Merge pull request kristeligt-dagblad#36 from BrianLz/master

Add BQML recommend feature for Recommender Systems workflows

README.md

@@ -74,6 +74,17 @@ with eval_data as (
 select * from {{ dbt_ml.predict(ref('model'), 'eval_data') }}
 ```
 
+If you're using a BQML **matrix_factorization** model, you can use the recommend macro in the same way.
+```sql
+# downstream_model.sql
+
+with predict_features AS (
+    ...
+)
+
+select * from {{ dbt_ml.recommend(ref('model'), 'predict_features') }}
+```
+
 The ML.DETECT_ANOMALIES function provides anomaly detection for BigQuery ML.
 
 ```sql

macros/recommend.sql

@@ -0,0 +1,10 @@
+{% macro recommend(relation, source, threshold = none) %}
+    ml.recommend(
+        model {{ relation }},
+        (select * from {{ source }})
+
+        {%- if threshold is not none -%}
+            , struct({{ threshold }} as threshold)
+        {%- endif -%}
+    )
+{% endmacro %}