update v2 configs for blog + movielens + hm

.github/workflows/upload_engine.yaml

@@ -17,12 +17,8 @@ jobs:
                 include:
                     - name: document-search
                       engine_config: document-search/notebooks/blog_engine_config.yaml
-                    - name: fashion-catalog-semantic-search
-                      engine_config: fashion-catalog/python/engine_configs/simple_semantic_search.yaml
                     - name: fashion-catalog-autotune
                       engine_config: fashion-catalog/python/engine_configs/autotune_lightgbm.yaml
-                    - name: movie-recommendations-v1
-                      engine_config: movie-recommendations/model/engine_configs/search_and_recommendations_v1.yaml
                     - name: movie-recommendations-v2
                       engine_config: movie-recommendations/model/engine_configs/search_and_recommendations_v2.yaml
         defaults:

apps/document-search/notebooks/blog_engine_config.yaml

@@ -1,22 +1,46 @@
 version: v2
-name: demo___blog_semantic_search
+name: shaped_blog_semantic_search_and_rag
+
 data:
   item_table:
     type: query
-    query: select toDateTime(toInt64(created_at) / 1000) as created_at_datetime, item_id, author, categories, chunk_metadata, content, created_at, featured, main_image, name, popular, article_id, read_length_in_mins, release_date, roles, slug, updated_at, url from blog_post_chunked order by created_at_datetime asc
+    query: |
+      select 
+        toDateTime(toInt64(created_at) / 1000) as created_at_datetime,
+        article_id,
+        item_id,
+        author,
+        categories,
+        chunk_metadata,
+        content,
+        created_at,
+        featured,
+        main_image,
+        name,
+        popular,
+        read_length_in_mins,
+        release_date,
+        roles,
+        slug,
+        updated_at,
+        url
+      from blog_articles_chunked
+      order by created_at_datetime asc
+
 index:
   lexical_search:
     item_fields:
-    - name
-    - content
-    - author
-    - categories
-    - roles
+      - name
+      - content
+      - author
+      - categories
+      - roles
+
   embeddings:
     - name: content_embedding
       encoder:
         type: hugging_face
-        model_name: sentence-transformers/all-MiniLM-L6-v2
+        model_name: Alibaba-NLP/gte-modernbert-base
         item_fields:
           - name
           - content

apps/fashion-catalog/python/engine_configs/autotune_lightgbm.yaml

@@ -1,25 +1,61 @@
+# this model demonstrates autotuning a lightgbm model
 version: v2
-name: hm__autotune_lightgbm
+name: hm_recommendations_and_search
+
 data:
-  item_dataset:
-    name: hm__articles
-    query: select * EXCEPT (article_id), article_id as item_id from hm__articles
-  user_dataset: 
-    name: hm__customers
-    query: select * EXCEPT (customer_id), customer_id as user_id from hm__customers
-  interactions_dataset: 
-    name: hm__transactions
-    query: select * EXCEPT (article_id, customer_id), article_id as item_id, customer_id as user_id from hm__transactions
-  index:
-    search:
-      item_fields:
-        - prod_name
-        - prod_type_name
-        - product_group_name
-        - colour_group_name
-        - department_name
-training:
-  models:
-    - policy_type: lightgbm
-    - policies:
-        - lightgbm
+  item_table:
+    type: table
+    name: hm_articles
+
+  user_table:
+    type: table
+    name: hm_customers
+
+  interaction_table:
+    type: query
+    query: select *, '5' as label from hm_transactions
+
+  schema_override:
+    item: 
+      id: item_id
+      features:
+        - name: image_url
+          type: Image
+
+index:
+  embeddings:
+    - name: image_embedding
+      encoder:
+        type: hugging_face
+        model_name: openai/clip-vit-base-patch32
+        item_fields:
+          - image_url
+    - name: text_content_embedding
+      encoder:
+        type: hugging_face
+        model_name: Alibaba-NLP/gte-modernbert-base
+        item_fields:
+          - prod_name
+          - index_name
+          - detail_desc
+          - section_name
+          - colour_group_name
+          - product_group_name
+          - graphical_appearance_name
+          - perceived_colour_value_name
+
+  lexical_search:
+    item_fields:
+      - prod_name
+      - index_name
+      - detail_desc
+      - section_name
+      - colour_group_name
+      - product_group_name
+      - graphical_appearance_name
+      - perceived_colour_value_name
+
+training: 
+  models: 
+    - name: lightgbm
+      policy_type: lightgbm

apps/movie-recommendations/model/engine_configs/search_and_recommendations_v2.yaml

@@ -1,34 +1,49 @@
 version: v2
-name: movielens__simple_semantic_search_2
-data: 
-  item_dataset:
-    name: movies_2018_enriched
-    query: |
-      SELECT r.movie_id as item_id, r.movie_title as title, r.poster_url,
-      r.imdb_url_new as imdb_url, r.genres as genres, r.description as
-      description, r.interests as interests, r.release_date as release_date,
-      r.cast as cast, r.directors as directors, r.writers as writers FROM
-      movies_2018_enriched r
-  index:
-    embeddings: 
-      - name: metadata_embedding
-        encoder: 
-          model_name: sentence-transformers/all-MiniLM-L6-v2
-          item_fields: 
-            - title
-            - description
-            - genres
-            - description
-            - directors
-            - writers
-            - cast
-          type: hugging_face
-    search: 
-      item_fields: 
-        - title
-        - description
-        - genres
-        - description
-        - directors
-        - writers
-        - cast
+name: movielens_recommendations_and_search
+
+data:
+  item_table:
+    type: query
+    query: select item_id as item_id, toDateTimeOrNull(release_date) AS release_datetime, cast, genres, imdbId, writers, directors, interests, poster_url, description, movie_title from movielens_items_enriched order by release_datetime asc
+
+  interaction_table:
+    type: query
+    query: SELECT r.user_id as user_id, r.item_id as item_id, r.timestamp AS created_at, r.label AS label FROM movielens_ratings_2018 r
+
+  schema_override: 
+    item:
+      id: item_id
+      features:
+        - name: poster_url
+          type: Image
+
+index:
+  embeddings:
+    - name: item_content_embedding
+      encoder:
+        type: hugging_face
+        model_name: Alibaba-NLP/gte-modernbert-base
+        item_fields:
+          - movie_title
+          - description
+          - cast
+          - genres
+          - writers
+          - directors
+          - interests
+    - name: poster_embedding
+      encoder:
+        type: hugging_face
+        model_name: openai/clip-vit-base-patch32
+        item_fields:
+          - poster_url
+
+  lexical_search:
+    item_fields:
+      - movie_title
+      - description
+      - cast
+      - genres
+      - writers
+      - directors
+      - interests