ZSL

ZSL is the declarative language for defining a DOSS application: its tables, views, workflows, permissions, and UI. The `zsl` CLI reconciles a desired state in a git repo against a live environment via plan-and-apply — identical in spirit to Terraform, but operating on business application primitives instead of cloud infrastructure. Everything downstream (ZebraDB tables, ZFlow workflows, Vapor-generated UI) is compiled from ZSL source.

Four layers, strictly dependent

1. 01

   Layer 0 — Schema

   Business Objects. Tables, columns, types, lookups, formulas. The data substrate every other layer builds on.

2. 02

   Layer 1 — Projections

   Views, joins, and queries over the Schema. Compiled with strongly-typed refs to Layer 0 so changes propagate at plan time.

3. 03

   Layer 2 — Workflows

   Business logic, automations, and triggers. Consumes Projections and Schemas; emits into ZFlow as durable functions.

4. 04

   Layer 3 — Interfaces

   UI components and layouts. Consumes Projections; triggers Workflows. Vapor takes over from here to generate the actual applet.

Declaring a table

Plain TypeScript using the `T` helper. No DSL, no YAML — just typed objects.

import { T } from "../types";
import { Users } from "./users";

export const Vendors = {
  slug: "vendors",
  displayName: "Vendors",
  columns: {
    company: T.string({ required: true, description: "Legal company name" }),
    isActive: T.boolean({ nullable: false, description: "Is vendor active?" }),
    primaryContact: T.lookup({
      target: () => Users,
      description: "The primary contact for this vendor",
    }),
    statusLabel: T.formula({
      formula: (refs) => `IF(${refs.isActive}, "Active", "Inactive")`,
      description: "A computed label for the status.",
    }),
  },
};

plan → apply → pull

1. 01

   zsl plan

   Loads `.zsl-state/state.json`, compiles desired state from source, reconciles UUIDs, diffs against live, writes `plan.json`.

2. 02

   zsl apply

   Reads `plan.json`, asks for confirmation, executes CREATE/UPDATE/DELETE against the live DOSS environment, then auto-pulls to refresh state.

3. 03

   zsl pull

   Fetches the live configuration back into `.zsl-state/state.json` so every teammate plans against the same reality.

Why TypeScript, not YAML

In the stack