cutip 2.20.0

Workflow automation framework — define infrastructure as YAML, automate with Python, execute in containers

CUTIP

Automate multi-step operations using containers as reproducible execution environments.

CUTIP is a workflow automation framework. You define container infrastructure as typed YAML artifacts, write orchestration logic in Python, and CUTIP handles the lifecycle: pre-build file generation, container startup ordering, health-check loops, and post-deployment verification. Every artifact is validated statically before any container runtime is contacted.

Containers are the medium, not the goal. The task is the goal — SSH into a VM, configure a service, patch a deployment, start a dev environment. CUTIP makes that operation structured, validated, and reproducible.

What cutip is NOT

• Not a Docker replacement. cutip uses Docker or Podman as backends. It does not build images differently or replace docker run.
• Not a Kubernetes orchestrator. cutip manages local container workloads. For K8s operations, use cutip-blocks which provides kubectl blocks over SSH.
• Not a CI/CD system. cutip runs workflows locally or in containers. It complements CI — it doesn't replace GitHub Actions or Jenkins.
• Not a Docker Compose replacement for simple stacks. If you're running postgres + redis + your app with no custom startup logic, use Compose. cutip solves a different class of problems.

Who is this for

• DevOps engineers automating VM or Kubernetes operations that span multiple tools (SSH, kubectl, REST APIs)
• Teams that outgrew bash scripts but don't need the weight of Ansible
• Developers who want reproducible multi-step procedures with validation, hooks, and visual inspection
• Anyone who needs to generate config files, exec into containers during startup, or branch on health state

The model

ImageCard   ─┐
NetworkCard ─┘──>  ContainerCard  ──>  Unit  ──>  Group  ──>  workflow.py

Layer	What it is
Card	A YAML file defining one container resource (image, network, or container config). Validated by Pydantic v2.
Unit	A named container instance — references one ContainerCard + optional pre/post hooks.
Group	A collection of Units + a workflow.py. The executable artifact: cutip run <group>.
Workflow	A Python function main(ctx) that starts containers, runs health checks, and orchestrates. Full Python — no DSL.

Every artifact is a versioned YAML file (apiVersion: cutip/v1). Every ref is resolved and validated before any backend is contacted.

Install

pip install cutip

Quick start

cutip init              # scaffold a workspace with example projects
cutip validate          # static validation — no container runtime needed
cutip plan hello-world  # dry-run — print what would happen
cutip run hello-world   # validate → connect → execute workflow

Before and after

Before (bash script):

#!/bin/bash
ssh root@$VM_IP "kubectl exec -n prod deploy/web -- cat /opt/config/handler.py > /root/patches/handler.py"
ssh root@$VM_IP "sed -i 's|get_host(req)|localhost:4000|g' /root/patches/handler.py"
ssh root@$VM_IP "chmod 777 /root/patches/handler.py"
ssh root@$VM_IP "kubectl get deployment -n prod web -o yaml | python3 -c 'import sys,yaml; ...' | kubectl apply -f -"
ssh root@$VM_IP "kubectl rollout status deployment -n prod web --timeout=120s"
# Hope nothing went wrong. No validation. No rollback. No visibility.

After (cutip workflow):

@orchestrator
def main(ctx):
    container.start(ctx, container="ops-runner")

    with ssh.session(ctx, container="ops-runner",
                     host=vm["ip"], username="root", password=pw) as sesh:

        stage("Pre-op Validation")
        ssh.probe(ctx, sesh)
        k8s.get_deployment(ctx, sesh, namespace="prod", deployment="web")

        stage("Operations")
        patch_handler(ctx, sesh)
        apply_deploy(ctx, sesh)

        stage("Post-op Validation")
        k8s.rollout_status(ctx, sesh, namespace="prod", deployment="web")

    container.stop(ctx, container="ops-runner")

Every step is a named @action. The workflow is validated before execution. Failed steps report which action and stage broke. The entire pipeline is visible as a DAG in CUTIP Desktop.

How it compares

	Bash script	Docker Compose	Ansible	CUTIP
Validation	None	Runtime only	YAML lint	Static graph validation — no runtime needed
Reproducibility	Hope-based	Image tags	Playbook idempotency	Deterministic: same YAML + same workflow = same result
Startup ordering	Sequential commands	depends_on with health poll	Task ordering	Python: exec into container, branch on result
Pre-build hooks	Makefile target	None	None	pre_build(ctx) per unit
Post-start hooks	None	None	Handlers (limited)	startup(ctx) per unit
Secrets	.env or hardcoded	.env flat substitution	Vault / vars	secrets.yaml with fail-fast validation
Visual inspection	set -x	None	--verbose	DAG visualization in CUTIP Desktop
Migration	Manual	—	Manual	cutip from-compose converts any compose file

Ecosystem

Project	Description
cutip	Core framework — YAML artifacts, Python workflows, CLI
cutip-blocks	Reusable workflow blocks — SSH, kubectl, file ops, container lifecycle (34 blocks, 9 categories)
cutip-desktop	Visual companion — DAG visualization, artifact inspection, container management, observability

CLI

Command	Description
cutip init	Scaffold workspace with example projects
cutip validate	Static graph validation (no runtime needed)
cutip plan <group>	Dry-run: print execution table
cutip run <group>	Validate, connect, execute workflow
cutip tree	Print workspace artifact tree
cutip show <type> <name>	Dump a resolved artifact
cutip from-compose <file>	Convert a Docker Compose file to cutip artifacts
cutip secrets set/list/check	Manage secrets.yaml

Documentation

joshuajerome.github.io/cutip

• Why CUTIP? — detailed comparison with Docker Compose
• Concepts — cards, units, groups, lifecycle, graph resolution
• Quickstart — end-to-end walkthrough
• Use Cases — real-world examples

License

MIT