vocabulary:
  name: Restate Vocabulary
  description: Domain vocabulary for the Restate durable execution engine
  version: 1.4.0
  created: '2026-05-02'
  modified: '2026-05-02'
  tags:
    - Durable Execution
    - Workflows
    - Distributed Systems

terms:
  - term: Durable Execution
    definition: The ability to make code execution reliable and fault-tolerant by recording every step in a journal. If a function crashes or fails, Restate replays the journal to resume exactly where it left off.
    category: Core Concept
    relatedTerms:
      - Journal
      - Replay
      - Exactly-Once Semantics

  - term: Journal
    definition: The persistent log maintained by Restate that records every step of a handler's execution, including service calls, state updates, timers, and their results.
    category: Core Concept
    relatedTerms:
      - Durable Execution
      - Replay

  - term: Replay
    definition: The process by which Restate recovers a failed handler by replaying its journal, skipping completed steps and resuming from the last recorded checkpoint.
    category: Core Concept
    relatedTerms:
      - Journal
      - Durable Execution

  - term: Exactly-Once Semantics
    definition: The guarantee that each side-effecting operation (external API call, database write, timer) is executed exactly once, even in the face of failures and retries.
    category: Guarantee
    relatedTerms:
      - Durable Execution
      - Idempotency

  - term: Deployment
    definition: A registered service endpoint URI that Restate routes invocations to. A deployment hosts one or more services and their handlers.
    category: Infrastructure
    relatedTerms:
      - Service
      - Handler

  - term: Service
    definition: A Restate service groups related durable handlers. Services can be of type Service (stateless), VirtualObject (stateful per key), or Workflow.
    category: Core Concept
    relatedTerms:
      - Handler
      - VirtualObject
      - Workflow

  - term: Handler
    definition: A durable function within a Restate service that processes invocations. Handler types include Exclusive (serialized per key), Shared (concurrent), and Workflow (exactly-once lifecycle).
    category: Core Concept
    relatedTerms:
      - Service
      - Invocation

  - term: VirtualObject
    definition: A Restate service type where each key represents a stateful entity. Exclusive handlers for the same key are serialized to prevent concurrent modification.
    category: Service Type
    relatedTerms:
      - Service
      - Handler

  - term: Invocation
    definition: A single call to a Restate service handler. Invocations are durably tracked and can be monitored, cancelled, killed, or purged via the Admin API.
    category: Execution
    relatedTerms:
      - Handler
      - InvocationId

  - term: InvocationId
    definition: A unique identifier for a specific Restate invocation instance, used to track and manage execution lifecycle.
    category: Identifier
    relatedTerms:
      - Invocation

  - term: Subscription
    definition: A Restate Kafka subscription that connects a Kafka topic (source) to a service handler (sink), enabling event-driven invocations with exactly-once processing.
    category: Integration
    relatedTerms:
      - Kafka

  - term: Kafka
    definition: Apache Kafka message broker integration that enables Restate services to be invoked by Kafka events via subscriptions.
    category: Integration
    relatedTerms:
      - Subscription

  - term: Idempotency
    definition: The property of an operation that can be safely retried without changing the result beyond the initial execution. Restate ensures idempotent execution via the journal.
    category: Property
    relatedTerms:
      - Exactly-Once Semantics

  - term: Purge
    definition: The Admin API operation to remove a completed invocation's state and journal from Restate storage.
    category: Operation
    relatedTerms:
      - Invocation

  - term: Bifrost
    definition: Restate's internal durable append-only log (write-ahead log) used for partition replication and fault tolerance.
    category: Internal Architecture

  - term: CRD
    definition: Kubernetes Custom Resource Definition used by the Restate Operator to manage RestateCluster, RestateDeployment, and RestateCloudEnvironment resources.
    category: Kubernetes
    relatedTerms:
      - RestateCluster
      - RestateDeployment

  - term: RestateCluster
    definition: A Kubernetes CRD that describes the configuration and desired state of a Restate cluster deployed via the Restate Operator.
    category: Kubernetes CRD
    relatedTerms:
      - CRD

  - term: RestateDeployment
    definition: A Kubernetes CRD representing a Restate service deployment managed by the Restate Operator.
    category: Kubernetes CRD
    relatedTerms:
      - CRD