do not affect existing recordings.
The people responsible for a meeting should control its recording
and transcript, and whoever initiated it should be able to access
it on their own. The starter owns the output and is responsible for
sharing it with their colleagues.
Recording members could not retrieve a recording before this
commit. No code path created such accesses until now, but one added
manually through the admin now grants read access.
One constraint is respected: the summary service creates a doc,
and the docs API can only attribute it to a single owner. The
summary therefore goes to the oldest access, and the starter's
access is created before the room owners/admins ones, so the
starter receives it. The lookup does not filter on role, since the
starter may only be a member. The previous lookup took `.first()`
with the default newest-first ordering, which with several owners
would have picked a room admin.
Alternatives considered:
* Granting access to every participant allowed to record, or to
those present when the recording starts: it requires listing
participants from LiveKit and creating an access for each, and
widens access by default, for a need no user has expressed.
* A `creator` attribute on the recording row: it only adds a
migration without adding value to access control, so it was
dropped.
* Restricting access to the link owner even when someone else
starts the recording: some organizations may want this for
security reasons, but it is YAGNI until a few large clients ask
for it.
I think we should contribute to Docs' api to allow sharing the
transcription document with several users.
Before this commit, a room update and a soft deletion racing each other
could leave the room active.
Both `perform_update`s now re-read the room with `SELECT ... FOR UPDATE`
inside a transaction to avoid this race condition.
pylint was pinned below 4.0 because pylint-django used to require it.
pylint-django 2.8.0 accepts pylint 4.
pylint 3.x kept too-many-lines pragma line numbers across modules: a C0302
disable in one file could make the check report on the wrong line
in another and fail depending on how files were spread across jobs.
pylint 4.1 resets them per module.
cf https://github.com/pylint-dev/pylint/pull/11191
Add a ROOM_DELETED_RETENTION_DAYS setting, unset by default. The
purge_inactive_rooms command now also deletes rooms soft deleted more
than that many days ago, whether they are inactive or not. Inactive
soft-deleted rooms are now purged under ROOM_INACTIVITY_DELETION_DAYS
too. As for inactive rooms, a room holding a saved recording that has
not expired yet is kept.
Before this commit:
1. Rooms could be hard-deleted from the internal API (unused by the
frontend).
2. The external API had no way to delete a room.
3. The admin could hard-delete rooms.
What this commit does:
1. Add a `deleted_at` field to Room. `Room.objects` hides soft-deleted
rooms; `all_objects` stays the default manager so deleted rooms keep
reserving their slug and PIN code.
2. Soft delete rooms from the internal API (replacing hard delete) and
the external API (new DELETE, owner only, `rooms:delete` scope). The
LiveKit room is closed and its participants disconnected; if closing
fails, the soft delete is rolled back and an error is returned.
3. Answer 410 Gone on detail operations on a deleted room (after
permission checks). Its can still be read but not modified.
4. Close LiveKit rooms recreated after deletion by participants still
holding a valid token (room_started webhook).
5. Frontend: participants in the call land on the feedback page without
the rejoin button, lobby participants see a "meeting deleted" screen,
and opening the link shows an error instead of recreating the room.
Untouched by this commit:
- The admin still hard-deletes rooms, without closing the LiveKit room,
and lists soft-deleted rooms alongside active ones.
- Purging soft-deleted rooms (hard deletion) is not handled.
- Scheduled deletion is not handled.
Feature flags are read from the analytics backend. Without one,
NoOpAnalytics returned no flag, so "summary-enabled" was always off and
a self-hosted instance could not get automatic summaries without PostHog.
NoOpAnalytics now accepts a static "feature_flags" mapping through the
existing ANALYTICS_BACKEND_SETTINGS, and get_analytics() passes the
settings to it. Nothing changes when no flag is declared.
Fixes#1610
Some containers were being exposed to local networks. Bind them
to localhost, to tighten security. Don't apply this binding to
LiveKit (exposed on 7880), as we access it using 127.0.0.1.nip.io:7880.
Case-insensitive user lookups during token issuance currently take
around 100 ms in production, accounting for roughly a third of the
total 300 ms token request trace.
Add an `UPPER(email)` index on the user table so these lookups can
use the index instead of scanning, cutting the dominant cost on
the token issuance path.
Emails were rendered and sent inline in the LiveKit webhook, so a
slow SMTP server blocked the request and a transient failure lost
the email. Delivery now runs in a worker, with retries and backoff.
The task is generic (template, subject, recipients, context,
language) so other emails can reuse it. The subject is translated
in the worker, and dates are formatted per owner before queuing,
so the arguments stay JSON-serializable.
Without Celery, the task still runs synchronously. Success now
means "queued", not "delivered".
gitlint prints a warning for each ignore-by-title rule because it will switch
its regexes from match (start of the string) to search (anywhere) semantics.
Both ignore-by-title regexes are already anchored with ^, so they match the
same titles either way. Setting regex-style-search opts in to the new behavior
now and removes the warning.
Fixes#96
Add a Recording tab to the settings dialog with two selectors, a
content profile and a resolution, building on the per-recording
encoding API.
The goal is to let an instance default to a cheaper profile and
resolution to save egress CPU, while users who really need a smooth
HD recording can still opt in. Most users only need a recording of
acceptable quality, so the default stays cheap and configuring it is
optional.
Profiles are labelled by content ("Talking heads", "Presentations and
text", ...) rather than by fps or bitrate: users know what they are
about to record, not which encoder settings fit it. Each profile shows
a one-line description of its trade-off, so picking one requires no
technical knowledge.
The panel gets a third row, matching the meeting language row of the
transcript panel. It is a shortcut to the settings, without adding
mental load about how to configure a recording. It reads as a short
statement with a trailing "Change" link, like the informational rows
above it, and stays on a single line in every locale: the statement
truncates if a translation gets too long, the link never wraps.
The selected options are intentionally not shown in the panel.
Several users can access it, and each would see their own local
configuration while the running recording may use another user's
encoding, which would be misleading. The issue is milder in the
settings tab, which is frozen while a recording is starting or in
progress.
The selectors are disabled from the moment a start request is sent
until the recording is saved, for both screen recordings and
transcripts. Settings are sent once, when the recording starts, so
changing them afterwards would have no effect and suggest otherwise.
Choices are persisted in localStorage to keep things simple,
consistent with the meeting language, which is already stored
client-side.
The backend now exposes the available resolutions and profiles, and
their defaults, in the frontend configuration, so the UI follows
whatever the operator configures instead of hardcoding values. Stored
choices are checked against it before being sent: a profile removed by
the operator falls back to the instance default instead of being
rejected by the API. No encoding is sent when per-recording encoding
is disabled, as the backend would reject it. Operator-defined values
without a translation are shown by their raw key.
Alternatives considered:
- Storing the preference in room metadata: it would turn a pre-start,
personal choice into shared editable state, requiring a write
endpoint, permission checks and handling of concurrent edits.
- Notifying participants of the chosen encoding: notifications are
one-off, so users who join or reload after the recording started
would miss them, and the information is noise for most participants.
Next steps:
This is a deliberately small first iteration, meant to deliver most
of the value and evolve with feedback and usage. Possible next steps:
- persist the preferences in the backend instead of localStorage
- store the encoding chosen for a recording on the room (or on a
session, if such a model is introduced) and read it from there
rather than from user settings, so every user sees the same value
React Aria renders the radio and switch inputs visually hidden and
absolutely positioned. Their labels were not positioned, so in the
side panel the inputs were anchored to the aside and did not scroll
with its content. Focusing one made Chrome scroll the aside, which
has no scrollbar, hiding the panel content.
Position the labels so each hidden input stays inside its control.
authenticate and authenticate_header hardcoded "Bearer" instead of
using the token type the backend is configured with. Setting
APPLICATION_JWT_TOKEN_TYPE, ADDONS_TOKEN_TYPE or USER_ACCESS_TOKEN_TYPE
to anything else made every token be ignored, since the Authorization
scheme never matched.
Store the token type on the backend and use it both to match the
Authorization header scheme (case-insensitively) and as the
WWW-Authenticate scheme.
generate_user_access_token was copy-pasted in seven test modules
(user access token authentication, and rooms create, list, retrieve,
update, participants management, rename/toggle and subtitle).
Extract it into core.tests.utils and reuse it everywhere. The shared
version keeps the optional application and claim overrides from the
authentication tests, and reads the token type claim from
USER_ACCESS_TOKEN_TYPE_CLAIM instead of hardcoding it.
The participant id returned by request-entry was the raw LiveKit UUID,
so anyone knowing it could poll the lobby and obtain the admitted
participant's LiveKit token.
Return the id signed with SECRET_KEY under a lobby-specific salt, and
reject unsigned, tampered or foreign-signed ids with a 400. The raw
UUID is still used internally for the cache and LiveKit identity.
Only run the transit_code exchange flow when the app is loaded in an
embedded context (i.e. inside an iframe).
Combined with the CSP rules that will restrict which origins are
allowed to embed the app, this gives us a client-side lever to
control which integrations can actually use this authentication
path.
Bind an accepted lobby entry to the username the participant had at
the moment of acceptance.
This prevents a participant, once accepted, from changing their
display name and reusing the same lobby grant to enter the room
under a different identity.
The lobby system relied on cookies to identify the participant
across the wait/enter cycle, which does not work in an iframe
context where our cookies are dropped.
Simplify the lobby behavior:
* The POST request that enters the lobby now returns the
participant id in the response.
* The frontend passes that id back on subsequent requests to keep a
sticky session while trying to enter the room.
This moves a bit more logic to the frontend but should be a
transparent refactoring, without decreasing the security of the
lobby flow.
Moving off cookie-based authentication surfaced several hard
issues, especially around loading virtual backgrounds: requests
used to be sent with cookies automatically, which trivially
authenticated those loads. With bearer tokens, those requests need
to be authenticated explicitly.
The situation is made harder by the fact that, when the custom
virtual background was introduced, some of the loading was done as
module-level, blocking imports that are not handled by React and
therefore live outside the normal auth flow.
Ship a functional patch to unblock third parties currently waiting
on this integration. The virtual background loading path should
definitely be refactored and simplified in a follow-up.
Wire the frontend to the backend's token exchange flow: when the
expected URL fragment is present, gate the app loading on exchanging
that fragment for a proper access token, which is then used to
interact with the API.
