Seeds & Reference Data

dbt seeds are static CSV files that dbt loads into the database as ordinary tables. nhic_dbt ships 40+ seed CSVs in the seeds/ directory covering facility mappings, administrative classifications, code lookups, and other reference data that changes infrequently.

What seeds are for

Seeds are the right choice when:

• The data is static or rarely changes (a new release is needed to update it)
• The data is small enough to commit to git as a CSV
• The data is authoritative reference data — code tables, classifications, approved facility lists

Seeds are the wrong choice when:

• The data is sourced from an external system and changes frequently — use a staging model instead
• The data is large (thousands of rows or more) — consider a staging model backed by a source table

Loading seeds

# Load all seeds
dbt seed
 
# Load a specific seed by name
dbt seed --select facility_mapping
 
# Full refresh (truncate and reload)
dbt seed --full-refresh

Seeds are loaded into the staging schema by default (configured in dbt_project.yml).

Using seeds in models

Reference a seed in a model using {{ ref() }} just like any other dbt model:

-- models/intermediate/int_visits__enriched.sql
select
    v.visit_id,
    v.facility_code,
    f.facility_name,
    f.district,
    f.province
from {{ ref('stg_cemr__visits') }} v
left join {{ ref('facility_mapping') }} f
    on v.facility_code = f.facility_code

dbt tracks this dependency in the DAG, so seeds are guaranteed to be loaded before any model that references them.

Adding or updating a seed

1. Add or edit the CSV file in seeds/.
2. If adding a new seed, declare column types in dbt_project.yml under seeds: to prevent dbt from inferring incorrect types.
3. Run dbt seed --select <seed_name> locally to verify it loads correctly.
4. Open a PR — CI will validate the seed as part of dbt build.