Migration from dbt

KelpMesh can automatically convert dbt projects to pure SQL. The import process converts Jinja macros, preserves materializations, copies seeds, and generates assertion tests from schema.yml.

Quick Start

`ash

Import a dbt project

kelpmesh import ./my-dbt-project --output ./my-kelpmesh-project

Or use the interactive migration wizard

kelpmesh migrate ./my-dbt-project --output ./my-kelpmesh-project

Then run the imported project

cd my-kelpmesh-project kelpmesh debug # validate config kelpmesh plan # check the DAG kelpmesh seed # load seed data kelpmesh run # execute all models kelpmesh test # run generated tests `

What Gets Converted

dbt Feature	KelpMesh Equivalent
{{ ref('model') }}	Plain table name (auto-resolved by AST)
{{ source('schema', 'table') }}	Plain table name
{{ config(materialized='table') }}	-- { materialized: table } header
{{ cents_to_dollars('col') }}	(col / 100)::numeric(16, 2)
{{ dbt.date_trunc('day', col) }}	DATE_TRUNC('day', col)
{{ dbt_utils.generate_surrogate_key(['a','b']) }}	generate_surrogate_key(a, b)
{{ is_incremental() }}	TRUE (use -- materialized: incremental header)
{{ var('name', 'default') }}	'default' (literal value)
{{ env_var('VAR', 'default') }}	'default' (literal value)
schema.yml tests (not_null, unique, etc.)	tests/*.sql assertion queries
Seeds (CSV)	Preserved as CSV + .sql wrapper
Sources (sources.yml)	Preserved + ephemeral model stubs
Snapshots	-- { materialized: table, strategy: snapshot }
Analyses	-- { materialized: ephemeral }
dbt_project.yml folder configs	Per-model -- materialized: headers

dbt-Compatible Macros (Runtime)

KelpMesh includes 12 dbt-compatible macros that work as plain SQL function calls (no Jinja). These are available at runtime in any kelpmesh model:

dbt Macro	KelpMesh SQL	Example
cents_to_dollars(col)	(col / 100)::numeric(16, 2)	SELECT cents_to_dollars(revenue_cents)
dollars_to_cents(col)	(col * 100)::bigint	SELECT dollars_to_cents(price)
dbt_current_timestamp()	CURRENT_TIMESTAMP	SELECT dbt_current_timestamp()
dbt_type_string()	'VARCHAR'	CAST(col AS dbt_type_string())
dbt_type_numeric()	'DECIMAL'	Type macro
dbt_type_bigint()	'BIGINT'	Type macro
dbt_type_timestamp()	'TIMESTAMP'	Type macro
dbt_type_date()	'DATE'	Type macro

Plus 32 built-in macros: surrogate_key, safe_divide, datediff, dateadd, date_trunc, haversine, median, percentile, and more.

Migration Report

After import, a MIGRATION_REPORT.md is generated in the output directory listing:

• Imported models and their materializations
• Folder-level materialization overrides (from dbt_project.yml)
• Leftover Jinja macros requiring manual review
• Next steps

Interactive Migration Wizard

ash kelpmesh migrate ./my-dbt-project

This wizard: 1. Scans the dbt project (counts models, seeds, tests, packages, snapshots) 2. Reports what was found 3. Asks for confirmation 4. Runs the full import with macro conversion 5. Generates MIGRATION_REPORT.md

Use --yes / -y for non-interactive mode (CI/CD):

ash kelpmesh migrate ./my-dbt-project --yes -o ./kelpmesh-project

dbt-Compatible Project Layout

Start a new project with a dbt-style directory structure:

ash kelpmesh init my_project --compat dbt

This creates: - models/staging/ — stg_ models (views) - models/marts/ — business marts (tables) - models/intermediate/ — intermediate transformations

Row-by-Row Comparison Against dbt

During migration, compare kelpmesh output against dbt:

`ash

First, compile dbt so target/ exists

cd ./my-dbt-project && dbt build

Then compare

kelpmesh compare orders --dbt ./my-dbt-project `

If dbt hasn't been compiled, kelpmesh will tell you to run dbt build first.

Why Migrate?

Feature	dbt	KelpMesh
Model syntax	Jinja-templated SQL	Pure SQL (no Jinja)
IDE autocomplete	Broken by Jinja	Full
AI assistant support	Broken by Jinja	Full
Column lineage	Enterprise only	Free
Schema drift detection	Manual	Built-in
Audit logging	Not included	Built-in
Security (RLS, masking, erasure)	Not included	Built-in suite
Learning curve	Weeks	Minutes
CI/CD (PR comments)	dbt Cloud only	Free, self-hosted
Price	Free-/mo	Free (core) / /mo (studio)

Manual Review Items

After import, some items may need manual attention:

1. Untranslated Jinja macros — listed in MIGRATION_REPORT.md. Run kelpmesh plan to see which models have parse errors, then replace the remaining {{ }} blocks with plain SQL.

2. dbt packages — Common macros from dbt_utils and dbt_date are auto-converted. Custom package macros need manual porting to macros/*.py or macros/*.yml.

3. Metricflow / dbt metrics — These are not auto-converted. Define metrics manually in metrics.yml.

4. Python models — dbt Python models need to be adapted to the kelpmesh def model(dbt, session) interface.