When no such fragment is present, the code path is a no-op and
should have minimal impact on load performance.
We now use `Authorization: Bearer <token>` to authenticate users
from the token exchange flow (used for iframe embeds).
Until now, the `Bearer` scheme was also reused for the alternative
LiveKit authentication, where a client presents its LiveKit token
issued by the backend to prove room membership on actions open to
any room participant. Sharing the scheme between the two flows is
not viable anymore.
Switch the LiveKit token authentication to a dedicated
`Authorization` scheme, so `Bearer` stays reserved for the iframe /
token-exchange flow.
Follow-up: a broader effort should look into harmonizing and
hardening the backend authentication stack of the app.
Some integrators render our videoconference inside an iframe, where
our cookie-based authentication does not work: our cookies are
SameSite=Lax/Strict, so the iframe drops them.
We looked at what Jitsi offers: a shared secret used to sign JWTs
that authenticate users coming from external services. Since we
already expose an external API where third parties authenticate as
a given user, it was simpler for us to add an exchange mechanism on
top of that.
Flow:
* Through the external API, mint a short-lived, single-use exchange
code for a user.
* The third party hands that code to the frontend as a URL fragment.
* The frontend exchanges the code for a longer-lived JWT that can be
used to query the regular API viewsets.
Known limitations and follow-ups:
* At some point it would be nice to shorten the JWT lifetime and
add a refresh mechanism. This will be handled in a follow-up PR
when actually needed.
* CSP rules to control which origins are allowed to embed the app
in an iframe still need to be added.
* This alternative authentication cannot easily be scoped to a
subset of endpoints without adding a lot of complexity, so it is
accepted globally on the API for now.
Load allowed domains once and compare the lowercased email domain
in Python. For restricted applications with an existing user,
this reduces successful token requests from four queries to three.
Keep domain loading after credential validation so rejected
credentials require only the application lookup.
Add endpoint query-count tests for restricted and unrestricted
successes and invalid credentials.
The client id and client secret are auto-generated with high entropy
when an application is created. Allowing them to be edited from the
Django admin let any administrator replace them with a short or
weak value, undermining that guarantee.
Make both fields read-only in the admin so they can only be
regenerated through the intended flow.
Application token authentication currently uses Django's default
password hasher. On our production hardware, token requests take
at least ~500 ms. Authentication happens per user of each
application, so this cost accumulates across frequently used
integrations.
Password hashers deliberately make guessing expensive to protect
human-chosen passwords after a database leak. Our application
secrets are generated server-side using a cryptographically secure
random generator, with a default length of 128 alphanumeric
characters. Guessing these secrets is already computationally
infeasible, making password stretching an unnecessary CPU cost.
Use salted SHA-256 for application secrets while retaining
constant-time comparison, and keep user password hashing
unchanged. Existing secrets migrate after successful verification
without requiring key rotation. Conditional updates prevent
migration from overwriting a concurrent rotation.
Store the new hash in `client_secret_sha256` while preserving the
original Django hash in `client_secret`. New applications populate
both fields, allowing the previous release to authenticate both
existing and newly created applications if we need to roll back.
Once every application has migrated and the rollback window has
closed, complete the migration by removing the legacy verification
code and `client_secret` field.
This assumes securely generated, high-entropy secrets. Deployments
that reduce `APPLICATION_CLIENT_SECRET_LENGTH` or supply
predictable secrets lose the offline guessing protection that the
previous slow hasher provided.
Explain the switch to django-lasuite marketing: removed BREVO_*
and MARKETING_SERVICE_CLASS settings, new LASUITE_MARKETING_BACKEND
and LASUITE_MARKETING_PARAMETERS variables, dummy backend default,
and the now-required Celery worker.
The `Unreleased` section of `UPGRADE.md` was not bumped when v1.33
was released, so follow-up changes landed under the wrong heading.
Reorganize the file so v1.33 is properly closed and a fresh
`Unreleased` section is opened on top.
`make run` no longer starts the metadata collector and the
multi-user transcriber. On my machine, this saves about 400 MB
of RAM and almost 2% CPU, out of the 11% the stack uses.
Start them with `make run-agents` when needed.
Disable `METADATA_COLLECTOR_ENABLED` by default in `common.dist` so
the backend does not dispatch jobs to an agent that is not running,
and document how to enable each agent in `developping_locally.md`.
The summary settings already default to the `redis` host, so
`redis-summary` was unused. Remove it and depend on `redis`.
Pin the summary Celery and task tracker URLs to DB 2 in
`summary.dist` to isolate them from LiveKit and the backend
Celery broker (DB 0) and the Django cache (DB 1).
Use the shared helm `dev-backend` chart to deploy the backend in
the Tilt dev stack, instead of maintaining our own copy.
The goal is to share more common pieces of the dev experience and
dev stack across projects, so we do not maintain N different
versions of similar setups.
The only adjustment needed was to backport Garage into the shared
chart.
Originally started by @rouja.
Improves PaginationControl: clearer structure, keyboard nav, a11y improved.
The main room and the picture-in-picture window share this control.
Ctrl+Shift+G focuses the pagination.
By default, `actions/checkout` saves the job's auth token
(`GITHUB_TOKEN` or the provided PAT) in the local git config so
later steps can run authenticated git commands. That token then
stays on disk for the rest of the job, where it can leak:
* If an artifact upload includes the checkout directory, the token
is packaged with it and anyone with artifact access can extract
it. On public repos that is anyone, and the token can be used
while the job is still running ("ArtiPACKED", flagged by
`zizmor` as `artipacked`).
* Any later step, third-party action, or build dependency can read
the token from the git config, which widens the impact of a
supply-chain compromise.
None of our workflows need authenticated git after checkout, so
disable credential persistence. If a step needs to push in
A separate vitest.config.ts keeps the tests off the build
plugins and the mediapipe version check in vite.config.ts.
make test now runs the frontend tests after the backend ones,
and test-front runs on Node 24, the current LTS.
Add vitest as a dev dependency, a test script that runs panda
codegen first, and a test-front job, so the frontend can carry
unit tests. One test covers normalizeRoomId, a plain function,
so no DOM library comes with it.
The hand button owns the lower-hand timer and the toast, and the
picture-in-picture window draws a second copy of that button, so two
offers go up and dismissing one leaves the other to lower the hand.
Move the watching into a component rendered once beside the room.
Surface the page's loading state to assistive technology, so screen
readers can announce that the page is still loading instead of
reading a partially rendered state as if it were complete.
When the user is updated, their lists on Brevo are overwritten with
the new value: this removes lists set by other products.
Switch to the common lib implementation from `django-lasuite`,
which manages this correctly.
This change was initially proposed by @qbey, but at the time our
deployment did not have a Celery worker running alongside the
backend. Since then, a Celery worker has been deployed, so the
switch to the common lib approach is now safe to adopt.
Highlight in the connection test when the user is connecting
through a TURN relay, especially over TLS or TCP. This usually
indicates that some network configuration is required on their
side, and gives them a concrete signal to pass to their IT team.
Suggested by a technical user, this is a first step toward making
users more autonomous when troubleshooting access to the tool.
Follow-up: show a similar warning in-product when we detect a
mid-meeting fallback to TURN/TLS. A one-time hint for first-time
users would likely be enough.
The home page only offers meeting creation to authenticated users, while the
backend already serves ephemeral rooms to anonymous visitors when
ALLOW_UNREGISTERED_ROOMS is enabled (the flag is checked in the room retrieve
view, not in the create one).
Expose the flag in the frontend configuration and, when it is on, show the
existing "Create a meeting" button to signed-out visitors. It navigates to a
freshly generated room id rather than calling POST /rooms/, which stays
reserved for registered rooms and for authenticated users.
The invite dialog opens for such a creator when the room is unregistered
(null id) and the navigation carries `create`. The `mode` computed in Room
is left untouched, so permissions and the join screen behave as before.
The home buttons row now wraps: signed-out visitors can see three controls
(create, join, login), and at the xsm breakpoint the fixed-width ProConnect
button leaves too little room for the other two.
A React Aria overlay is rendered at `top: 0; left: 0` until
`useOverlayPosition` computes its position, and `data-placement` is
only set once that succeeds. When the pointer moves quickly between
triggers, a closing tooltip can mount for its exit animation without
ever being positioned, React Aria does not retry, so it stays
stuck in the top-left corner.
Hide tooltips until they have a `data-placement` set, so unpositioned
tooltips never flash in the corner. Use `visibility` rather than
`display: none`, so the element stays measurable for the positioning
pass.
Replace the deprecated `enable_tracing=True` with `traces_sample_rate`,
read from a new `sentry_traces_sample_rate` setting (default 0.1,
validated to the 0.0–1.0 range).
Previously, tracing sampled 100% of transactions, which is costly and
unnecessary in production. The rate can now be tuned per environment
without a code change.
Sentry attaches stack frame locals to its events. When storing a
transcript in S3 failed, the full transcript held in `data` and
`transcript` was sent to Sentry.
Keep locals for debugging, but scrub variables and nested dict keys
known to hold transcripts, summaries, LLM prompts, participants'
personal data or pre-signed URLs. Share the Sentry init between the
API and the Celery worker, and never send request bodies.
Since boto3/botocore 1.36, the S3 client computes CRC32 checksums on
uploads by default (request_checksum_calculation="when_supported").
PutObject requests are then sent with aws-chunked encoding, a trailing
x-amz-checksum-crc32 header and a STREAMING-UNSIGNED-PAYLOAD-TRAILER
content hash.
Our production storage (S3NS, storage.s3nsapis.fr) is built on Google
Cloud Storage and exposes it through GCS's S3-compatible XML API, which
does not support these flexible checksums. It rejects the request with
a 403 SignatureDoesNotMatch ("Invalid argument"), so storing transcripts
failed in the Celery worker. Garage, used locally, supports them, which
is why the issue only appeared in production after switching the
client to boto3.
Add aws_request_checksum_calculation and
aws_response_checksum_validation settings, defaulting to
"when_required", and pass them to the botocore Config of the summary S3
client and the backend S3 client. Checksums are then only sent for
operations that require them, restoring the pre-1.36 behavior.
