Cassini

Warning

Cassini is in beta. We run it daily on our own Nextcloud, but expect rough edges. Read the changelog before you update, and please open an issue if something breaks.

Cassini is a recording backend for Nextcloud Talk that puts you in control of your meeting data.

• Record and transcribe on your own infrastructure. Cassini runs on the hardware you give it. Speech-to-text runs in-process with sherpa-onnx and NVIDIA Parakeet models, from a CPU or a CUDA image.
• AI is opt-in. Summaries and LLM-powered insights run only once you configure a model - local or third-party - in the app.
• A self-contained, portable meeting file: audio, transcript, summary and tags are packed into one file, so you can take your meetings with you.

Meetings are where decisions get made, and these are easily lost unless someone writes it down afterwards. We believe voice is the most valuable context a team produces, and the least captured. In an LLM age you should be able to build on it without handing your meetings to a third party.

What you get

Cassini is an ExApp which can be installed from the Nextcloud App Store. Once it's running, it shows up as a new app icon inside your Nextcloud suite:

• Works with any Nextcloud Talk room, group calls and 1:1 calls. Just press Record and Cassini is listening.
• Transcripts and synced audio playback with speaker IDs.
• Configurable recording access: everyone with a Nextcloud account by default, or room-based permissions for private recordings when participant access is enabled. Public-room recordings remain visible to all signed-in accounts.
• One portable meeting file per meeting that can be opened without Cassini.
• Search and tags across meetings.
• Configure in-app AI providers for per-meeting summaries and cross-meeting insights.
• A CLI and an agent skill, to build workflows with an external harness.

Limitations

• No live transcription or captions. Transcription starts after recording stops. The operator schedules processing around live recordings to protect capture.
• Audio only. Cassini records the video streams, but the meeting file, transcript and viewer are audio only for now.

Transcription starts off. Configure a model with progress, or prepare and import it offline. Models persist across application upgrades.

What leaves your server

Recording and transcription run on your own hardware. No audio and no transcript leaves the host for those steps.

If you configure a language model, transcript text goes to it in two cases: automatically, to summarise each meeting, and on request, when someone asks a question about meetings they have access to.

There is no telemetry.

Full details are in docs/privacy.md.

The meeting file

Each published meeting is one ordinary .opus file. Any audio player plays it.

The transcript, speaker names, summary and tags travel inside the same file, so a meeting you copy off your server is still a complete meeting, and it stays readable without Cassini.

The format is an open specification at format.gocassini.com.

What Cassini writes is described in docs/portable-meeting-format.md.

Requirements

• Nextcloud 32 to 35 with AppAPI and a HaRP deploy daemon.
• Talk with the High-performance backend (standalone signalling). Cassini joins calls as an internal signalling client, so it needs the signalling server's internalsecret. This is the one value you have to supply by hand.
• Nextcloud Files sharing enabled. Cassini uses built-in direct shares; no Team folders or Everyone Group app is needed.
• Optional: an NVIDIA GPU with the Container Toolkit (x86_64) for faster transcription. CPU is the default and runs on amd64 and arm64.

Install

1. Install Cassini from the App Store. In the deploy options, paste your signalling server's internalsecret. On Nextcloud All-in-One, print it with docker exec nextcloud-aio-talk printenv INTERNAL_SECRET. On a standalone signalling server, it is internalsecret under [clients] in server.conf. You can also save the secret later under Operator › Publish pipeline › Talk authentication. Every other option can stay empty.
2. Open Cassini as an administrator. It creates the cassini service account that owns the meeting archive. Nextcloud 34.0.2 and later ask you to confirm with your password first.
3. Check recording sharing, under Operator › Settings. Cassini shares each recording with the local people, groups and Teams captured from its Talk room. For public meetings, those participants can share onward when your Nextcloud allows resharing. Cassini creates no public link. Existing installations can follow the one-user cutover.
4. Point Talk at Cassini. Back up Talk's current recording_servers value, then apply the one Cassini generates for you.
5. Record a test call in a private room and watch it arrive in Cassini.

New to external apps? Start with Before installing and the first ExApp walkthrough. After installation, use Operator › Publish pipeline to check the connection and follow the repair instructions.

Each step, with the commands, the verification checklist and GPU setup, is in docs/exapp-install.md.

CLI and skills

cassini meetings reads published recordings from outside Nextcloud, as a Nextcloud user with an app password. It sees exactly what that account can see in the app, and nothing more.

cassini meetings list --from 2026-08-01
cassini meetings search "offer" --tag hiring   # finds where it was said
cassini meetings context <meeting-id>          # transcript and summary, ready for an agent
cassini meetings fetch <meeting-id> --out standup.opus

An agent skill ships at .claude/skills/cassini-meetings/SKILL.md. It teaches a coding agent when and how to use these commands. Claude Code loads it from a checkout; for other agents, point them at the file.

Setup and worked examples are in docs/agent-meeting-access.md.

Development

You need Go 1.24. From a checkout, ./bin/cassini builds the CLI and runs it in your current directory:

./bin/cassini --help
./bin/cassini doctor     # checks your environment before a long build

• docs/README.md: the documentation index for contributors
• docs/cli.md: record, build, publish, serve and inspect from a checkout
• harness/README.md: a local Nextcloud Talk lab, driven by cassini dev
• deployment/README.md: a Docker Compose bundle for development and staging. To run Cassini on a Nextcloud, use the app install above.

Contributing

Issues and pull requests are welcome. Read CONTRIBUTING.md first; it covers the branch and changelog conventions this repo uses.

• License: GNU AGPLv3
• Changelog: CHANGELOG.md
• Security reporting: SECURITY.md