Library
The library uses ESM imports and has no runtime dependencies. It supports Node.js 22.6+ on Linux and macOS.
Tubeless
Typed pipelines for Node.js. Define dependencies, preview execution, and inspect step results.
Inspired by cycling, designed to keep work rolling with fewer flats and less resistance.
$ npm install tubeless
$ bunx tubeless plan ./examples/minimal-pipeline.ts --target normalize --explain
Pipeline minimal: plan (ok=true, dryRun=false, steps=2) - load: run (required by normalize for target normalize) - Read rows supplied by the caller. - normalize: run (target normalize) - Trim rows and remove empty entries.
Runtime
Build a pipeline with createSteps, step, and definePipeline.
Preview the selected steps with plan(), then execute with
run() or runOrThrow(). The CLI uses the same pipeline definitions.
CLI
Inspect, preview, and run pipelines from the terminal. The CLI requires Bun 1.3.14 or later.
Execution
Pipelines run in your Node.js process. No separate server is required.
Define
Connect steps through their inputs and outputs. TypeScript checks the values passed between them, and Tubeless checks for missing dependencies, cycles, and duplicate IDs before execution.
import { createSteps, definePipeline } from "tubeless";
interface ImportOptions {
lines: readonly string[];
}
const { step } = createSteps<ImportOptions>();
const load = step("load", {
run: (_inputs, context) => context.options.lines,
});
const normalize = step("normalize", {
dependsOn: [load],
run: ({ load: rows }) =>
rows.map((row) => row.trim().toLowerCase()).filter(Boolean),
});
export const ImportPipeline = definePipeline({
id: "import",
steps: [load, normalize],
});Plan
plan() lists the steps selected for a target without
executing them. Export the dependency graph as Mermaid with
toMermaid().
flowchart LR
required optional input + failure gate
View Mermaid source
flowchart LR step0["build-artifact"] step1["validate-artifact"] step2["publish-artifact"] step0 --> step1 step0 --> step2 step1 -. optional input + failure gate .-> step2
Run
run() returns step results, failures, and timings.
The terminal reporter displays progress during execution.
Run history can be stored in SQLite or exported as NDJSON.
$ bunx tubeless run ./live-tui.ts
Discover Sources catalogs 4/4
Fetch Records sources 4/4
Transform Rows rows ready
Publish Artifacts files 3/3
Studio
Open recorded runs in your browser to inspect step timelines, logs, errors, and child pipelines. Register commands to preview and launch new runs.
Use cases
See how Tubeless fits into release scripts, data jobs, and maintenance tasks.
CI/CD
Run the same release logic locally and in your CI runner.
ML flows
Coordinate dataset preparation, training jobs, and evaluation.
LLM workflows
Retrieve context, call a model, and validate its response.
Data pipelines
Import records, enrich them through an API, and write the results.
Operational workflows
Run tenant setup, migrations, and other maintenance commands.