What vx is, and what it refuses to be
Every monorepo tool eventually describes itself with a paragraph of nouns: caching, task graph, remote execution, affected detection, generators, a dashboard, a cloud. vx's description is one sentence. vx runs and caches a task graph, correctly, and stops there. That sentence is a design, not a slogan, and this post walks through what it commits us to. The pipeline A run is a pipeline with five…
vx is a monorepo tool that runs and caches a task graph correctly, with five stages in its pipeline. The stages consist of discovering projects in the workspace, evaluating vx.config.ts files, building a task graph, deriving content-addressed keys, scheduling the graph, and executing tasks with bounded parallelism. A task in vx is defined as a shell command with declared inputs and outputs, and caching is opt-in, requiring both inputs and outputs to be specified.
Tasks have three rules: caching is opt-in, no cache block means no caching, and each task must have a single command. Project boundaries are hard, with no crossing of project directories. vx does not have a cloud, dashboard, daemon, auto-inferred inputs, JavaScript-function tasks, named or global inputs, or global environment variables.
Configs are TypeScript, and the core repository does not contain anything distributed. Remote caches, execution, telemetry, and agent protocols are all plugins in separate packages. The boundary matters because a tool that owns the cloud may prioritize adequacy over correctness and speed, while the vx design focuses on being correct and fast on the machine in front of you.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
