Quickstart

There are three ways in. Pick the one that fits what you need today.

Try the demo#

The demo on the home page runs the engine in your browser. Pick a scenario (shop, hiring or help desk), change the record count and the seed, turn masking on, and export CSV. Nothing is sent to a server, so it's a safe place to see what connected, seeded data looks like before you write a spec of your own.

Build the CLI from source#

The CLI is free on every plan and needs no account. Until the first cli-v release publishes binaries (Linux x64 and arm64, macOS x64 and arm64, Windows x64, with a SHA256SUMS file), build it from the repository. The source isn't public yet, so this step needs repository access.

cd dogm-engine
cargo build --release -p dogm-cli
./target/release/dogm --version

That prints dogm 0.1.0 for the current engine. Put target/release on your PATH, or copy the binary somewhere that is. The examples below call it dogm.

Write a spec#

Save this as shop.yaml. It describes 50 customers, up to four orders each, and a rule that every order total adds up:

spec_version: 1
locale: en-PH
tables:
  customers:
    rows: 50
    primary_key: { column: customer_id, format: "CUS-{n:04}" }
    columns:
      full_name: { gen: person.full_name, pii: person.name }
      email:     { gen: person.email, from: full_name, unique: true, pii: person.email }
      city:      { gen: ph.city, weights: population }
  orders:
    rows_per_parent: { parent: customers, dist: uniform, min: 0, max: 4 }
    primary_key: { column: order_id, format: "ORD-{n:05}" }
    columns:
      customer_id: { ref: customers.customer_id }
      quantity:    { gen: int, between: [1, 5] }
      unit_price:  { gen: choice, values: [249, 399, 1250], type: money }
      total_php:   { expr: "quantity * unit_price" }
rules:
  - "orders.total_php == orders.quantity * orders.unit_price"

The spec language reference covers every field. orders.customer_id is a foreign key to the owning parent, so it's valid by construction, and the emails use RFC 2606 example domains, so none of them can reach a real inbox.

Compile it#

dogm compile shop.yaml

spec_hash 423ed75a1229058e25651e1ebe7ef456fe488d322c9175669af701c67f40c6ac
seed      0
table              rows  chunks  parent
customers            50       1
orders              105       1  customers
rules     1

Child row counts depend on the seed, and the summary uses seed 0 unless you pass --seed. A mistake comes back with its JSON path, a stable code and, often, a hint. Misspell unit_price in the expression and you get:

/tables/orders/columns/total_php E_EXPR_UNKNOWN_COLUMN: unknown column 'unit_pric'
    hint: did you mean 'unit_price'?
1 error

The exit code is 2 for compile errors, which is what a CI step should key on. The CLI reference lists the others.

Generate and validate#

dogm generate shop.yaml --seed 4242 --out ./data

wrote 2 parts, 145 rows, to ./data
output_sha256 560ff432f35c664b14692c0f673a17590c62f0e0cbc4fdeee6356cfb51915e2f
checks 16 passed, 0 failed

./data now holds customers.part-00000.csv, orders.part-00000.csv, manifest.json, and the checks report as report.json, report.junit.xml and report.sarif. --format jsonl and --format sql write the other two formats.

dogm validate re-reads a directory and checks it against the spec again, which is how you check data that has been copied or loaded somewhere:

dogm validate shop.yaml ./data

Same seed, same bytes#

Run the same command again, on any machine, and output_sha256 comes out identical. That's the promise: the same spec, seed and row overrides give the same bytes within an engine minor version, whatever the worker count. Change the seed and you get a different, equally valid dataset. Scale it with --rows customers=5000 and the orders follow.

Hosted runs#

Sign up, let onboarding create your org and first project, start from a template (e-commerce, hiring or support) or your own spec, approve it, run it, and download the files. Each run records its seed and engine version and shows a Reproduce command that gives the same output_sha256 with the CLI. During the beta, every workspace has Pro limits at no cost.

From there, the GitHub Action runs a spec in CI, and the masking guide covers real data you already hold.