Tenant isolation and data protection

A universe separates one team’s Lightspeed records from another team’s within the same deployment. Sessions, profiles, workspaces, bots, credentials, channels, and environment records belong to a universe. The runtime processes and infrastructure serving those universes can be shared.

Consider Acorn and Cedar, each with a workspace named release-notes. A request authenticated for Acorn resolves that name inside Acorn’s universe. It does not reach Cedar’s workspace, even when the two names are identical. The same boundary follows requests into storage and workflows.

How the universe follows a request

A universe API key selects its one stored universe. A deployment key can address a universe by supplying its UUID in x-lightspeed-universe, subject to the key’s method groups. The Platform uses this second path after checking the person’s membership and permissions. The runtime resolves a service bound to that universe before dispatching ordinary resource operations.

Layer	How resources remain separate
Runtime API	Key authentication resolves the request’s universe and allowed method groups.
PostgreSQL	Universe-bound queries and composite keys/foreign keys associate tenant records with universe_id.
Content-addressed storage	The blob catalog includes the universe in its key. External objects live under <prefix>/universes/<uuid>/cas/….
Blob cache	Cache lookups include both the universe UUID and blob reference.
Temporal	Workflow identities include the universe, and workers resolve the corresponding universe service.
Encrypted secrets	Authenticated encryption binds ciphertext to the universe UUID, secret ID, and secret kind.

A session’s Temporal workflow ID is <universe-uuid>/<session-id>. Bots and channel conversations also use universe-prefixed workflow IDs. A delegated child stays in its parent’s universe.

Identical blobs in two universes are stored separately; there is no physical blob deduplication across universes. Knowing Acorn’s content digest does not let a caller read those bytes through Cedar’s service. Inside Acorn, however, blobs/read accepts a digest without a session visibility check. A caller allowed to read blobs can retrieve known digests in that universe. This is one reason private session visibility is a narrower property than isolation between tenants.

PostgreSQL isolation uses application queries and schema constraints, not PostgreSQL row-level security policies. Database and storage administrators remain able to access the infrastructure they operate. The Platform also gives its administrators Admin access in every universe. These are trusted administrative boundaries, not accounts constrained by an ordinary member’s session visibility.

What a deployment shares

Universes share runtime processes, a PostgreSQL pool, optional object storage, provider clients, and other process resources. Session, bot, and channel task queues provide common worker capacity. Creating a universe does not reserve CPU, memory, queue throughput, or provider quota for it. A bot’s own limits can bound its work without establishing tenant-wide resource quotas.

Some registries have deployment-wide meaning: universes and API keys support routing, and environment providers are registered once before being bound to universes. Daemon identities and provider-native channel account identities have deployment-wide uniqueness.

Separate deployments are the appropriate unit when teams need independent operator access, database credentials, encryption keys, infrastructure capacity, or failure domains. If deployments share a Temporal namespace, give all three runtime task queues distinct names. A separate namespace also separates workflow identity space and namespace settings. See deployment configuration.

Credentials and encryption

Secret values and grant tokens use the deployment’s LIGHTSPEED_SECRETS_MASTER_KEY. Binding ciphertext to its universe and secret identity prevents moving it to a different identity and decrypting it there. There is no independently managed encryption key per universe. This mechanism encrypts secret material; encryption of database storage, backups, and CAS objects belongs to the infrastructure configuration. Preserve the master key as part of the recovery set.

A universe can configure its own model credentials. When a built-in provider record is absent, a deployment fallback key can supply the request, so several universes may consume the same provider account and quota. A disabled or unusable universe record blocks fallback rather than silently using that shared key. Models and credentials explains credential selection.

Credentials used by tools deserve the same attention as credentials used to enter Lightspeed. Sharing an environment also shares access to credentials injected into processes on that environment. Keeping a session private does not narrow the external permissions of those credentials.

Bot trigger metadata also carries access material. Current trigger read and list responses include bearer webhook ingest paths and chat pairing codes, and the Platform permits those methods from Viewer upward. Treat those values as available to universe members; they are not restricted to bot managers. See Bots and triggers for trigger setup and the separate HMAC and pairing mechanisms.

Machines, files, and external services

An environment record belongs to a universe, while its machine has the filesystem, network access, and operating-system permissions supplied by its operator or provider. Registering a machine does not create an additional operating-system sandbox around it. Choose templates, host permissions, network rules, and credential bindings for the work that should run there.

The VFS and an environment’s filesystem are separate domains. Files transfer only through explicit operations; session visibility does not make a shared workspace or machine private. See using environments, environment credentials, and networking and ingress.

Model providers and external tools receive the inputs sent to them. Universe scoping inside Lightspeed does not impose a retention policy or access model on those services. Their credentials and deployment configuration determine that part of the boundary.

Check an installation

Create two disposable universes with different content under the same workspace name. Read each through its own universe key, and verify that the other universe’s content is unavailable. Inspect the corresponding universe-prefixed workflow IDs. These checks exercise the installation’s routing; they do not establish every isolation property on their own.

For adopting existing universes, preserving their UUIDs during recovery, and retiring a tenant, follow Managing universes.