| Term | Meaning |
|---|---|
| Actor | Who sent a command: an API key name (key:<name>), a supervisor id or a UI user. Recorded in the audit |
| Admission | The decision to start, queue or refuse an execution. Made by the scheduler against the concurrency limits |
| At-least-once | A step may run more than once but never zero times after it started. Exactly-once is not achievable across the side effects of arbitrary plugins |
| Attempt | One try of a node on one message. Nodes with retries label each message with its attempt number |
| Back-pressure | A slow consumer making upstream producers wait, which keeps memory bounded |
| Breakpoint | A rule that holds a node before the first message matching a value |
| Checkpoint | A node’s recorded output, kept so a restart or resume can replay it instead of recomputing |
| Command | An instruction to a live execution: pause, resume, step, skip, stop, restart, signal and others. Applied in order and audited |
| Control bus | The live command and event path of the server |
| Control plane | A supervisor that decides one command per event, using rules, a model or both |
| Dead letter | A message that failed after all attempts and was routed to the error port instead of being lost |
| Drain | Stopping gracefully: work already in progress finishes before the execution or worker ends |
| Durable end marker | A small record written and forced to disk with a node’s output when the node completes. A resume trusts only this marker |
| Edge | A one-way connection from a source node to a target node, optionally from a named output port |
| Engine policy | The reserved engine object in a node’s configuration: retry, error routing, time limits, buffering and placement |
| Execution | One live run of a workflow, with its own executionId, a status and a status per node |
| Fan-out | One node feeding several children. Paced by the slowest child |
| Fencing number | A counter that grows each time high-availability leadership changes hands, so a stale leader can be recognised |
| Heartbeat | A signal from a long-running node or worker that restarts its idle clock |
| Idempotency key | A client-supplied value that makes a repeated start return the first execution |
| Idempotent plugin | A plugin whose repeated execution for the same message has the same effect as one |
| Join rule | How a node with several parents treats a failed parent: ALL_SUCCESS, ALL_DONE or ONE_SUCCESS |
| Lease | A time-limited right: held by a worker to stay registered, or by a server to be the leader |
| MCP | Model Context Protocol: how AI agents call tools. Yukta exposes its control and read operations through it |
| Message | The unit of data on an edge: a payload plus labels, with a stable id |
| Node | One step of a workflow: an id, a plugin type and a configuration |
| Output port | A named output of a node. Edges can listen to one port. error is built in |
| Payload | The data inside a message |
| Placement | Running a processor node on a remote worker pool instead of the server |
| Plugin | The implementation behind a node type: a trigger, a processor or a terminal |
| Processor | A node that transforms, routes or calls out |
| Reviewer | A policy hook that can refuse any command before it runs |
| Scheduler | The part of the server that grants slots to executions and queues or refuses the rest |
| Session | A scope that groups workflows, executions and configuration |
| Signal | A named message sent to a waiting node of a running execution, such as an approval |
| Slot | The right to run one execution under the concurrency limits |
| SSE | Server-sent events: a one-way live stream to browsers, scripts and agents |
| Supervisor | A program that watches the event feed of all executions and returns commands |
| Terminal | A node that ends a branch: it consumes messages and has no outputs |
| Trigger | A node that starts a workflow: it has no inputs and emits messages |
| Valve | The gate in front of each node that implements pause, step, skip and breakpoints |
| Workflow | A directed acyclic graph of nodes and edges, stored once and run many times |
| Worker | A separate process that runs placed nodes for the server |
Reference
Glossary
Terms used across the Yukta documentation.