dbt-sqlx

🚀 dbt-sqlx

A CLI to convert SQL models across database dialects in your dbt projects.

🔍 Why dbt-sqlx?

Tired of rewriting SQL logic every time your data platform changes?

Whether you're:

• Migrating from one SQL type to another type like Snowflake to Redshift and many more
• Porting models between versions such as oracle 11g to oracle 19c.
• Maintaining compatibility across clouds

dbt-sqlx automates the hard part — letting you focus on insights, not syntax.

✨ Features

✅ Translate dbt models across supported SQL dialects. ✅ Translate dbt models across supported SQL dialects
✅ Retains dbt Jinja templating: {{ ref('...') }}, {{ var('...') }}
✅ Bulk model conversion support
✅ Intuitive CLI: dbt run -m-like syntax
✅ LLM-powered translation via OpenAI, Groq, Google, etc.
✅ Fully configurable through CLI or .env
✅ SQL version-aware translation (e.g., Oracle 11g vs 19c)
✅ Auto-detects source dialect from dbt metadata

📋 Pre-requisite

• Python 3.10+

• A dbt project with models

• API key and model name for one of the following providers:

  • Groq — Free Tier
  • OpenAI — Paid
  • Anthropic — Paid
  • Google-GenAI-Free Tier
  • MistralAI-Free Tier

🔧 CLI Commands

dbt-sqlx provide two main method config and trasnpile. Both method support multiple options. Below are the details:

  dbt-sqlx --help

1. config

Set or update default LLM provider, model and Key. It store configuration at ~/.dbt-sqlx/.env.

dbt-sqlx config --help

🛠️ CLI Command Options (config)

Option	Required?	Description	Default Value
--llm-provider	⚪ Optional	Set or update the default LLM provider	Not set
--llm-model	⚪ Optional	Set or update the default LLM model	Not set
--api-key	⚪ Optional	Provide or update your provider API key	Not set

📌 Example:

Prompt

dbt-sqlx config

Output

Updating dbt-sqlx environment settings...
Select model provider:
  1. OpenAI
  2. Groq
  3. Anthropic
  4. Mistral
  5. Cohere
  6. Google
  7. Azure
Enter your choice (1 to 7): 1
Enter the model name (e.g., gpt-4o, mixtral-8x7b): gpt-4o
The provider OpenAI API Key already configured, Do you want to overwrite? [Y-Yes, N-No]: Y
Enter API key for OpenAI: 
Successfully configured below configuration:
Default Provider -> OpenAI
Default LLM Model -> gpt-4o
Default Provider API Key -> sk-proj-******************************ht4GS5YA

Single Command

dbt-sqlx config --llm-provider OpenAI --llm-model gpt-4o --api-key sk-xxxxxxxxxx

2. transpile

Convert dbt models to the target dialect. It create new directory named as models_target_sql in your dbt project to avoid unintentially overwrite existing models.

dbt-sqlx transpile --help

Options

🛠️ CLI Command Options (transpile)

Option	Required?	Description	Default Value
--target-sql	🟢 Required	Target SQL dialect (e.g., oracle, snowflake, redshift)	—
--target-sql-version	⚪ Optional	Target SQL version (e.g., 11g, 19c for Oracle)	latest
--source-sql	⚪ Optional	Source SQL dialect (auto-detected if omitted)	Auto-detected
--dbt-project	⚪ Optional	Path to your dbt project	Current directory (pwd)
--models	⚪ Optional	Comma-separated list of specific dbt models to transpile	All models
--llm-provider	⚪ Optional	Override default LLM provider (e.g., OpenAI, Groq)	Configured provider
--llm-model	⚪ Optional	Override default LLM model	Configured model
--verbose	⚪ Optional	Enable logging of LLM Provider and Model during execution	False

📌 Example:

Below is the exmaple of transpile specific models dim_customer & dim_order of the dbt project named as dbt-ecom into Oracle.

dbt-sqlx transpile --target-sql oracle --dbt-project ~/dbt/dbt-ecom/ --models dim_customer,dim_order

⚡Quick Start

📦 Installation

Install the dbt-sqlx from PyPI.

pip install dbt-sqlx

✅ Verify Installation

dbt-sqlx --version

output

dbt-sqlx version x.x.x

⚙️ Configuration

Set up your default LLM provider, model, and API key:

dbt-sqlx config

You'll be prompted to enter:

