dbt

Name: dbt
Rating: 92
Author: clawdata

You can build and manage dbt data transformations using the clawdata CLI. Use this when the user wants to create, run, or test data models — not for ad-hoc queries.

For querying the database directly, use the duckdb skill. For orchestration, use the airflow skill.

Commands

Task	Command
Run all models	`clawdata dbt run`
Run specific models	`clawdata dbt run --models slv_customers gld_customer_analytics`
Run data quality tests	`clawdata dbt test`
Test specific models	`clawdata dbt test --models slv_orders`
Compile SQL (no run)	`clawdata dbt compile`
Load seed CSVs	`clawdata dbt seed`
Generate documentation	`clawdata dbt docs`
Debug dbt connection	`clawdata dbt debug`
List available models	`clawdata dbt models`
Health check	`clawdata doctor`
Task status	`clawdata status`

Append --json to any command for structured JSON output.

When to use

•User wants to transform data → clawdata dbt run
•User wants to validate data quality → clawdata dbt test
•User wants to create a new model → write the .sql file, then clawdata dbt run
•User asks about model lineage or available models → clawdata dbt models
•Something isn't working → clawdata dbt debug

Creating new models

•Bronze models → dbt/models/bronze/ (materialised as views — raw ingestion)
•Silver models → dbt/models/silver/ (materialised as views — cleaned & validated)
•Gold models → dbt/models/gold/ (materialised as tables — business aggregates)
•Update dbt/models/schema.yml to add tests and docs

Example silver model (dbt/models/silver/slv_sales.sql):

sql

SELECT
    order_id,
    customer_id,
    amount::DECIMAL(10,2) AS amount,
    order_date::DATE      AS order_date
FROM {{ ref('brz_orders') }}
WHERE amount > 0

After creating models, run clawdata dbt run to build them.

Typical workflow

•
Check existing models:
bash
```
clawdata dbt models
```
•
Run transformations:
bash
```
clawdata dbt run
```
•
Validate with tests:
bash
```
clawdata dbt test
```

•

If something fails, debug:

bash

clawdata dbt compile    # see the generated SQL
clawdata dbt debug      # check connection

Error recovery

Problem	Fix
`dbt: command not found`	`pip install dbt-core dbt-duckdb`
`clawdata: command not found`	Rebuild: `cd clawdata && npm run build && npm link`
Connection error	`clawdata dbt debug`
Model compile error	`clawdata dbt compile` to see the SQL error, then fix the `.sql` file

Project structure

code

dbt/
├── dbt_project.yml    ← project config
├── profiles.yml       ← connection config (DuckDB default)
└── models/
    ├── schema.yml     ← tests & documentation
    ├── bronze/        ← views (raw data ingestion)
    ├── silver/        ← views (cleaned & validated)
    └── gold/          ← tables (business-ready aggregates)

Notes

•dbt writes to the same DuckDB warehouse at data/warehouse.duckdb
•Staging models are views; mart models are tables
•After dbt run, query results via the duckdb skill