The nginx-unprivileged:1.30.4-alpine3.24 base image ships
pcre2 10.48-r0, which is affected by CVE-2026-103111 (HIGH,
out-of-bounds write via crafted regular expression). No newer
base image tag is available yet.
Upgrade pcre2 from the Alpine v3.24 repository with a minimum
version constraint (>=10.49-r0) so the build fails instead of
silently shipping a vulnerable version if the fix is unavailable.
Recordings were failing to download in the local stack because of
several small issues stacked together:
* nginx: add a `/media/recordings/` location that authorizes
against `recordings/media-auth/`. Before, every media request
went to `files/media-auth/`, which returned 403 for recording
paths.
* nginx: call `proxy_hide_header Content-Disposition` before
`add_header Content-Disposition "attachment"`. Garage stored the
header as `inline`, which combined with nginx's value into
`inline, attachment` and broke browser downloads.
* frontend: `mediaUrl()` now uses the frontend origin, so
recording links go through the Vite `/media` proxy instead of
hitting Django directly on `:8071`.
* frontend: include the file extension in the download filename.
co-author: cameldev
The devstack was pulling the full `node:22` image, around 1.6 GB.
Switch to `node:22-alpine`, which is much smaller, to speed up the
devstack bootstrap time and reduce disk usage.
The devstack was pulling two different versions of the Redis image.
Align everything on a single version to save a few MB of network
bandwidth when bootstrapping the stack.
Late-night minor optimization.
Add new options to query start-start recording API. A resolution
("540p", "720p", "1080p") and a profile ("talking_heads", "text", "mixed")
are resolved to provide a width, height, fps and bitrate which
are passed on to the encoder. Using profiles allows for some flexibility
on quality if necessary without changing front facing user config.
Co-authored-by: sarthakbahal <sarthakbahal.45@gmail.com>
The python:3.14.7-slim base image ships libpcre2-8-0
10.46-1~deb13u2, which is affected by CVE-2026-103111 (HIGH):
an out-of-bounds write triggered by a crafted regular expression.
Explicitly install libpcre2-8-0 in the base stage so apt pulls
the patched 10.46-1~deb13u3 from trixie-security. All stages
(builder, development, production) inherit the fix.
This line can be dropped once an upstream python slim image
ships the patched package.
Address the following HIGH severity CVEs reported by Trivy on the
backend image:
* Django 5.2.16 → 5.2.17
- CVE-2026-15307 — remote code execution via GeoDjango spatial
lookups.
* urllib3 2.7.0 → 2.8.0
- CVE-2026-97687 — traffic interception via HTTPS proxy TLS
configuration override.
- CVE-2026-97689 — denial of service via unbounded memory
allocation in the chunk parser.
The trivy scan fails on HIGH vulnerabilities found in the Debian 13
base images (util-linux, acl, ncurses, systemd and perl-base). None
of them has a fixed version available yet, so there is nothing we
can upgrade to clear them.
List these CVEs in a shared .github/.trivyignore and pass it to
every image scan, so the scan stays blocking for any new HIGH or
CRITICAL vulnerability. Remove the entries once Debian ships a fix.
Instead of referencing the shared CI repo on `main`, pin it to the
initial tagged version `v0.0.1`, so the CI behavior is stable and
does not silently change when the shared repo is updated.
Get the shellcheck CI job to pass again by addressing the issues it
flagged across our shell scripts.
Note: I am not 100% sure of every fix applied here. Reviewers should
feel free to challenge specific changes and suggest better ones
where relevant.
Get the spellcheck CI job to pass again by:
* Fixing the actual spelling issues it caught in the project.
* Excluding generated files from the scan, since they are not
written by us.
* Excluding translation files, whose content is not necessarily in
English and would trigger false positives.
Wire Menshen into the CI to scan the GitHub Actions we use and flag
vulnerable ones, following the same approach as other projects that
recently adopted it.
Note: I am not fully sure about the current setup. Reviewers should
feel free to adjust the configuration or the integration point as
they see fit.
The Crowdin workflows were not used by the project and had turned
into dead CI code.
Remove them to reduce noise and keep the CI configuration limited
to what is actually running.
First iteration of a migration toward a centralized repository
containing our shared CI logic.
Goals:
* Manage GitHub Actions version upgrades in one place.
* Make it easier to audit what actually runs in CI from a security
perspective.
* Avoid duplicating CI logic across projects and having each
repository slowly diverge over time.
* Centralize as many of our custom GitHub Actions as possible,
including some that still live in the old `numerique-gouv`
organization.
* Centralize the Renovate configuration alongside the workflows.
Inspired by the Accounts project, which recently simplified and
reorganized its CI setup.
This PR starts moving meet's CI to the shared repository so we can
validate the approach on a real project. For now, reusable
workflows are pinned to `main`; once we agree on the structure and
content of the central repository, they should be pinned to a
specific commit SHA instead.
Address the following CVEs in util-linux, reported by Cyberwatch on
the agents image. The python:3.14.6-slim tag is no longer rebuilt
and still ships util-linux 2.41-5. Bump the base image to
python:3.14.7-slim, which ships the patched 2.41.5-0+deb13u1
(DSA-6442-1), to cover them all:
* CVE-2026-53612 (7.0) — TOCTOU in mount post-mount ownership/mode
changes.
* CVE-2026-53613 (7.0) — TOCTOU in mount via ancestor directory swap.
* CVE-2026-53614 (7.0) — SUID mount(8) nosuid/noexec bypass via
`LIBMOUNT_FORCE_MOUNT2`.
* CVE-2026-13595 (5.3) — flaw in the libblkid library.
* CVE-2026-27456 (4.7) — TOCTOU in SUID mount(8) when setting up
loop devices.
Permissions on the recording panel checkboxes were not properly
enforced. A user with only partial access to some recording modes
could still tick a mode's checkbox and, for example, launch a
transcription from the recording panel even though they were not
authorized to.
Gate each checkbox on the user's actual permissions so that only
authorized modes can be started from the panel.
The per-minute room creation throttle absorbs bursts but does not stop
a compromised account from steadily creating rooms over hours or days.
Add RoomCreationDailyUserRateThrottle, a per-user throttle with its own
"room_creation_daily" scope, applied to room creation only alongside
the existing short-term throttle. It defaults to 1000 rooms per day and
is configurable via ROOM_CREATION_DAILY_THROTTLE_RATES.
Tests use a controllable clock and patch rates with monkeypatch so they
are restored after each test.
Add throttling on the endpoint used to generate meeting links, so a
compromised authenticated account cannot silently generate thousands
of links without hitting any suspicious errors or alerts.
The limits are set high enough not to affect legitimate usage, while
capping the damage a leaked account can do.
The CHANGELOG entries ended up in the wrong order after a rebase
performed while merging a recent PR.
Restore the intended organization of the entries so the file reads
correctly again.
Bump PyJWT from 2.13.0 to 2.14.0 to address the following CVEs
reported by Trivy:
* CVE-2026-102268 (CRITICAL)
* CVE-2026-102266 (HIGH) — authentication bypass via empty HMAC
key acceptance.
* CVE-2026-102267 (HIGH) — verification key substitution via
unvalidated JWKS redirects.
* CVE-2026-102271 (HIGH) — authentication bypass via acceptance
of DER public keys as HMAC.
* CVE-2026-102272 (HIGH) — token forgery via improper Unicode
byte-order mark handling.
* CVE-2026-102273 (HIGH) — token forgery via acceptance of
public JWK containers as HMAC.
Since posthog-js 1.356.0, feature flags are reloaded every 5
minutes by default while the tab is visible. Meet sessions are
long-lived visible tabs, so this multiplied the number of `/flags`
requests we send.
Restore the pre-1.356.0 behavior: only (re)load flags on init and
on identity
Add a `purge_inactive_rooms` management command which permanently
deletes the rooms that were not started for
`ROOM_INACTIVITY_DELETION_DAYS` days. Rooms never started are aged from
their creation date.
Important points :
- The setting is unset by default, which disables the purge.
- Rooms holding a recording their users may still access are kept,
- It refuses to run while unregistered rooms are allowed, as the link
of a purged room would turn into a public unregistered room,
Rooms pile up in database with no way to tell the ones still in use from
the abandoned ones. Record on the room the last time LiveKit reported it
as started, in a new `last_started_at` field which stays NULL until the
first time the room is started on a LiveKit node.
The migration stamps existing rooms with the migration time to avoid
deleting preexisting rooms.
Garage has no web console, so inspecting recordings, transcripts and
summaries locally meant writing aws-cli commands by hand.
For each of recordings, transcripts and summaries:
- `make <folder>-list` lists the files from the most recent, and
- `make <folder>-download-latest` downloads the latest one into
data/<folder>.
Only files with the folder's expected extensions are kept, so the Egress
manifests stored next to recordings are skipped.
The media ingresses and their ExternalName services defaulted to the
MinIO service of the development stack, which is now Garage.
Self-hosted relying on these defaults must set serviceMedia*.host and the
upstream-vhost annotations explicitly, see UPGRADE.md.
The development stacks now run Garage instead of MinIO which is
deprecated.
Garage is a bit stricter than MinIO:
- key IDs and secrets must be at least 8 and 16 characters long
- requests must be signed for its region, so every development env now sets
AWS_S3_REGION_NAME=local;
- cross-origin requests are denied unless the bucket CORS rules allow
them, so a one-shot aws-cli container allows the frontend origin to
upload files straight to the bucket.
Switch from the minio client to the boto3 python client, in the metadata
collector agent and the summary service. AWS_S3_REGION_NAME is passed
as-is to boto3, the region is not looked up from the bucket.
In the dev stack, configure the backend to call the summary service
using its v2 API by default.
Also strip the trailing `/` from the task endpoint, which was
triggering a redirect on every call.
The organization now provides default GitHub templates for pull
requests and issues at the org level, so individual repositories no
longer need to duplicate the same Markdown files.
Delete the local templates so this project uses the organization
defaults, reducing the amount of code we maintain and centralizing
the practice across repositories.
Address the following CVEs reported by Trivy on the LiveKit agent
image against `libssl3t64` 3.5.7-1~deb13u2:
* CVE-2026-63073 — CRITICAL (CVSS 9.8)
* CVE-2026-63072
Bump `libssl3t64` to the patched version to pick up both fixes.
