[Regression] Primary Key Constraint applied without Contract

[kmarq]

Is this a regression in a recent version of dbt-core?

• I believe this is a regression in dbt-core functionality
• I have searched the existing issues, and I could not find an existing issue for this regression

Current Behavior

Executing on dbt-cloud.

Model Primary key constraint is populating without the enforce_contract set to true and a full model contract was provided.

Example model:

• name: pk_example
  columns:
  - name: primary_key_column
  description: My column
  tests:
  - not_null
  - unique
  constraints:
  - type: primary_key
  columns: [primary_key_column]

This fails because it tries to apply the PK constraint but in this case we didn't actually also set the required not_null constraint.

{"app": "dbt", "dbt_version": "2026.1.26+c924ee8", "dbt_databricks_version": "1.11.4post1+ec229f85e80de8737ba308b58059fef56cf9e3fd", "databricks_sql_connector_version": "4.1.3", "profile_name": "user", "target_name": "prod", "connection_name": "master"}

Expected/Previous Behavior

Previous behavior, the primary_key constraint is not applied without having enforce_contract: true (see further examples here feature request where I had actually requested this capability be added).

The PK constraint was intended to be applied but previous behavior was that it wasn't without a full contract. This caused our jobs to fail because we had not set non-nullable on the PK fields, but had never needed to because the constraint was never actually applied.

Prior successful runs that didn't apply the PK
{"app": "dbt", "dbt_version": "2026.1.19+d634a9b", "dbt_databricks_version": "1.11.3post1+cd5d71f50cd9f003e2f879dc6729a3aa93e1cdaf", "databricks_sql_connector_version": "4.1.3", "profile_name": "user", "target_name": "prod", "connection_name": "master"}

Steps To Reproduce

1. dbt cloud on latest

2. define a model with a PK constraint but do not set enforce_contract

3. Run - model will be created with the PK constraint

4. dbt cloud compatible version

5. run same model - the PK constraint will not be added

Relevant log output

Environment

- dbt (working version): 2026.1.19+d634a9b
- dbt (regression version):2026.1.26+c924ee8

- dbt-databricks (working): 1.11.3post1+cd5d71f50cd9f003e2f879dc6729a3aa93e1cdaf
- dbt-databricks (regression): 1.11.4post1+ec229f85e80de8737ba308b58059fef56cf9e3fd

Which database adapter are you using with dbt?

other (mention it in "Additional Context")

Additional Context

I'm mainly raising because I do not see any release notes here or on the dbt-databricks adapter that lead me to believe this was an intended change. I've actually requested this behavior Feature request, and so am glad if this was intended but also want to make sure it doesn't just get removed since it wasn't.

[jeremyyeo]

I was not able to reproduce this... full project:

# dbt_project.yml
name: analytics
profile: all
version: "1.0.0"

models:
  analytics:
    +materialized: table

# models/schema.yml
models:
  - name: foo
    constraints:
      - type: primary_key
        columns: [id]
    columns:
      - name: id
        data_type: int

-- models/foo.sql
select 1 id

$ dbt build

18:18:01 Began running node model.analytics.foo
18:18:01 1 of 1 START sql table model dbt_jyeo.foo ...................................... [RUN]
18:18:01 Databricks adapter: DatabricksDBTConnection(session-id=None, name=model.analytics.foo) - Creating connection
18:18:01 Acquiring new databricks connection 'model.analytics.foo'
18:18:01 Began compiling node model.analytics.foo
18:18:01 Writing injected SQL for node "model.analytics.foo"
18:18:01 Began executing node model.analytics.foo
18:18:01 MATERIALIZING TABLE
18:18:01 Writing runtime sql for node "model.analytics.foo"
18:18:01 On model.analytics.foo: /* {"app": "dbt", "dbt_version": "2026.1.26+c924ee8", "dbt_databricks_version": "1.11.4post1+ec229f85e80de8737ba308b58059fef56cf9e3fd", "databricks_sql_connector_version": "4.1.3", "profile_name": "user", "target_name": "default", "node_id": "model.analytics.foo"} */

  
    
        create or replace table `dev`.`dbt_jyeo`.`foo`
      
      
    using delta
  
      
      
      
      
      
      
      
      as
      select 1 id
  
18:18:01 Opening a new connection, currently in state init
18:18:01 Databricks adapter: Connection(session-id=01f0fc75-ae31-122f-b498-937762d06c8a) - Created
18:18:09 SQL status: OK in 8.090 seconds
18:18:09 Databricks adapter: Cursor(session-id=01f0fc75-ae31-122f-b498-937762d06c8a, command-id=01f0fc75-ae38-106d-8067-b9f12f07738f) - Closing
18:18:09 Applying tags to relation None
18:18:09 On model.analytics.foo: Close
18:18:09 Databricks adapter: Connection(session-id=01f0fc75-ae31-122f-b498-937762d06c8a) - Closing
18:18:09 1 of 1 OK created sql table model dbt_jyeo.foo ................................. [�[32mOK�[0m in 8.36s]
18:18:09 Finished running node model.analytics.foo

[jeremyyeo]

@kmarq - try deleting everything in your dbt project - such that it only represents the thing you're testing (like me above, simple dbt_project.yml + schema.yml + model.sql file). Perhaps the contract enforcement is set from somewhere else - e.g. dbt_project.yml file.

[kmarq]

I was able to replicate this with a simplified model. I believe I narrowed this down to what seems to be a change in the behavior of the persist_docs option. It also seems to only matter when it is an incremental model. A full refresh does not have issues.

So no contract enforcement enabled anywhere

persist_docs:
relation: "True"
columns: "True"

Running on Compatible mode in dbt cloud the model executes without issue. On Current, it fails due to the nullable primary key field.

dbt_project.yml
models:
databricks_analytics_dbt:
+on_schema_change: "append_new_columns"
+persist_docs:
relation: True
columns: True

model:
{{
config(
materialized='incremental',
unique_key = ['seqnbr', 'insertdts'],
)
}}
select
cast(seqnbr as decimal(18,0))seqnbr,
lower(trim(clientid))clientid,
lower(trim(recordid))recordid,
cast(insertdts as date)insertdts,
from source('finance', 'src_financial_accounting_gl_detail')

yaml:
models:

• name: c_financial_accounting_gl_detail
  tests:

  • dbt_utils.unique_combination_of_columns:
    combination_of_columns:
    - seqnbr
    - insertdts

  constraints:

  • type: primary_key
    columns: [seqnbr,insertdts]

Database Error in model c_financial_accounting_gl_detail
Cannot create the primary key c_financial_accounting_gl_detail_pk because its child column(s) seqnbr is nullable.