Getting Started

Ripline is a workflow engine for repeatable AI agent pipelines. You define each step explicitly, keep context isolated by default, and validate boundaries with contracts where needed.

Installation

# From the repository
git clone https://github.com/craigjmidwinter/ripline
cd ripline
npm install
npm run build

Once built, the ripline CLI is available via node bin/ripline.js or, if the package is installed globally, as the ripline command.

Hello World

The minimal pipeline has three steps: input (load JSON), transform (shape data), and output (write artifact).

Pipeline (pipelines/examples/hello-world.yaml):

id: hello_world
name: Hello World Pipeline
entry:
  - intake
nodes:
  - id: intake
    type: input
    description: Provide `person` and optional `goal` fields
  - id: enrich
    type: transform
    expression: "({ greeting: `Hello, ${inputs.person}!`, goal: inputs.goal ?? 'explore Ripline' })"
  - id: finalize
    type: output
    path: hello.result
    source: enrich
edges:
  - from: { node: intake }
    to:   { node: enrich }
  - from: { node: enrich }
    to:   { node: finalize }

Run it:

ripline run --pipeline pipelines/examples/hello-world.yaml \
  --input '{"person":"World","goal":"get started"}'

Or use the built-in demo (uses a stub agent, no configuration needed):

npm run demo

Outputs are written to .ripline/runs/<runId>/run.json and, if you pass -o <path>, to that file.

What to understand next

• Pipelines are explicit. Nodes, edges, retries, and templates are all declared in the pipeline file.
• Agent steps are isolated by default. Set resetSession: false only when you intentionally want continuity across steps.
• Contracts are optional but useful. Add JSON Schema at pipeline or node boundaries when you want stricter guarantees.

Next steps

Topic	Guide
Learn the pipeline YAML format	Pipeline Reference
Explore all CLI commands and flags	CLI Reference
Set up config files and env vars	Configuration
Connect an AI model or agent runner	Agent Integration
Set up profiles for different projects	Pipelines & Profiles
Run pipelines on a schedule	Automation & Cron
Trigger and inspect runs over HTTP	HTTP API