Dorbar
Dorbar is a self-hosted audio and video meeting platform for governments and regulated enterprises that need meetings, recordings, and chat to stay inside their own infrastructure. Every room is fixed at schedule time as either Official or Sealed, and that single choice decides whether the platform may record, transcribe, and moderate the session or stays completely blind to it.
Dorbar is a sovereign video-meeting platform for governments and regulated enterprises, run inside your own boundary. A meeting's security class — and who may join it — is fixed when it is scheduled, not left to participants, and the interface runs in Bengali or English at full parity.
What Dorbar does
Whoever schedules a meeting fixes it as Official or Sealed up front, and that classification travels with the room's join credential so the client and server always agree on what the room is allowed to do.
Sealed rooms are designed to encrypt every media frame on each participant's device using MLS (RFC 9420) group keying, SFrame media encryption, and a post-quantum X-Wing hybrid key exchange, so the meeting server relays only ciphertext and cannot see the content.
In an Official room a host can record the session to in-cluster object storage, with a start confirmation, an always-visible recording badge for every participant, and a per-join notice; Sealed rooms cannot be recorded by design.
Completed Official recordings are transcribed by an in-cluster speech model that runs fully offline with no cloud speech service, producing a timestamped, language-detected transcript.
Official rooms can show real-time captions from an offline in-cluster speech model in English or Bengali with a host-selected language, shipped switched off by default and never offered in Sealed rooms.
Transcripts are scanned offline for categories such as personal, payment-card, and health data, extendable with operator-defined patterns, and the sensitive text is masked on read for anyone who lacks the elevated view permission.
Hosts admit or deny people from a waiting room, lock the room, mute or remove participants, reassign in-room roles, and spotlight a speaker, with the join gate enforced before anyone reaches the media server at all.
Every consequential action is appended to a SHA-256 hash-chained audit log that can be recomputed and verified on demand, so any alteration, deletion, or reordering of the record is detectable.
The entire interface is available in English and Bengali at full parity with self-hosted fonts and no content delivery network, so the platform reads natively for its users with no external dependency.
The meeting server, database, object storage, STUN/TURN, and speech models all run inside a single cluster with no external runtime dependency, packaged for installation into a disconnected network from a mirrored registry.
The two-class room model
Dorbar does not treat every meeting the same. At schedule time each room is fixed as one of two mutually exclusive classes, and that choice is carried end to end. Official rooms are the on-premises-trusted default. The meeting server inside your own hardened cluster is the trust anchor, media is protected in transit, and because the server can read the media it can record, transcribe, live-caption, and moderate the session. Official rooms are protected in transit only and are never end-to-end encrypted. Sealed rooms are for the most sensitive sessions. Media is end-to-end encrypted on each participant's device, so the meeting server relays only ciphertext and every server-side feature that would touch content, including recording, transcription, captions, and breakout rooms, is refused by design. The two classes cannot be blurred or mixed within one room. Assurance limit: the Sealed end-to-end encryption runs at the browser tier and is currently rated medium assurance pending an independent external cryptographic review; a hardened native-client tier is a later phase.
Sovereign identity and access
What are Official and Sealed rooms in Dorbar?
Two room classes, fixed when a meeting is scheduled. Official rooms are server-trusted and protected in transit — the media server handles the streams. Sealed rooms are end-to-end encrypted so the server cannot read them; that path runs at the browser tier and is rated medium assurance, pending external cryptographic review. Official rooms are never end-to-end encrypted.
How does Dorbar control who can join a meeting?
Classification and access are set on the meeting at scheduling, not left to participants, with a lobby and waiting-room control and host moderation. Administrative reach and in-meeting authority are two independent role dimensions, so holding one never grants the other.
Does Dorbar work in Bengali?
Yes — the interface runs in Bengali or English at full parity, which is the point for Bangladesh government and regulated-sector deployments.
Can meetings be recorded and transcribed?
Official rooms support recording and offline transcription, over a tamper-evident audit trail of the security-relevant events. Recording cannot be paused mid-session, and local client-side recording is deliberately not built.
Sealed-room encryption runs at the browser tier: post-quantum confidentiality on stock browser crypto, medium assurance pending external cryptographic review, with a hardened native client a later phase. Recording cannot be paused mid-session, and local recording is deliberately not built.