https://livekit.io logo
Join Slack
Powered by
Which LiveKit agents/pods could be self-hosted whi...
# ask-ai
a
Which LiveKit agents/pods could be self-hosted while keeping LiveKit Cloud? What's the minimum needed to keep the LiveKit Cloud dashboard?
t
Thanks for reaching out, Mohammed. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — I’m responding to your two questions: which LiveKit components can be self-hosted while keeping LiveKit Cloud, and the minimum needed to keep the LiveKit Cloud dashboard. • You can self-host LiveKit agents/pods (agent processes) on your infrastructure while still using LiveKit Cloud for media transport and Dashboard/observability. Self-hosted agents connect to LiveKit Cloud via WebSocket and appear in the Cloud dashboard when configured correctly. • You cannot self-host the LiveKit server (SFU/media node) and retain Agent Insights / full Cloud dashboard observability. Agent observability and dashboard insights require using LiveKit Cloud for media transport. Minimum to keep the LiveKit Cloud dashboard and Agent Insights: 1. Use LiveKit Cloud as your media transport (do not replace the Cloud SFU with a self-hosted server). 2. Have an active LiveKit Cloud project with valid API keys/secrets. 3. Run agents that connect to the LiveKit Cloud server URL (not a self-hosted server). 4. Enable Agent Observability in the project settings and use supported SDK/agent versions (required versions for observability are documented). 5. Ensure sessions complete successfully (observability data is uploaded after session end; crashes before completion may prevent upload). If you want, tell me which specific agents/pods you plan to self-host (agent type, language/runtime, or manifest), and I’ll give exact config steps and the SDK versions you’ll need. Sources: Self-hosted deployments | LiveKit Documentation | Agents on LiveKit Cloud | Deploy voice AI on global infrastructure | LiveKi... | Where to find your Agent Logs | LiveKit | Where to find your Agent Logs | LiveKit | Why is my Agent insights data missing? | LiveKit, with interactive elements
a
so self-hosting app LiveKit agents and recording (egress) ones is fine for the scenario I mentioned, right?
t
Yes — you can self-host both app agents and recording (egress) agents and still use LiveKit Cloud’s dashboard and observability, as long as the agents connect to the LiveKit Cloud server URL (i.e., LiveKit Cloud remains the media transport). Two important caveats: • Do not self-host the LiveKit server/SFU — that breaks Agent Insights and dashboard observability. • Ensure agents use supported SDK/agent versions and Agent Observability is enabled in the project settings; observability data uploads after sessions complete. Would you like configuration steps for your agent types or the SDK version matrix for recording agents? , with interactive elements
a
I'm specifically asking about the 
livekit/egress
Docker service (the headless Chrome compositor that records rooms to S3) — not recording implemented via the Agents SDK. Can I run self-hosted 
livekit/egress
workers pointed at a LiveKit Cloud project as the SFU? The egress docs say redis must be the same instance used by the LiveKit server. Since LiveKit Cloud doesn't expose its internal Redis, it seems like self-hosted egress workers would never receive dispatch jobs from a Cloud SFU. Is that correct, or is there a supported mechanism (e.g. a different dispatch path, a Redis endpoint LiveKit Cloud exposes to customers, or a direct HTTP API on the egress service) that makes this work?
t
You're asking whether a self-hosted 
livekit/egress
Docker worker can receive dispatch jobs from a LiveKit Cloud SFU. Short answer: No — with the information available, self-hosted 
livekit/egress
workers cannot receive jobs from a LiveKit Cloud SFU. Egress workers rely on Redis Pub/Sub shared with the LiveKit server for job dispatch, and LiveKit Cloud does not expose its internal Redis instance. There is no documented alternative dispatch path (public Redis endpoint, HTTP dispatch API, or other mechanism) to connect self-hosted egress to a Cloud SFU. Options you can consider: • Use LiveKit Cloud’s built-in egress (Cloud-managed) for recordings. • Self-host both the LiveKit server (SFU) and your 
livekit/egress
workers so they share the same Redis instance. • Re-implement recording client-side (Agents SDK) or run custom recording services that initiate uploads to S3 without relying on SFU-dispatched egress jobs. If you’d like, tell me which option you prefer (Cloud egress, full self-host, or a client-side recording approach) and I’ll provide concrete setup steps and required versions. Sources: Universal Egress | LiveKit | Universal Egress | LiveKit, with interactive elements
a
For Option 2 (self-hosted SFU + self-hosted 
livekit/egress
): what LiveKit Cloud features, if any, are still accessible? Specifically — can I still use the LiveKit Cloud dashboard for room monitoring and Agent Insights if my SFU is self-hosted, or is the dashboard completely unavailable? Are there any LiveKit Cloud services (TURN infrastructure, observability, analytics) that work alongside a self-hosted SFU?
t
You're asking which LiveKit Cloud features still work if you self-host the SFU and egress workers. Short answer: • LiveKit Cloud dashboard, Agent Insights (recordings, traces, logs, waveforms), and the Cloud observability features are not available when you self-host the LiveKit server (SFU). • TURN access from Cloud and any Cloud-managed dispatch for 
livekit/egress
are not confirmed available for self-hosted SFU deployments. • You can export telemetry from the self-hosted SFU (server telemetry endpoint) and build your own monitoring/analytics, and you can self-host egress + SFU together (they share Redis) to retain egress job dispatch. What remains usable: • The open-source SFU’s telemetry endpoint for exporting metrics (you can build custom dashboards/analytics). • Fully self-hosted egress + SFU workflows (they work together if they share the same Redis). If you want, I can: • Provide steps to configure a self-hosted SFU + 
livekit/egress
(Redis, Docker examples). • Show how to export SFU telemetry and suggestions for building a custom dashboard. Which would you prefer? Sources: Agent deployment overview | LiveKit Documentation | Announcing LiveKit Cloud | LiveKit | Announcing LiveKit Cloud | LiveKit | Why is my Agent insights data missing? | LiveKit | Why is my Agent insights data missing? | LiveKit, with interactive elements
2 Views
Open in Slack
PreviousNext
Powered by