Project & Analysis Approval
Hub supports administrator approval workflows for projects and analyses. These ensure that federated computations are reviewed before execution.
Project Approval
Projects represent collaborative research efforts. When a user creates a project:
- The project enters a pending state
- An administrator reviews the project details (description, risk assessment, data requirements, selected nodes)
- The administrator approves or rejects the project
- Only approved projects can have analyses created

Analysis Approval
Analyses are distributed computations within a project. Every node an analysis runs on decides for itself whether the analysis may run there. A node approves the built image, not just the code:
- The analysis owner locks the configuration and builds the image. The build does not wait for approvals.
- Once the build has finished, each node's operators review the image (hash, size, OS) together with the code and approve or reject it. Approving before the build has finished is refused.
- Distribution starts only after every node has approved. Aggregator nodes are approved automatically.
Starting a new build resets every node's approval (and rejection), because the new image needs a fresh decision. Unlocking the configuration resets them as well.

Analysis Execution Flow
An analysis goes through:
- Configure — storage buckets are provisioned, configuration is locked
- Build — the worker builds a Docker image from the analysis code and master image
- Approve — every node approves the built image
- Distribute — the container image is pushed to target node registries
- Execute — nodes pull and run the container against their local data
- Complete — results are collected and stored
Skipping Approval (Development)
For development and testing, approval workflows can be disabled via environment variables on the Core API:
SKIP_PROJECT_APPROVAL=true
SKIP_ANALYSIS_APPROVAL=trueWARNING
Never disable approval workflows in production environments.