Address the following MEDIUM severity CVEs reported against
`djangorestframework` 3.17.1:
* CVE-2026-73228 (CVSS 5.3)
* CVE-2026-73229 (CVSS 4.3)
Bump `djangorestframework` to the patched version to pick up the
fixes.
Bump `libexpat` from 2.8.4-r0 to 2.8.5-r0 to address the following
HIGH severity CVE, reported by Trivy on the frontend image
(alpine 3.24.1):
* CVE-2026-93990 — expat: XML injection via malformed UTF-16
input.
https://avd.aquasec.com/nvd/cve-2026-93990
Most call sites of `update_metadata` already wrap the call in a
try/except that logs the failure at info level.
Remove the warning log inside `update_metadata` itself to avoid
redundant logs, without losing any information.
`EGRESS_ABORTED` and `EGRESS_FAILED` events were previously ignored, leaving
recordings indefinitely in `ACTIVE` state and potentially blocking subsequent
recordings with 409 errors. Add handling and logging for failed and aborted
egresses, discarding failed recordings while preserving the existing behavior
for savable recordings.
Rename `handle_complete` to `handle_savable` to reflect that it handles both
`EGRESS_COMPLETE` and `EGRESS_LIMIT_REACHED`.
Slight refactor to separate LiveKit event handling from recording concerns as
part of a general separation concern to allow for future SFU swapping.
NB:
- FAILED recordings are currently discarded although exploitable media files
may exist
- There is a theoretical hole: if stop observes EGRESS_FAILED before the
egress_ended webhook is processed, the recording is immediately marked as
FAILED. Since only ACTIVE and STOPPED recordings are savable, the webhook
then skips the failure notification and LiveKit error log. In that rare race
condition, participants may therefore not see the failure toast. We accept
this trade-off for now, as this should be very infrequent.
- Another theoretical hole: There is a short race window where the user
clicks stop while the limit-reached status is being processed. Since the user
explicitly requested the stop, we consider skipping the limit notification
acceptable and do not handle this case.
fix(recording): log aborted worker events at info level
When a user starts a recording or a transcription while no track is
published yet, the recording stays in a "starting" state until an
appropriate track is available.
Show an explicit message on start explaining that the recording
will remain in "starting" state until a track of the required type
is published. The expected track type depends on the recording
type (audio-only vs. audio + video).
This situation was generating a lot of support requests, with users
asking why the recording did not actually start.
`VideoResolutionSubscription` applied the saved reception resolution on
`RoomEvent.TrackPublished`. livekit-client does not raise that event for cameras
that were already sending when the local participant joined, so a user who had
chosen Low definition still received High definition from everyone already in
the meeting, and Low definition only from whoever joined after them. Nothing in
the UI showed the discrepancy: the setting kept displaying Low definition.
Apply the preference to the publications we already know about when the effect
runs, and keep listening on `TrackPublished` — which stays the earliest point to
cap a camera that starts after us — plus `TrackSubscribed`, which is the first
event raised for the cameras that were already sending.
That initial pass also covers a change of preference mid-call, which
`VideoTab.updateExistingRemoteVideoQuality` was doing separately. Removed, it is
now the same code path for joining and for changing the setting.
The three entry points overlap on purpose; the `publication.videoQuality` guard
makes the repeats free. It reads as High definition when nothing was ever
requested, so the default case costs no signal round trip either.
Fixes#1606.
`DockerflowMiddleware` serves the endpoints `/__heartbeat__`, `/__lbheartbeat__
that we use for the kubernetes probes. Sitting at the bottom of
MIDDLEWARE, every Kubernetes probe traversed all middlewares which is not
efficient.
The backend and summary probes had the two Dockerflow endpoints the wrong way
round. `/__lbheartbeat__` returns an unconditional 200 as soon as the server is
up and touches no dependency, while `/__heartbeat__` runs the Dockerflow checks
and answers 500 when one of them errors.
Wired as they were, a database error made `/__heartbeat__` fail on every
backend pod at once, restarting them all. Since a restart cannot fix a
database outage, it's better to use these check on the readiness probe
and start routing traffic when the database is reachable.
- Probe liveness on `/__lbheartbeat__` and readiness on `/__heartbeat__`
- Add a startup probe on `/__lbheartbeat__`, polled every 5s with a
`failureThreshold` of 12, leaving the pod a minute to boot
- Drop `initialDelaySeconds` from liveness and readiness, now that the startup
probe holds them off until the server answers
- Set `timeoutSeconds` to 5s on every probe, up from the 1s Kubernetes default
- Set the readiness `failureThreshold` to 3
The `meet.probes.abstract` helper was missing `periodSeconds` block which means
Kubernetes fell back to its 10s default instead of the chart value.
It's now possible to configure the `failureThreshold` and `successThreshold`.
Builds through the Docker API of the Podman service receive none of the
proxy variables in their RUN steps and fail systematically because of
the proxy rejection. Plain HTTP connections are also rejected by the
proxy with a HTTP 405 method error.
- Passes http_proxy, https_proxy and no_proxy from the shell as build args
- Make the Debian mirror of the agents image a build argument and
override it for bureautix to force https usage
Add the minimal requirements to build and run the project locally on NixOS
- `devenv update` update devenv using NixOS 26.05 stable repositories
- `devenv shell` activates the devenv
- `devenv --profile <profile>` shell uses additional packages when
activated (profile=agent|summary|k8s)
Pin the LiveKit rtc section: the browser reaches the server through
podman's published ports on loopback while egress and the agents reach
it over the podman network, and those two have no address in common.
`advertise_internal_ip` keeps the container's own interface address as a
host candidate alongside the node_ip one, so LiveKit offers both and ICE
picks whichever works. Without it egress only ever sees 127.0.0.1, which
is the egress container itself, and its peer connection timeouts
`use_external_ip` is turned off since it would advertise the STUN-discovered
public IP, which no local peer can hairpin to.
Rootless podman maps container UID 0 to the host user and every other
container UID to a subuid that owns nothing in the worktree, so the usual
DOCKER_USER=$(id -u):$(id -g) makes every bind mount effectively
read-only.
Add a compose.podman.yml to override the compose.yml and set
`userns_mode: keep-id` in order to map the host user to the same UID and UID
inside the container. The merge of the docker compose file is now done with
the COMPOSE_FILE environment variable
The waiting room has its own sound in notifications.mp3, and nothing could play
it: the sprite is named "waiting" while triggerNotificationSound passes a
NotificationType, and howler returns without playing when the sprite id is
unknown. ParticipantWaiting was also absent from the sound settings, so the
check on the store would have refused it first. The toast borrowed the
participant joined sound instead.
The sound was tied to the waiting list going from empty to non empty, so a
second person arriving while someone was still waiting was silent, which is the
case the issue describes.
Name the sprite after the notification type, register the type in the settings,
and sound every arrival, detected on the participant ids so that an admission
and an arrival between two refreshes do not cancel each other out.
closes#1705
get_release() read the version from a version.json file but nothing generates
it during the CI Docker image build, therefore the release reported to Sentry
was always "NA".
Read the version from pyproject.toml instead, which is bumped at each
release and copied into the image.
Incorrect file permissions in the frontend Docker image caused
problems when running the project, and were surfaced by @briquet
while setting it up with Podman.
Adjust the ownership and permissions applied during the build so
the image works cleanly under Docker and Podman alike.
Docker Hub now denies anonymous pulls of minio/minio (pull access
denied), which fails the test-back job for every pull request. The
quay.io/minio/minio mirror remains publicly available, accepts the
same environment variables, and the container lookup in the Configure
MinIO step still matches the image name.
thx @mmaudet
Load the Crisp JavaScript module only once the frontend is idle,
instead of during the initial page load.
Keeps the critical path lighter and prevents Crisp from competing
with the app's own bootstrap for network and CPU on slow devices.
The resource server backend returned any user matching the token's
`sub` claim without checking `User.is_active`. The upstream lasuite
backend only validates the token's introspection `active` claim, so a
deactivated Django account kept API access until its token expired.
Raise `SuspiciousOperation` in `get_or_create_user` when the user is
inactive, which the authentication class turns into a 401, consistent
with `BaseJWTAuthentication`. Add unit and end-to-end tests.
Add a configuration knob for the trace sampling rate, so we can
enable tracing on middleware and cache spans when debugging slow
requests in production.
Sampling is set to 0 by default, so tracing stays fully off unless
explicitly enabled.
Attach the LiveKit SIDs (room and participant) to the connection
analytics event.
Makes it easier to debug problematic sessions and to correlate a
room session with the corresponding LiveKit logs.
Around 0.76% of incoming LiveKit webhooks were being flagged as
unprocessable and returned a 422, even though LiveKit was sending
legitimate data — just with event types we do not handle. This
inflated error metrics and made real webhook issues harder to spot.
Return a 200 for these webhooks instead. When a new, unhandled
event type shows up, log a warning so we can decide whether it is
worth adding explicit handling.
The request summary info logs were spamming the log stream, making
around 46% of the total volume, without carrying any exploitable
information.
Silence them so the remaining logs are easier to explore and cheaper
to store; roughly halves the overall log volume.
On a busy morning, `/me` alone produced 72k warning logs — 97% of
all warnings. They all come from anonymous requests to `/me`
without credentials, which is normal: `/me` is how the app
determines the current auth status.
These warnings carry no diagnostic value on this endpoint, so
silence them there to cut down on log volume.
AUTHENTICATED_PARTICIPANTS_CAN_EDIT_DISPLAY_NAME was only enforced at
LiveKit token generation and by hiding the name field in the frontend.
The `rooms/{id}/rename/` endpoint never checked it, so any authenticated
user with a valid room token could rename themselves via the API even
when the self-hoster had disabled it.
Return 403 from the rename action for authenticated users when the
setting is disabled, mirroring the `can_edit` rule in
`core.utils.generate_token`. Anonymous participants are unaffected, as
they have no account name to fall back on.
Add tests covering the disabled/enabled cases for authenticated users
and the anonymous exception.
When the lobby is disabled mid-meeting (e.g. the room is switched to
public), the waiting participants list stopped being refetched, so
the previously cached list stayed visible with stale data.
Trigger a refetch in that case as well, so the list is cleared and
the moderator UI no longer shows waiting participants for a lobby
that is no longer active.
Highlighted by a suggestion from @florent, the waiting participant
list was not sorted, so moderators could see participants in an
arbitrary order.
