Pipeline Dependencies
Pipeline dependencies let you model upstream/downstream relationships between pipelines within the same environment. VectorFlow uses these relationships to warn you at deploy time, visualize data flow, and prevent configuration mistakes.
What are dependencies?
A dependency means "Pipeline B depends on Pipeline A." In this relationship:
- Pipeline A is the upstream -- it produces data that Pipeline B consumes.
- Pipeline B is the downstream -- it relies on Pipeline A being deployed and running.
Dependencies are always within a single environment. You cannot create cross-environment dependencies.
Creating a dependency
Open the pipeline editor
Navigate to the pipeline you want to add a dependency to and open it in the editor.
Open the Dependencies tab
In the pipeline detail panel, click the Dependencies tab to see existing upstream dependencies.
Add a dependency
Click Add Dependency and select the upstream pipeline from the dropdown. Only pipelines in the same environment are shown as candidates.
Save
The dependency is saved immediately. You can optionally add a description to document why the relationship exists.
Adding or removing dependencies requires the Editor role or above on the team.
Removing a dependency
From the Dependencies tab, click the remove button next to any upstream dependency. The relationship is removed immediately.
DAG visualization
The Dependency Graph page (accessible from the environment sidebar) renders all pipelines and their dependencies as a directed acyclic graph (DAG). Each node in the graph represents a pipeline, and edges show the upstream-to-downstream flow.
The graph helps you:
- Understand the overall data flow across your environment
- Identify pipelines with many dependents (high fan-out)
- Spot isolated pipelines with no relationships
Pipeline nodes are color-coded by status: deployed pipelines appear in their normal color, while draft (undeployed) pipelines are visually distinct.
Deployment order implications
VectorFlow uses dependencies to surface warnings during deployment:
- Deploying a downstream pipeline -- If any of its upstream dependencies are not yet deployed (still in draft), VectorFlow shows a warning listing the undeployed upstreams. You can still proceed, but the warning helps you avoid deploying a consumer before its producer is ready.
- Undeploying an upstream pipeline -- If any downstream pipelines are currently deployed and depend on the pipeline you are undeploying, VectorFlow warns you that those pipelines may lose their data source.
These are advisory warnings only -- they do not block deployment. The goal is to keep you informed so you can coordinate rollouts across related pipelines.
Cycle prevention
VectorFlow enforces that the dependency graph must be a DAG (directed acyclic graph). When you attempt to add a dependency that would create a circular loop, the operation is rejected with an error:
Cannot add: would create a circular dependency
For example, if Pipeline A depends on Pipeline B, and Pipeline B depends on Pipeline C, you cannot make Pipeline C depend on Pipeline A because that would form a cycle (A -> B -> C -> A).
The cycle detection algorithm runs automatically whenever you add a new dependency. It traverses the existing graph to verify that no loop would be created before persisting the relationship.
Self-references are also prevented -- a pipeline cannot depend on itself.