Browse documentation
Vocabulary and concept map
A precise lookup for the terms used throughout Septa.
Concept map
| Familiar concept | Septa term | Key difference to retain |
|---|---|---|
| A service’s state owner | Durable actor | Identity and saved state can outlive an executing process |
| A long-running operation | Workflow | Progress is retained at supported boundaries |
| An I/O operation | Effect | The typed interaction is visible to the compiler and runtime |
| Permission to act | Scoped grant | Typed capability access still requires authority and admission |
| A package version | Content identity | Identifies exact content rather than only a name or label |
| A test report | Evidence receipt | Describes observed inputs, outcomes, origin and applicability |
| A runtime trace | Observed view | Describes retained observations with explicit coverage |
Actor
A state owner with an identity, mailbox and message handlers. A durable actor preserves state and progress under a persistence and recovery contract. Actor isolation does not imply a distinct operating-system sandbox or deployment.
Admission
The decision whether a particular request may proceed under the runtime’s applicable checks, authenticated context and authority.
Capability
A typed interface or handle for requesting operations. Its use is governed by scoped grants and runtime admission. Possessing a handle or declaring an effect does not itself grant authority.
Grant
An authorization under a defined scope and policy context. Runtime checks determine whether it permits a particular request. A grant and a typed capability handle serve different roles.
Cast and call
A cast sends a message without asking for a typed reply. A call requests a typed reply. Delivery, admission, timeout and recovery remain part of the operation’s contract.
Content identity and pin
A content identity names a defined object by its content. A pin selects that exact identity. Source, library, archive and bundle identities cover different objects and inputs; they are not interchangeable labels.
Declared, permitted and observed
- Declared: the interactions described by the checked program.
- Permitted: the operations allowed in a named authority context.
- Observed: the interactions represented by retained execution evidence.
Deterministic core
The supported execution surface with defined semantics, arithmetic and effect boundaries. Repeatability depends on the applicable code and execution contract, including recorded inputs.
Effect
A typed interaction understood by Septa’s compiler and runtime. “First-class” does not imply native Rust language support for arbitrary algebraic-effect handlers.
Evidence applicability
A comparison between an observation’s relevant recorded inputs and the current target. Applicability is separate from whether the observed outcome passed or failed.
Graph
In commands such as graph check, the application source graph and its libraries. This command group is distinct from the reusable temporal graph database built on Septa.
Managed core and native component
The managed core is code executing under Septa’s supported profile and runtime contract. A native component executes through its separate native route, with its own build process and operating-system authority.
Profile
The documented set of Rust syntax, types, operations and constraints admitted by a particular execution route. Acceptance outside an enumerated stable surface does not automatically establish a compatibility guarantee.
Workflow
Work progressing toward completion with a supported place to suspend and resume. An actor commonly continues receiving messages; a workflow commonly represents a particular operation.
Zero-build development
The installed runtime handles supported application and library edits without a separate native application build per edit. Parsing, checking and lowering still occur. Native components retain their own build path.
Build software you can reason about.