Add an explicit `entered_at` attribute on each waiting participant,
so the list can be sorted by arrival time. Participants are now
shown in a stable order of arrival, both across polls and across
moderators.
Compute the position of the login hint at render time so it is
always displayed close to the login button, regardless of the
button's placement or the current viewport size.
When a moderator accepts or rejects a lobby entry that no longer
exists, emit a tracking event so we can measure how often it
happens.
This signal will help tune the lobby polling interval: too many
"not found" events means the moderator side is working from a stale
list. Keep raising the error to the client on top of tracking it,
so the frontend still surfaces the issue (its current handling of
this case is still incomplete).
Include the time taken by each request in the Gunicorn worker
access logs, so we can spot slow endpoints and correlate latency
patterns directly from the logs.
The previous presence cache lookup keyed off a scan over the whole
cache, so its cost was O(db_size) rather than O(room_size).
Combined with the recent switch to cursor-based `SCAN` at an
inappropriate page size, this caused a lot of Redis round-trips and
noticeably slowed down the backend pods under load.
Refactor the presence cache to keep a per-room set of all its
participant keys. Lookups now iterate that set instead of scanning
the whole database.
Complexity is now bounded by room size, not database size, which
should restore the backend performance to its previous levels while
keeping the lobby behavior unchanged.
The previous lobby lookup keyed off a scan over the whole cache, so
its cost was O(db_size) rather than O(room_size). Combined with the
recent switch to cursor-based `SCAN` at an inappropriate page size,
this caused a lot of Redis round-trips and noticeably slowed down
the backend pods under load.
Refactor the lobby storage to keep a per-room set of all its lobby
keys. Lookups now iterate that set instead of scanning the whole
database:
* Membership in the set acts as a memory of who is supposedly in
the lobby for a given room.
* Individual keys are then read to check who is actually still
waiting or accepted.
Complexity is now bounded by room size, not database size, which
should restore the backend performance to its previous levels while
keeping the lobby behavior unchanged.
The `/me` endpoint was called without a trailing slash, so every
request was going through a 301 redirect before hitting the actual
endpoint.
This endpoint is called by every user at least once per session, so
based on the logs, avoiding the redirect should cut the volume of
requests hitting it by around 10%.
Increase the polling interval used by the lobby feature, on both the
waiting participant side and the moderator side.
The goal is to reduce the volume of requests the lobby generates,
trading a bit of data freshness for better performance.
It will de facto reduce pressure on the backend.
We will observe the impact in production, and revisit these
intervals if the delays turn out to be too aggressive.
Bump the frontend base image to `1.30.4-alpine3.24`, which picks up
fixes for the CVEs listed below and lets us drop the individual
dependency pins that were only there to address earlier known CVEs.
Address the following HIGH severity CVEs in libuuid / util-linux,
reported by Trivy. Bumping to 2.41.6-r1 (bundled in the new base
image) covers all of them:
* CVE-2026-53612 — TOCTOU in mount post-mount ownership/mode
changes.
* CVE-2026-53613 — TOCTOU in mount via ancestor directory swap.
* CVE-2026-53614 — SUID mount(8) nosuid/noexec bypass via
LIBMOUNT_FORCE_MOUNT2.
* CVE-2026-76642 — failed external mount helper still runs
privileged X-mount post-hooks.
* CVE-2026-78408 — nsenter --join-cgroup leaks root cgroup
migration authority (fixed in 2.41.6-r1).
* CVE-2026-78410 — restricted bind mounts do not pin the source,
allowing X-mount.owner/group/mode escalation.
A minor layout regression appeared when switching to the Marianne
font: the feedback buttons wrapped onto two lines instead of
staying on one.
Adjust the layout so the buttons stay on a single line regardless
of the font in use.
The previous implementation centered initials by measuring `<text>`
with `getBBox()`, which returns the font's advance-width by
ascent-to-descent band, not the ink of the glyphs. On fonts with an
asymmetric band, initials rendered off-center. Marianne is a
particularly clear case: ascent 1131 / descent 256 puts the band
center 87.5/1000 above the caps' optical center, so every avatar
sat ~4.6 viewBox units (~1.5–1.8 px) too low. A follow-up rewrite
using canvas `actualBoundingBox*` metrics fixed it but pulled in a
runtime measurement rig (shared canvas, ref, state, layout effect,
`fonts.load` + `loadingdone`, plus combining-mark stripping) just
to place two uppercase letters.
Since initials are always uppercased, optical centering has a
closed form: `baseline = center + capHeight/2`. Real-world text
fonts have cap heights in a narrow ~0.66–0.73 em band (Marianne is
0.70), so:
translateY(calc(var(--avatar-cap-height, 0.7) * 0.5em))
is exact for the stock font and within ~0.4 px for any plausible
replacement. No JS, SSR-safe, no first-paint jump, no font-loading
race. Accents float above the cap box instead of dragging the
letter down.
The single font-dependent number remaining (cap height) is exposed
as a CSS variable, so self-hosters overriding the font can override
it next to the font itself, or leave the default. Once
`text-box: trim-both cap alphabetic` ships broadly, even the
variable can go.
Following the re-rendering optimization refactoring, the automatic
lower-hand feature broke: the way `isSpeaking` was read no longer
made sense once we limited how often components in the app
re-render.
Fix the detection so the raised hand is again lowered automatically
when the participant starts speaking, without relying on frequent
re-renders.
The user `sub` field was rejecting some ASCII characters that are
actually valid according to the OIDC spec.
Loosen the validation to accept the full ASCII range except control
characters, so the field is compliant with the RFC and works with
any spec-compliant identity provider.
Based on the Stack Overflow discussion in question 279832.
Closes#1609.
The compose stack referenced livekit/livekit-server without a tag,
which resolved to :latest. Docker doesn't re-resolve a mutable tag it
already holds locally, so images earlier than v1.12.0 crashed
on startup:
could not parse config: yaml: unmarshal errors:
line 20: field allow_restricted_peer_cidrs not found in
type config.TURNConfig
LiveKit parses its config in strict mode, and allow_restricted_peer_cidrs
only exists since v1.12.0. The TURN block added in bd81c994 therefore
carried an minimum version which was not reflected.
The Tilt/Helm stack builds its own image from docker/livekit/Dockerfile
to inject the mkcert root CA, and that was still based on v1.9.4. Bump
it to the same version so both dev stacks run the same server.
Adds support for serving media behind Traefik, which currently cannot work
at all.
The media-auth subrequest views read the original request URL from a
hardcoded HTTP_X_ORIGINAL_URL header. That header is an nginx-ingress
convention. Traefik's ForwardAuth middleware sends X-Forwarded-Uri instead
and has no mechanism to emit X-Original-URL, so behind Traefik every
recording download and file attachment is rejected with a bare 403 --
indistinguishable from a legitimate permission denial, which makes it
painful to diagnose.
Add MEDIA_AUTH_ORIGINAL_URL_HEADER, defaulting to HTTP_X_ORIGINAL_URL so
existing nginx-ingress deployments are unaffected. Traefik deployments set
it to HTTP_X_FORWARDED_URI. It is used in both places that resolve the
header: RecordingViewSet._auth_get_original_url and the file attachment
_authorize_subrequest. The log message on a missing header now names the
header actually expected, which is what makes the failure diagnosable.
This mirrors the setting the sibling Docs project already exposes
(suitenumerique/docs, MEDIA_AUTH_ORIGINAL_URL_HEADER) for the same reason.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The sending resolution selector stopped at 720p while `VideoPresets` already
exposes `h1080` (1920x1080), so publishers on a good uplink could not make use
of the capacity they had. Add "Very high definition (1080p)" above the existing
entries, translated in the five supported locales.
The default stays `h720`, so nothing changes unless a user goes and picks the
new entry. Being explicit about what that costs, since 1080p roughly doubles a
publisher's uplink: this is a per-user choice, and an instance operator has no
way today to decline it. Whether that warrants a server-side setting alongside
the existing `ApiConfig` flags is a call for maintainers — happy to add one if
you want it, rather than change the API contract unasked in a frontend PR.
While here, make the option list harder to get wrong. Resolutions now come from
a single `VIDEO_RESOLUTIONS` tuple that `VideoResolution` derives from, the
selector items are built by mapping over it against a
`Record<VideoResolution, string>` of labels — so a resolution cannot be added
to one and forgotten in the other — and a persisted value that is not in the
tuple falls back to `h720` instead of reaching `VideoPresets[...]` as
`undefined`, since `loadUserChoices` spreads localStorage without validating
it.
Known limitation, unchanged by this patch: `restartTrack` passes the resolution
as an `ideal` constraint, so a camera that cannot reach the selected height
degrades silently. That is already true of 720p on a 480p webcam; 1080p is the
first step where the gap is the common case rather than the edge one.
`handleVideoResolutionChange` did all of its work inside `if (videoTrack)`,
including `saveVideoPublishResolution`. With the camera off there is no camera
publication, so choosing a resolution did nothing at all: it was neither applied
nor recorded, while the selector went on showing the value the user had just
picked. Turning the camera back on then published at the old resolution, and so
did the next session.
Found on a self-hosted instance: a user set the sending resolution with the
camera off, turned it back on, and the publisher kept sending 720p. Nothing in
the UI suggested the choice had been dropped.
Persist the choice first and unconditionally, then restart the track only when
there is one to restart.
Persisting alone is not enough within a session. `roomOptions` is only read by
`new Room(...)`, so a store update never reaches a room that is already built.
Sync the VideoDeviceControl with the userChoiesStore resolution, as we did for
the device id and the processor configuration.
The early return is the honest shape here: with no live track there is nothing
to await, and the defaults above already cover what happens next.
Introduce an in-app devtool that monitors WebRTC statistics in
real time and lets developers simulate various network scenarios,
including constraining the uplink and downlink bandwidth.
Makes it much easier to reproduce and investigate connectivity or
quality issues locally without depending on external tools.
The code was AI generated, and might contain some smell.
It's only enabled in dev, and not included in the production
build. Feel free to enhance it as needed.
Wire a TURN server into the local LiveKit server used by the dev
stack, so ICE negotiation has more candidate types available during
local testing.
Makes it easier to reproduce connectivity scenarios that would
otherwise only show up on stricter networks in production.
The summary `uv.lock` had not been updated in the last few releases.
