prefect-deployments-toolkit 0.0.5

CLI toolkit for detecting and applying Prefect deployment changes in CI/CD pipelines

Prefect Deployments Toolkit

A lightweight CLI toolkit for managing Prefect deployment lifecycles in CI/CD pipelines. It detects deployments that were added, modified, or removed between two git references, and applies (creates, updates, or deletes) those deployments against Prefect Cloud or a self-hosted Prefect server — via either the Prefect CLI or direct REST API calls.

Built for teams running many Prefect flows out of a monorepo, where each deployment is defined in its own YAML file alongside a shared prefect_base.yaml, and CI needs to figure out which deployments actually changed before re-registering them.

Features

• Change detection — diff deployments between two git refs (a merge request source vs. target branch, or HEAD~1 vs. HEAD after a merge) and get back exactly which deployments were added, modified, or removed.
• Unified push / pull-request logic — one code path handles both pull_request (compare local files against a fetched base branch) and push/pull_request_target (compare HEAD~1 against HEAD) event styles.
• Dual deploy backends — apply deployments using the prefect CLI as a subprocess, or bypass it entirely with direct Prefect Cloud REST API calls for faster, more controllable deploys.
• Dev/prod environment overrides — automatic name prefixing, work pool substitution, and schedule pausing for dev environments.
• Concurrent deployment application — apply multiple deployments in parallel with clean, non-interleaved log output per deployment.
• Duplicate deployment name detection — warns (or optionally cleans up) when the same deployment name resolves to more than one flow.

Installation

pip install prefect-deployments-toolkit

Requires Python >= 3.10.

Usage

1. Detect modified deployments

python -m prefect_deployments_toolkit.get_modified_deployments \
  --modified-by push \
  --base-ref main \
  --deployments-dir deployments

Prints a comma-separated list of changed deployment names to stdout, and (if running inside GitHub Actions) writes DEPLOYMENT_NAMES, NEW_OR_MODIFIED_DEPLOYMENT_NAMES, and REMOVED_DEPLOYMENT_NAMES to GITHUB_ENV.

--modified-by accepts:

• pull_request — compares your local working tree against the fetched base branch (use during an open MR/PR).
• push or pull_request_target — compares HEAD~1 against HEAD (use after a merge to a protected branch).

2. Apply deployments

python -m prefect_deployments_toolkit \
  --deployment-names "flow_a,flow_b" \
  --enable-schedule false \
  --tag dev \
  --reference feature-branch-1 \
  --repo-name my-flows-repo \
  --custom-image "" \
  --deployments-dir deployments \
  --dev-work-pool my-dev-work-pool \
  --backend rest \
  --enforce-unique-deployment-names false

Expected Repository Layout

deployments/
├── prefect_base.yaml # shared config merged into every deployment
├── flow_a.yaml
├── flow_b.yaml
└── ...

Each deployment YAML file must declare a deployments: list with at least a name and entrypoint key, following the standard Prefect deployment YAML schema.

Backends

Backend	How it works	When to use
cli	Shells out to prefect deploy	Simplest, matches local prefect deploy behavior exactly
rest	Talks directly to the Prefect Cloud REST API	Faster for many concurrent deployments, avoids CLI subprocess overhead

Environment Variables

Variable	Required	Purpose
PREFECT_API_URL	Yes	Full Prefect Cloud/server API base URL
PREFECT_API_KEY	Yes	API key for authentication

License

Licensed under PolyForm Noncommercial 1.0.0. You may view, use, and modify this software for noncommercial purposes. Commercial use — including building products, services, or automation on top of it — requires a separate license. Contact [e-mail] for commercial licensing inquiries.

Contributing

Issues and pull requests are welcome for noncommercial use cases. Please open an issue before submitting a large PR to discuss scope.