contracts

dbt-forge contracts generate creates dbt data contracts (v1.5+) by introspecting column types from your warehouse. Use it to enforce schema stability on public models without writing contract YAML by hand.

Command

dbt-forge contracts generate [MODEL] [--all-public] [--dry-run] [--yes] [--target TARGET]

What it does

The command connects to your warehouse, reads column names, data types, and nullability for each model, then updates the model’s YAML file with:

• contract: { enforced: true } in the model config
• data_type on each column
• not_null test for non-nullable columns (if not already present)

Existing descriptions, tests, and other YAML content are preserved.

Arguments

MODEL

Generate a contract for a specific model.

dbt-forge contracts generate orders

Options

--all-public

Generate contracts for all models with access: public defined in YAML.

dbt-forge contracts generate --all-public

--dry-run

Preview the generated YAML without writing to disk.

dbt-forge contracts generate orders --dry-run

--yes, -y

Auto-accept all generated contracts without interactive review.

dbt-forge contracts generate --all-public --yes

--target

dbt profile target for the warehouse connection. Defaults to dev.

dbt-forge contracts generate orders --target prod

Interactive review

When run without --yes, the command presents each generated contract for review. You can accept or skip each model, following the same pattern as docs generate.

Generated output

For a model orders with columns id (INTEGER, NOT NULL), amount (NUMERIC, NULL), and created_at (TIMESTAMP, NOT NULL):

version: 2
models:
  - name: orders
    config:
      contract:
        enforced: true
    columns:
      - name: id
        data_type: INTEGER
        data_tests:
          - not_null
      - name: amount
        data_type: NUMERIC
      - name: created_at
        data_type: TIMESTAMP
        data_tests:
          - not_null

Finding public models

The --all-public flag scans all YAML files under models/ for models with access: public (either at the top level or inside config:). This follows the dbt Mesh convention where public models are the contract boundary between projects.

Behavior and limits

• Must run from inside an existing dbt project (walks up to find dbt_project.yml).
• Reads connection credentials from profiles.yml using the introspect module.
• Requires the model to exist in the warehouse (the table/view must be queryable).
• Schema is inferred from the model’s directory path (e.g., models/staging/ uses schema staging).
• If the YAML file already exists, existing content (descriptions, tests, other models) is preserved.
• If the YAML file does not exist, a new file is created with version: 2.
• The not_null test is only added for non-nullable columns and only if not already present.
• Supported warehouses: all 8 adapters (BigQuery, Snowflake, PostgreSQL, DuckDB, Databricks, Redshift, Trino, Spark).