• LLM Provider (e.g., OpenAI, Groq)
• Model Name (e.g., gpt-4, mixtral)
• API Key (input hidden for security)

Alternatively, you can use one line command to configure default Provider and Model:

dbt-sqlx config --llm-provider your-llm-provider --llm-model your-llm-model --api-key your-api-key

# Example 
dbt-sqlx config --llm-provider Groq --llm-model llama-3.3-70b-specdec  --api-key ] gsk_ob**********LhiB

🚀 Usage

Convert all dbt Project's models

dbt-sqlx transpile --target-sql your-sql-type --dbt-project /path/to/dbt-project

# Example
dbt-sqlx transpile --target-sql oracle --dbt-project /path/to/dbt-project

🎯 Convert Specific Models

dbt-sqlx transpile --target-sql snowflake --dbt-project /path/to/project --models model1,model2

🎥 Demo

Check out dbt-sqlx in action! 👇

Blog

Check out dbt-sqlx in blog! 👇

GenAI + dbt = dbt-sqlx: The Easiest Way to Switch SQL Dialects 💫

🎯 Use Cases

🧾 Input (Snowflake SQL):

SELECT
    user_id,
    first_name,
    CURRENT_TIMESTAMP AS refreshed_at
FROM {{ ref('dim_customers') }}

dbt-sqlx transpile --target-sql redshift --dbt-project your-dbt-project-path

🔁 Output (Redshift):

SELECT
    user_id,
    first_name,
    GETDATE() AS refreshed_at
FROM {{ ref('dim_customers') }}

🧾 Input (Snowflake SQL):

SELECT customer_id,
       LISTAGG(DISTINCT first_name, ', ') WITHIN GROUP (ORDER BY first_name) AS customers
FROM {{ ref('dim_customers') }}
GROUP BY customer_id;

dbt-sqlx transpile --target-sql oracle --target-sql-version 11g --dbt-project your-dbt-project-path

🔁 Output (Oracle 11g):

SELECT customer_id,
       RTRIM(XMLAGG(XMLELEMENT(e, first_name || ', ') ORDER BY first_name).EXTRACT('//text()'), ', ') AS customers
FROM (
    SELECT DISTINCT customer_id, first_name
    FROM {{ ref('dim_customers') }}
) 
GROUP BY customer_id;

dbt-sqlx transpile --target-sql oracle --target-sql-version 19c --dbt-project your-dbt-project-path

🔁 Output (Oracle 19c):

SELECT customer_id,
       LISTAGG(first_name, ', ') WITHIN GROUP (ORDER BY first_name) AS customers
FROM (
    SELECT DISTINCT customer_id, first_name
    FROM {{ ref('dim_customers') }}
) subquery
GROUP BY customer_id;

Sample Configuration

Below are some sample configuration of LLM providers and models:

Groq

LLM_Provider = "Groq"
LLM_Name = 'llama-3.3-70b-versatile'
LLM_Provider_Key = 'gsk_*************************TLhiB'

Open AI

LLM_Provider = "OpenAI"
LLM_Name = 'gpt-4o'
LLM_Provider_Key = sk-proj-*****************************5YA

Google GenAI

LLM_Provider = "Google_Genai"
LLM_Name = 'gemini-2.0-flash'
LLM_Provider_Key = 'AI******************************7k'

Mistral AI

LLM_Provider = "MistralAI"
LLM_Name = 'mistral-small-latest'
LLM_Provider_Key = 'a2**************************ya0'

🧪 Supported Dialects (so far)

Here’s what’s currently supported dialect, use it as --target-sql values

• Redshift
• Snowflake
• BigQuery
• Postgres
• MySQL
• Oracle
• Spark-SQL
• SQL-Server
• Db2
• ClickHouse
• DuckDB
• Databrick-Sql
• Trino
• Vertica
• Athena
• Presto
• Google-Alloy-DB
• MariaDB
• Presto
• Apache-Hive
• SQLite

⚠️ Important Notes

• dbt-sqlx uses LLM models — do not use if your code is under strict data security policies.
• Accuracy may vary depending on the LLM — always review and test translated code.
• It does not overwrite original models. Output is stored in a direcotry named as models with suffix target SQL type 'models_<target_SQL>' like models_oracle/.

📄 License

This project is licensed under the MIT License – see the LICENSE file for details.

📬 Contact

👨‍💻 Author: Nikhil Suthar
📧 Email