Skip to main content

Server

The Studio server handles the API, user authentication, assistant configuration, and conversation storage.

Core

Database

The database stores conversations, assistant configurations, and analytics data. Both the Studio server and the event ingestion service use this same database connection.

Authentication

Studio uses Keycloak for identity and access management.

Model Server

MS_API_URL is the address the Studio backend uses to reach the Rasa Pro model server inside the cluster. With the bundled model server, that Service is rasapro. Try your assistant opens a WebSocket from the browser to the model server on the public hostname, at the path /talk. On a Helm install, that hostname comes from rasa.rasa.ingress.hostName when it is set, and otherwise from config.ingressHost. A values file that sets only config.ingressHost still needs that same hostname on the /talk ingress. See Expose Try your assistant.

Conversation Management

Voice

To enable voice features, you must set both DEEPGRAM_API_KEY and CARTESIA_API_KEY, even if you are only using one of the providers for both speech-to-text and text-to-speech. In such a case, set the unused provider’s environment variable to a placeholder value (e.g., CARTESIA_API_KEY=unused).

Observability

Advanced

Event Ingestion

Studio consumes Rasa events from a Kafka topic to power analytics and conversation review. Ingested events are persisted to the same PostgreSQL database configured in the Database section above.

Kafka