The CI is now refactored to use `uv sync --locked`, which fails when
the lockfile is out of sync with `pyproject.toml`.
Regenerate the lockfile so `uv sync --locked` passes again.
The rule joins the list it fetches into an alternation, so any regex character
in a gitmoji entry would change what the pattern matches. re.escape treats each
one as a literal, which is what the match was always meant to do.
lint-git runs gitlint through uvx, which fetches it for that one command and
puts nothing in the runner's Python. lint-summary and test-summary sync from a
new src/summary/uv.lock, so the versions those jobs resolve are the versions in
the tree, which is what src/backend and src/agents already do. The agents sync
refuses source distributions, since that package installs no project of its own.
Follows suitenumerique/menshen#73.
The rule downloaded the gitmoji list with requests, so lint-git had to install
that package before the linter could run. urllib.request is in the standard
library and answers the same call, leaving one fewer package fetched on the
runner before the job's own command starts.
--no-sync already stops uv run resolving an environment of its own,
and that is what the version findings were about. Building is a
separate guarantee that nothing on the line carried, so --no-build
now says it outright. It is inert beside --no-sync, and the three
lint jobs are unchanged.
The repository is already called meet, so the file name said nothing
about what the workflow holds. The print-statement check now excludes
the whole workflows directory rather than one file by name: its own
grep carries the literal print(, so the rename would otherwise match
it on the deleted lines and fail the job.
A tag and a branch both move, so actions/checkout@v6 and
numerique-gouv/action-trivy-cache@main ran whatever the owner had last
pushed. Each of the 65 uses now names a 40-character commit, with the
version it resolved to in a trailing comment.
dependabot.yml keeps the hash and that comment moving together.
gitlint-core keeps the linter at 0.19.1, which the wheels-only pin was
resolving down to 0.18.0 through an sdist-only sh. --ignore-scripts
stops three npm ci, one yarn install and one npm install -g running
the scripts of what they fetch, and yarn is pinned to 1.22.22. curl
holds the dockerize download and its redirects to https.
A package could run arbitrary code on the runner while installing,
and the versions were resolved rather than taken from the pins.
The two pip calls now take wheels only, and uv run no longer
resolves an environment of its own.
The component exposes the local microphone and camera state so external
tools automating the frontend can read it. It does not expose whether
publication is allowed, so such a tool cannot tell a muted microphone from
one it is not permitted to unmute, and ends up offering a control that
silently does nothing.
Expose canPublishMicrophone and canPublishCamera in the event detail and as
data attributes, reusing the useCanPublishTrack hook that already gates the
web client controls. The permissions are part of the effect dependencies so
a mid-meeting permission change emits the event.
The make lint target did not cover the summary and agents components. Update
the linting workflow to include both services and harmonize Makefile target
names. Harmonize Docker compose user declarations.
Add Spanish to the language selection system on both trees, with 872
frontend phrases and 145 backend ones translated from the French source.
The frontend to backend language map gains it too, so a Spanish browser
does not write an empty language to the account on every load.
@@ -7,6 +7,13 @@ To get started with the project, please refer to the [README.md](https://github.
Please also check out our [dev handbook](https://suitenumerique.gitbook.io/handbook) to learn our best practices.
## Before You Start
- A small fix, a typo or an obvious bug, can go straight to a pull request.
- Before a larger change, or one written with the help of AI, agree on the approach with a maintainer on the [Matrix channel](https://matrix.to/#/#meet-official:matrix.org), as the [README](https://github.com/suitenumerique/meet/blob/main/README.md#contributing) asks.
- [Develop locally](https://documentation-meet.beta.numerique.gouv.fr/contributing/develop/) sets up the code and explains every command. With an AI coding assistant, point it at that page and at [What CI Checks](#what-ci-checks).
## Creating an Issue
When creating an issue, please provide the following details:
@@ -18,50 +25,146 @@ When creating an issue, please provide the following details:
5.**Labels**: Add appropriate labels to categorize the issue (e.g., bug, feature request, documentation).
## Workflow
1. Fork [suitenumerique/meet](https://github.com/suitenumerique/meet) on GitHub, clone your fork, and add the project as a second remote:
```bash
git clone https://github.com/<you>/meet.git && cd meet
2. Create a branch from the project's latest `main`:
```bash
git fetch upstream
git switch -c my-change upstream/main
```
3. Commit in steps, each one change that builds and passes the tests on its own: maintainers may merge commits as they are.
4. Add a line to `CHANGELOG.md`, as [Changelog Update](#changelog-update) describes.
5. Go through the [Checklist](#checklist).
6. Push to your fork and open the pull request against `main`.
7. To bring the branch up to date, rebase it rather than merging `main` into it:
```bash
git fetch upstream
git rebase upstream/main
git push --force-with-lease
```
## Commit Message Format
All commit messages must adhere to the following format:
`<gitmoji>(type) title description`
```
<gitmoji>(<scope>) <title>
* <**gitmoji**>: Use a gitmoji to represent the purpose of the commit. For example, ✨ for adding a new feature or 🔥 for removing something, see the list here: <https://gitmoji.dev/>.
***(type)**: Describe the type of change. Common types include `backend`, `frontend`, `CI`, `docker` etc...
<description>
```
* <**gitmoji**>: Use a gitmoji to represent the purpose of the commit. For example, ✨ for adding a new feature or 🔥 for removing something. Copy it from the list at <https://gitmoji.dev/>: some, such as 🔒️, end in an invisible character that an emoji picker can leave out, and gitlint then rejects the title.
* **(scope)**: The part of the repository the change touches, such as `backend`, `frontend`, `summary`, `helm`, `ci` or `docs`.
* **title**: A short, descriptive title for the change, starting with a lowercase character.
* **description**: Include additional details about what was changed and why.
To see the scopes in use, most used first:
```bash
git log -300 --format=%s upstream/main | sed -n 's/^[^(]*(\([^)]*\)).*/\1/p' |
sort | uniq -c | sort -rn
```
CI checks every commit with gitlint, using [`.gitlint`](https://github.com/suitenumerique/meet/blob/main/.gitlint) and the rule in [`gitlint/gitlint_emoji.py`](https://github.com/suitenumerique/meet/blob/main/gitlint/gitlint_emoji.py). Run the same check before pushing:
The [devenv shell](https://documentation-meet.beta.numerique.gouv.fr/contributing/develop/#nix-with-devenv) installs gitlint as a commit-message hook, so it runs on every commit.
### Example Commit Message
```
✨(frontend) add user authentication logic
✨(frontend) add user authentication logic
Implemented login and signup features, and integrated OAuth2 for social login.
```
## Changelog Update
Please add a line to the changelog describing your development. The changelog entry should include a brief summary of the changes, this helps in tracking changes effectively and keeping everyone informed. We usually include the title of the pull request, followed by the pull request ID to finish the log entry. The changelog line should be less than 80 characters in total.
Please add a line to the [changelog](https://github.com/suitenumerique/meet/blob/main/CHANGELOG.md) describing your development, under `## [Unreleased]`, in the `### Added`, `### Changed` or `### Fixed` section. The changelog entry should include a brief summary of the changes, this helps in tracking changes effectively and keeping everyone informed. The line is usually the commit title, sometimes ended with the pull request number.
CI fails the pull request when:
- `CHANGELOG.md` is unchanged, unless a maintainer adds the `noChangeLog` label.
- A line other than a link reference is 80 characters or more. The `wc -L` command in [What CI Checks](#what-ci-checks) checks it: in a UTF-8 terminal it counts most gitmojis as two columns, so it can read more than CI does, and a line under 80 there passes.
### Example Changelog Message
```
## [Unreleased]
## Added
### Added
- ✨(frontend) add AI to the project #321
```
## What CI Checks
These jobs run on each pull request to `main`, most through shared workflows from [suitenumerique/ci](https://github.com/suitenumerique/ci). Each one runs locally:
| Workflow | Run it locally |
|---|---|
| [Project quality](https://github.com/suitenumerique/meet/blob/main/.github/workflows/code-quality.yml), commits | `git diff upstream/main..HEAD -- src/backend src/summary src/agents \| grep -E '^\+.*\bprint\('` must print nothing, then the gitlint command of [Commit Message Format](#commit-message-format) and the `fixup!` check of [Code Review](#code-review) |
| CI, SDK | `cd src/sdk/library && npm ci && npm run lint && npm run check && npm run build` |
| [Docker Hub](https://github.com/suitenumerique/meet/blob/main/.github/workflows/docker-hub.yml) | `npm run build` in `src/frontend` for the type errors; `docker build --build-arg DOCKER_USER=$(id -u):$(id -g) --target backend-production .` for the backend, and `--target production` in `src/summary` or `src/agents` when you touch them. CI also fails an image on a HIGH or CRITICAL Trivy finding |
It is nice to add information about the purpose of the pull request to help reviewers understand the context and intent of the changes. If you can, add some pictures or a small video to show the changes.
### Don't forget to:
- check your commits
- check the linting: `make lint && make frontend-lint`
- check the tests: `make test`
- add a changelog entry
### Checklist
- [ ] Every job in [What CI Checks](#what-ci-checks) passes locally.
- [ ] The change has tests.
- [ ] A model change comes with its migration, from `make makemigrations`.
- [ ] An External API change is reflected in [`docs/reference/openapi.yaml`](https://github.com/suitenumerique/meet/blob/main/docs/reference/openapi.yaml), and in `resource_server.yaml` beside it when the user's OIDC token routes change.
- [ ] Every new or changed string is translated in the five frontend languages, and in the backend catalogs when it is a backend string, see [Translations](https://documentation-meet.beta.numerique.gouv.fr/contributing/develop/#translations).
- [ ] An interface change works with the keyboard alone, every control has a name a screen reader reads, and the description shows it with screenshots.
- [ ] A new or changed setting is documented in [Environment variables](https://documentation-meet.beta.numerique.gouv.fr/reference/env-variables/), and a user-visible change on the page that describes the feature.
- [ ] A change that asks operators to do something when they upgrade is described in [`UPGRADE.md`](https://github.com/suitenumerique/meet/blob/main/UPGRADE.md).
- [ ] No secret, password or key is committed.
Once all the required tests have passed, you can request a review from the project maintainers.
