Pipeline
Pipeline
Also called: 流水线 · Processing pipeline
Processing stages chained in a fixed order, each stage's output feeding the next; often used for data-processing or content-generation agent tasks.
Structure
A pipeline is the simplest workflow shape: stage A → B → C, linear, no branches, fixed order. Each stage has a single responsibility and a clear input/output contract. In an agent context some stages may be one LLM call ("expand bullet points into paragraphs"), others plain code ("fetch the page," "validate format").
When it fits and when it doesn't
When a task is naturally "process in sequence" with no need to go back, a pipeline is clear and testable. Once you need branching (different paths by type), parallelism (independent sub-tasks at once), or loops (redo until a condition holds), you need a DAG workflow or an agent with autonomous decisions.
Common misconceptions
- A pipeline is a subset of workflow (the linear special case), not a separate thing alongside it.
- When stage interfaces are unstable, a pipeline's "easy to test" advantage disappears.
Related terms
Sources
Compiled 2026-08-29 · This glossary is compiled from public papers, official specifications, and common industry definitions, and is updated as the field evolves. Corrections welcome.