## Code Review
A maintainer reviews the pull request. Keep it to one change: a fix and a refactor go in two pull requests, so neither waits on the other.
CI rejects `fixup!` and `squash!` commits, so a review fix goes into the commit it belongs to, not a new one on top:
```bash
# Record the fix against the commit it amends
git commit --fixup=<sha of that commit>
# Fold every fixup into its target: save the list the editor opens.
# Keep -i: Git before 2.44 ignores --autosquash without it.
git rebase -i --autosquash upstream/main
# Check that no fixup! or squash! title remains: this prints nothing
Please maintain consistency in code style. Run any linting tools available to make sure the code is clean and follows the project's conventions.
@@ -70,6 +173,10 @@ Please maintain consistency in code style. Run any linting tools available to ma
Make sure that all new features or fixes have corresponding tests. Run the test suite before pushing your changes to ensure that nothing is broken.
## Security
Never report, discuss or fix a vulnerability in a public issue or pull request first: report it privately, as [SECURITY.md](https://github.com/suitenumerique/meet/blob/main/SECURITY.md) describes.
## Asking for Help
If you need any help while contributing, feel free to open a discussion or ask for guidance in the issue tracker. We are more than happy to assist!
Powered by [LiveKit](https://livekit.io/), LaSuite Meet offers Zoom-level performance with high-quality video and audio. No installation required—simply join calls directly from your browser. Check out LiveKit's impressive optimizations in their [blog post](https://blog.livekit.io/livekit-one-dot-zero/).
Powered by [LiveKit](https://livekit.io/), LaSuite Meet offers Zoom-level performance with high-quality video and audio. No installation required—simply join calls directly from your browser. Check out LiveKit's impressive optimizations in their [blog post](https://blog.livekit.io/livekit-one-dot-zero/).
> [!TIP]
> New here? Start by introducing yourself in our Matrix channel:
@@ -49,21 +49,23 @@ Powered by [LiveKit](https://livekit.io/), La Suite Meet offers Zoom-level perfo
- Telephony integration
- Secure participation with robust authentication and access control
- Customizable frontend style
- LiveKit Advances features including:
- speaker detection
- LiveKit advanced features including:
- speaker detection
- simulcast
- end-to-end optimizations
- selective subscription
- SVC codecs (VP9, AV1)
LaSuite Meet is fully self-hostable and released under the MIT License, ensuring complete control and flexibility. It's simple to [get started](https://visio.numerique.gouv.fr/) or [request a demo](mailto:visio@numerique.gouv.fr).
LaSuite Meet is fully self-hostable and released under the MIT License, ensuring complete control and flexibility. It's simple to [get started](https://visio.numerique.gouv.fr/) or [request a demo](mailto:visio@numerique.gouv.fr).
We’re continuously adding new features to enhance your experience, with the latest updates coming soon!
⭐ **Find LaSuite Meet useful?** Click the Star button at the top right of this page. It helps other people and public organizations discover the project.
### 🚀 Major roll out to all French public servants
On the 29th of January 2026, Prime Minister Sébastien Lecornu, announced the full deployment of Visio—the French government’s dedicated Meet platform—to all public servants. ([Source in English](https://www.nytimes.com/2026/01/29/world/europe/france-zoom-alternative-visio.html))
On the 29th of January 2026, Prime Minister Sébastien Lecornu announced the full deployment of Visio—the French government’s dedicated Meet platform—to all public servants. ([Source in English](https://www.nytimes.com/2026/01/29/world/europe/france-zoom-alternative-visio.html))
## Table of Contents
@@ -79,30 +81,19 @@ On the 29th of January 2026, Prime Minister Sébastien Lecornu, announced the fu
## Docs
We're currently working on both technical and user documentation for La Suite Meet. In the meantime, many of the essential aspects are already well covered by the [LiveKit documentation](https://docs.livekit.io/home/) and their [self-hosting guide](https://docs.livekit.io/home/self-hosting/deployment/). Stay tuned for more updates!
The documentation, for users, administrators and integrators, is at [documentation-meet.beta.numerique.gouv.fr](https://documentation-meet.beta.numerique.gouv.fr/).
## Self-host
### LaSuite Meet is easy to install on your own servers
### LaSuite Meet is easy to install on your own servers
We use Kubernetes for our [production instance](https://visio.numerique.gouv.fr/) but also support Docker Compose. The community contributed a couple other methods (Nix, YunoHost etc.) check out the [docs](/docs/installation/README.md) to get detailed instructions and examples.
We use Kubernetes for our [production instance](https://visio.numerique.gouv.fr/) but also support Docker Compose. The community contributed a couple of other methods (Nix, YunoHost, etc.); check out the [docs](/docs/self-hosting/index.md) for detailed instructions and examples.
**Questions?** Open an issue on [GitHub](https://github.com/suitenumerique/meet/issues/new?assignees=&labels=bug&template=Bug_report.md) or join our [Matrix community](https://matrix.to/#/#meet-official:matrix.org).
> [!NOTE]
> Some advanced features (ex: recording, transcription) lack detailed documentation. We're working hard to provide comprehensive guides soon.
#### Known instances
We hope to see many more, here is an incomplete list of public La Suite Meet instances. Feel free to make a PR to add ones that are not listed below🙏
| [visio.numerique.gouv.fr](https://visio.numerique.gouv.fr/) | DINUM | French public agents working for the central administration and the extended public sphere. ProConnect is required to login in or sign up |
| [visio.suite.anct.gouv.fr](https://visio.suite.anct.gouv.fr/) | ANCT | French public agents working for the territorial administration and the extended public sphere. ProConnect is required to login in or sign up |
| [visio.lasuite.coop](https://visio.lasuite.coop/) | lasuite.coop | Free and open demo to all. Content and accounts are reset after one month |
| [mosacloud.cloud](https://mosa.cloud/) | mosa.cloud | Demo instance of mosa.cloud, a dutch company providing services around La Suite apps. |
| [Clever Cloud](https://www.clever.cloud/product/visio/) | clever cloud | Openvisio is a sovereign video conferencing solution based on LaSuite Meet offered by [Clever Cloud](https://www.clever.cloud/). |
| [Email.eu](https://email.eu/) | Email.eu | Sovereign business workspace. |
The public instances are listed in the [documentation](https://documentation-meet.beta.numerique.gouv.fr/#known-public-instances). Feel free to make a PR adding yours to [`docs/index.md`](docs/index.md).
# Contributing
@@ -135,7 +126,7 @@ Open source is much more than code. Writing documentation, improving onboarding,
When you’re ready, here are a few ways to get involved:
* 👋 **Say hello** and share your ideas with the community and maintainers on our [Matrix channel](https://matrix.to/#/#meet-official:matrix.org)
* 🛠️ **Open a PR** by following our guide to [develop LaSuite Meet locally](https://github.com/suitenumerique/meet/blob/main/docs/developping_locally.md)
* 🛠️ **Open a PR** by following our guide to [develop LaSuite Meet locally](https://github.com/suitenumerique/meet/blob/main/docs/contributing/develop.md)
* 💡 **Suggest an idea** by opening a [feature request](https://github.com/suitenumerique/meet/issues/new?assignees=&labels=enhancement&template=Feature_request.md)
* 🐛 **Report a bug** by opening a [bug report](https://github.com/suitenumerique/meet/issues/new?assignees=&labels=bug&template=Bug_report.md)
@@ -149,9 +140,9 @@ Whether you’re actively contributing or just curious about the project, you’
## Philosophy
We’re relentlessly focused on building the best open-source video conferencing product—LaSuite Meet. Growth comes from creating something people truly need, not just from chasing metrics.
We’re relentlessly focused on building the best open-source video conferencing product—LaSuite Meet. Growth comes from creating something people truly need, not just from chasing metrics.
Our users come first. We’re committed to making LaSuite Meet as accessible and easy to use as proprietary solutions, ensuring it meets the highest standards.
Our users come first. We’re committed to making LaSuite Meet as accessible and easy to use as proprietary solutions, ensuring it meets the highest standards.
Most of the heavy engineering is handled by the incredible LiveKit team, allowing us to focus on delivering a top-tier product. We follow extreme programming practices, favoring pair programming and quick, iterative releases. Challenge our tech and architecture—simplicity is always our top priority.
@@ -165,7 +156,7 @@ To learn more, don't hesitate to [reach out](mailto:visio@numerique.gouv.fr).
### Help us!
Come help us make LaSuite Meet even better. We're growing fast and [would love some help](mailto:visio@numerique.gouv.fr).
Come help us make LaSuite Meet even better. We're growing fast and [would love some help](mailto:visio@numerique.gouv.fr).
`env.d/production.dist/hosts` and the OIDC endpoints of
`env.d/production.dist/common` now read `IDP_HOST`. If you refresh `common`
from the new templates, rename `KEYCLOAK_HOST` to `IDP_HOST` in your `.env`
too, or the OIDC endpoints get an empty host and login fails.
`hosts` also no longer carries `BACKEND_INTERNAL_HOST`,
`FRONTEND_INTERNAL_HOST` and `LIVEKIT_INTERNAL_HOST`. A deployment that still
mounts `docker/files/production/default.conf.template` needs them in its
`.env`, or the frontend nginx fails to start. The compose examples now use
`docs/examples/meet/nginx-routing.conf` in its place.
### Room access tokens moved to POST /join/
LiveKit access tokens are no longer returned when reading a room. They are henceforth issued by a dedicated `POST
/api/v1.0/rooms/{id}/join/` endpoint, which receives the display name in its JSON body. Display names therefore no
longer appear in URLs and token retrieval by logged-in users is now protected against CSRF.
-`GET /api/v1.0/rooms/{id}/` no longer returns the `livekit` block, and neither do the list, create and update
responses
- The `username` field of `POST /api/v1.0/rooms/{id}/request-entry/` is now limited to 255 characters
- When the caller may enter the room directly, the response contains the room data plus a `livekit` block
- When the caller must go through the lobby, the response has no `livekit` block
- Session-authenticated callers must send the CSRF token (`X-CSRFToken` header and `csrftoken` cookie)
- Every join returning a `livekit` block mints a new token, valid for 6 hours, and anonymous callers get a new
participant identity each time.
- Joining again returns no `livekit` block to callers admitted through the lobby.
- Unregistered rooms are always entered directly: joining one always returns a `livekit` block.
- Unregistered rooms are no longer described with an `is_administrable` field,
### Unregistered rooms can no longer be named after a UUID
When `ALLOW_UNREGISTERED_ROOMS` is enabled, reading or joining a room whose identifier normalizes to an empty slug or to
a UUID now returns a 404 instead of describing or letting callers join the unregistered room.
This closes a lobby bypass since a registered LiveKit room is named after its room id. An identifier crafted to not
parse as a UUID, yet to normalize to an existing room's id, matched no registered room and was treated as an
unregistered one. The token minted for that unregistered room was valid for the registered room: anyone could join a
trusted or restricted room without being admitted through the lobby.
## v1.37.0
### Crisp support chat replaced by the LaSuite support form
The Crisp chat is gone, the support button, now open the LaSuite support form. They are hidden unless these `FRONTEND_SUPPORT` keys are all set.
The former help links for recording/transcription are now redirecting to the adhoc page in the official documentation
### Pages hidden from search engines
Every page except `/` now carries a `noindex` tag and, with the provided nginx configurations, an `X-Robots-Tag` header to stay out of search indexes.
A new `robots.txt` also blocks the main AI crawlers. If you maintain your own nginx configuration, copy the `location = /` and `location = /home.html` blocks from `src/frontend/default.conf`, and the `X-Robots-Tag` header from its `location /`.
### LiveKit server v1.13.11 and egress v1.15.0
The development stack now runs LiveKit server v1.13.11 and LiveKit Egress v1.15.0. When upgrading your own deployment:
- Deploy this release of Meet first, then upgrade egress, then the server. Older Meet releases discard the partial recordings described below.
- Transcripts no longer go through Chrome. Since egress v1.13, audio-only recordings always use the egress SDK pipeline.
- A transcript now waits for actual audio: it stays "starting", and participants are asked to turn on their microphone, while every microphone is muted.
- Recordings cut short are now kept. Since egress v1.14, a transcript that lost the room, or any recording drained because its egress host ran out of CPU, still uploads what it recorded, even though egress may report it as failed. Meet saves and processes it as usual, and tells participants the recording was interrupted. A screen recording that loses the room is still reported as complete, and saved without a warning.
- Telephony: since server v1.13.8, a call that matches no dispatch rule, for example while no room is active, is rejected with a SIP 404 instead of being silently dropped.
## [v1.35.0]
### Purging deleted rooms
Deleting a room now only soft deletes it: it is hidden but keeps its slug and PIN code. The `purge_inactive_rooms` command also permanently deletes soft-deleted rooms. See [the room purge documentation](docs/self-hosting/configuration/room-purge.md).
- Inactive soft-deleted rooms are purged along with the other inactive rooms: if `ROOM_INACTIVITY_DELETION_DAYS` is already set, nothing else is needed.
- The new `ROOM_DELETED_RETENTION_DAYS` setting purges soft-deleted rooms that many days after their deletion, whether they are inactive or not. It is unset by default.
- As for inactive rooms, a deleted room holding a saved recording its users may still access is kept until that recording expires.
### Recording permission
We've introduced the ability for any authenticated participant to start a recording or transcription, but this behavior remains configurable.
You can keep recording and transcription restricted to room admins and owners by setting `RECORDING_AUTHENTICATED_PARTICIPANTS_ENABLED` to `false`.
## [v1.34.0]
### Marketing / Brevo integration now uses `django-lasuite`
The in-house marketing service (`core.services.marketing`) has been removed and
replaced by the shared implementation from `django-lasuite`
(`lasuite.marketing`). This fixes a bug where updating a user's contact on
Brevo overwrote their list memberships, removing lists set by other
LaSuite products. Existing lists are now preserved and merged.
**Celery worker required.** Newsletter signup on login
(`SIGNUP_NEW_USER_TO_MARKETING_EMAIL=True`) is now dispatched as an
| `RECORDING_ENCODING_ENABLED` | Nothing. A default encoding is now always built (see below). **Not**`RECORDING_CUSTOM_ENCODING_ENABLED`, which gates a different feature. |
| `RECORDING_ENCODING_WIDTH` | The `width` of the entry selected by `RECORDING_ENCODING_DEFAULT_RESOLUTION` in `RECORDING_ENCODING_AVAILABLE_RESOLUTIONS`. |
| `RECORDING_ENCODING_HEIGHT` | The `height` of that same entry. |
| `RECORDING_ENCODING_FRAMERATE` | The `fps` of the profile selected by `RECORDING_ENCODING_DEFAULT_PROFILE` in `RECORDING_ENCODING_AVAILABLE_PROFILES`. |
| `RECORDING_ENCODING_VIDEO_BITRATE_KBPS` | That profile's `kbps`. |
`RECORDING_ENCODING_AUDIO_BITRATE_KBPS` and `RECORDING_ENCODING_KEY_FRAME_INTERVAL_S`
keep their names and meaning. The keyframe interval now defaults to `0` (unset,
encoder's choice) instead of `4.0`.
#### If you never set `RECORDING_ENCODING_ENABLED=True`
The shipped defaults (`RECORDING_ENCODING_DEFAULT_PROFILE=full`,
`RECORDING_ENCODING_DEFAULT_RESOLUTION=720p`) match LiveKit's built-in
`RECORDING_CUSTOM_ENCODING_ENABLED` (default `False`) toggles whether the
start-recording API accepts an `encoding` object
(`{"resolution": "720p", "profile": "talking_heads"}`, `profile` optional. It
falls back to `RECORDING_ENCODING_DEFAULT_PROFILE`) that overrides the default for
a single recording. It does not enable or disable the
default encoding, which is built from the two `RECORDING_ENCODING_DEFAULT_*`
settings either way. Leaving it at `False` preserves the previous behavior, where
every recording uses the server-side encoding: requests carrying
`options.encoding` are rejected with a `400` before the recording is created, so
nothing is persisted and no egress is started.
Before enabling it:
- clients can only pick keys you declared; there is no way to send a raw width or bitrate
- as of this implementation, the frontend never sends `encoding`
-`encoding` is accepted but ignored for `transcript` recordings, whose audio-only
egress has no video encoding to configure.
See [docs/self-hosting/configuration/recording.md](docs/self-hosting/configuration/recording.md#configuring-recording-encoding)
for the full setting reference, the shipped profile table and the tuning caveats.
## v1.33.0
### Purging inactive rooms
Rooms now keep track of the last time they were started (`last_started_at`), fed by LiveKit's `room_started` webhook. A new `purge_inactive_rooms` management command permanently deletes the rooms that have not been started for `ROOM_INACTIVITY_DELETION_DAYS` days. See [the room purge documentation](docs/self-hosting/configuration/room-purge.md).
- The feature is **disabled by default**: nothing is deleted unless you set `ROOM_INACTIVITY_DELETION_DAYS`.
- The migration marks every existing room as started at the time of the upgrade, so no existing room can be purged before a full inactivity period has elapsed after upgrading.
- Rooms holding a saved recording their users may still access are kept: any saved recording, or, when `RECORDING_EXPIRATION_DAYS` is set, a saved recording created within that window.
- Inactivity is measured from LiveKit's `room_started` webhook: if it is not delivered to your backend, rooms in daily use look inactive and get purged.
- When a room is purged, all it's configuration and access rights are also deleted. Its slug becomes available again and can be reused when a meeting is created from that same URL.
* With `ALLOW_UNREGISTERED_ROOMS=false`, only an authenticated user can navigate to a previously existing link after the room has been purged. Doing so recreates the room in the database with a fresh configuration, with that user associated with it and granted admin rights.
* With `ALLOW_UNREGISTERED_ROOMS=true`, any user can reopen the purged room by navigating to the same URL. In that case, the room is created dynamically and no corresponding room entry is persisted in the database.
### Local development: MinIO replaced by Garage
The development stacks now use [Garage](https://garagehq.deuxfleurs.fr/) instead of MinIO as S3 storage. Garage keeps its own format in `data/media/meta` and `data/media/data` and cannot read what MinIO left there, so local recordings and files will be lost.
To migrate a local environment:
1. Stop the stack and remove its containers, including the former `minio` one: `docker compose down --remove-orphans`
2. Optionally reclaim the space used by MinIO: `rm -rf data/media && make data/media`
3. In your `env.d/development/*` files, replace `minio:9000` by `garage:9000`, the `meet` / `password` credentials by `meet-access-key` / `meet-secret-access-key`, and add `AWS_S3_REGION_NAME=local` (or delete these files and run `make create-env-files`)
4. Run `make create-env-files` to generate `env.d/development/garage`, which holds a random RPC secret for Garage.
5. Rebuild the images, since the summary and agent images now install boto3 instead of minio
### Summary service and metadata collector: boto3 replaces the minio client
The summary service and the metadata collector agent now talk to S3 through boto3 instead of the minio client, with the same settings.
Requests are now signed for `AWS_S3_REGION_NAME` as-is. When it is not set, the region is no longer looked up from the bucket: boto3 falls back to `AWS_DEFAULT_REGION`. If you left `AWS_S3_REGION_NAME` unset, set it to your provider's region before upgrading, or providers that check the signing region will reject the transcripts, summaries and meeting metadata uploads, as well as their signed URLs.
Also:
- Signed URLs to transcripts and summaries are now always path-style (`<endpoint>/<bucket>/<key>`), whereas the minio client used virtual-hosted-style URLs
- The metadata collector now accepts `AWS_S3_ENDPOINT_URL` with or without a scheme, like the summary service: the scheme always follows `AWS_S3_SECURE_ACCESS`.
### Helm chart: media services default to Garage
The `meet` chart now defaults `serviceMedia.host` and `serviceMediaFiles.host` to `garage.meet.svc.cluster.local`, and the `upstream-vhost` annotation of `ingressMedia` and `ingressMediaFiles` to `garage.meet.svc.cluster.local:9000`. If you relied on the former `minio.meet.svc.cluster.local` defaults, set these values explicitly to your S3 service before upgrading, or recordings and files stop being served under `/media`.
## v1.30.0
### Removing S3 storage-event webhooks for recordings
Recordings were previously confirmed as saved by an S3 storage-event webhook posting to `/api/v1.0/recordings/storage-hook/`. That endpoint has been removed: recordings are now always finalized from LiveKit's own `egress_ended` webhook, which has been the default path since v1.22.0.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.