fix(ingestion): heal positional schema drift in class contracts

src/ingestion/connectors/ai/chatgpt-team/dbt/chatgpt_team__ai_assistant_usage.sql

@@ -38,9 +38,7 @@ SELECT
     lower(trim(email))                                              AS email,
     toDate(date)                                                    AS day,
     'chatgpt'                                                       AS tool,
-    'ChatGPT'                                                       AS tool_label,
     'chat'                                                          AS surface,
-    'Chat'                                                          AS surface_label,
     -- chat is not session-bounded in the API → session_count NULL
     CAST(NULL AS Nullable(UInt32))                                  AS session_count,
     CAST(NULL AS Nullable(UInt32))                                  AS conversation_count,

src/ingestion/connectors/ai/chatgpt-team/dbt/chatgpt_team__ai_dev_usage.sql

@@ -42,7 +42,6 @@ SELECT
     CAST(NULL AS Nullable(String))                      AS api_key_id,
     toDate(date)                                        AS day,
     'codex'                                             AS tool,
-    'Codex'                                             AS tool_label,
     -- Codex threads ≈ coding sessions. Non-nullable UInt32 per contract.
     toUInt32(coalesce(toUInt32OrNull(toString(n_threads)), 0))        AS session_count,
     toUInt32OrNull(toString(n_threads))                               AS conversation_count,

src/ingestion/connectors/ai/claude-admin/dbt/claude_admin__ai_dev_usage.sql

@@ -99,7 +99,6 @@ SELECT
     CASE WHEN u.actor_type = 'api_actor' THEN k.api_key_id END AS api_key_id,
     u.day,
     'claude_code'                                           AS tool,
-    'Claude Code'                                           AS tool_label,
     toUInt32(u.sessions_sum)                                AS session_count,
     toUInt32OrNull(toString(u.sessions_sum))                AS conversation_count,
     toUInt32(u.lines_added_sum)                             AS lines_added,

src/ingestion/connectors/ai/claude-enterprise/dbt/claude_enterprise__ai_assistant_usage.sql

@@ -90,9 +90,7 @@ chat AS (
         email,
         day,
         'claude'                                                        AS tool,
-        'Claude'                                                        AS tool_label,
         'chat'                                                          AS surface,
-        'Chat'                                                          AS surface_label,
         -- chat is not session-bounded in the API → session_count NULL
         CAST(NULL AS Nullable(UInt32))                                  AS session_count,
         toUInt32OrNull(toString(chat_conversation_count))               AS conversation_count,
@@ -140,9 +138,7 @@ excel AS (
         email,
         day,
         'claude'                                                        AS tool,
-        'Claude'                                                        AS tool_label,
         'excel'                                                         AS surface,
-        'Excel'                                                         AS surface_label,
         toUInt32OrNull(toString(excel_session_count))                   AS session_count,
         CAST(NULL AS Nullable(UInt32))                                  AS conversation_count,
         toUInt32OrNull(toString(excel_message_count))                   AS message_count,
@@ -177,9 +173,7 @@ powerpoint AS (
         email,
         day,
         'claude'                                                        AS tool,
-        'Claude'                                                        AS tool_label,
         'powerpoint'                                                    AS surface,
-        'PowerPoint'                                                    AS surface_label,
         toUInt32OrNull(toString(powerpoint_session_count))              AS session_count,
         CAST(NULL AS Nullable(UInt32))                                  AS conversation_count,
         toUInt32OrNull(toString(powerpoint_message_count))              AS message_count,
@@ -214,9 +208,7 @@ cowork AS (
         email,
         day,
         'claude'                                                        AS tool,
-        'Claude'                                                        AS tool_label,
         'cowork'                                                        AS surface,
-        'Cowork'                                                        AS surface_label,
         toUInt32OrNull(toString(cowork_session_count))                  AS session_count,
         CAST(NULL AS Nullable(UInt32))                                  AS conversation_count,
         toUInt32OrNull(toString(cowork_message_count))                  AS message_count,
@@ -259,9 +251,7 @@ cross_surface AS (
         email,
         day,
         'claude'                                                        AS tool,
-        'Claude'                                                        AS tool_label,
         'cross'                                                         AS surface,
-        'Cross'                                                         AS surface_label,
         CAST(NULL AS Nullable(UInt32))                                  AS session_count,
         CAST(NULL AS Nullable(UInt32))                                  AS conversation_count,
         CAST(NULL AS Nullable(UInt32))                                  AS message_count,

src/ingestion/connectors/ai/claude-enterprise/dbt/claude_enterprise__ai_dev_usage.sql

@@ -53,7 +53,6 @@ SELECT
     CAST(NULL AS Nullable(String))                                     AS api_key_id,
     toDate(parseDateTimeBestEffortOrNull(date))                        AS day,
     'claude_code'                                                      AS tool,
-    'Claude Code'                                                      AS tool_label,
     toUInt32(coalesce(code_session_count, 0))                          AS session_count,
     toUInt32OrNull(toString(code_session_count))                       AS conversation_count,
     toUInt32(coalesce(code_lines_added, 0))                            AS lines_added,

src/ingestion/connectors/ai/claude-team/dbt/claude_team__ai_dev_usage.sql

@@ -54,7 +54,6 @@ SELECT
     CAST(NULL AS Nullable(String))                      AS api_key_id,
     toDate(metric_date)                                 AS day,
     'claude_code'                                       AS tool,
-    'Claude Code'                                       AS tool_label,
     toUInt32(coalesce(total_sessions, 0))               AS session_count,
     toUInt32OrNull(toString(total_sessions))            AS conversation_count,
     toUInt32(coalesce(total_lines_accepted, 0))         AS lines_added,

src/ingestion/connectors/ai/cursor/dbt/cursor__ai_dev_usage.sql

@@ -39,7 +39,6 @@ SELECT
     CAST(NULL AS Nullable(String))                  AS api_key_id,
     toDate(fromUnixTimestamp64Milli(CAST(date AS Int64))) AS day,
     'cursor'                                        AS tool,
-    'Cursor'                                        AS tool_label,
     toUInt32(1)                                     AS session_count,
     CAST(NULL AS Nullable(UInt32))                  AS conversation_count,
     toUInt32(coalesce(acceptedLinesAdded, 0))       AS lines_added,

src/ingestion/connectors/ai/github-copilot/dbt/copilot__ai_dev_usage.sql

@@ -104,7 +104,6 @@ SELECT
     CAST(NULL AS Nullable(String))                                         AS api_key_id,
     toDate(parseDateTimeBestEffortOrNull(m.day))                           AS day,
     'copilot'                                                              AS tool,
-    'GitHub Copilot'                                                       AS tool_label,
     -- session_count: Copilot doesn't expose a per-day session counter;
     -- presence of an activity row implies at least one active session.
     -- Match Cursor's convention: 1 per active day.

src/ingestion/connectors/crm/hubspot/dbt/hubspot__crm_accounts.sql

@@ -44,6 +44,8 @@ WITH src AS (
             'annualrevenue',     coalesce(toString(properties_annualrevenue), ''),
             'archived',          toString(coalesce(archived, false))
         ))                                              AS metadata,
+        -- Envelope parity with salesforce__crm_* (no HubSpot custom-fields blob).
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         updatedAt                                       AS updated_at,
         data_source,

src/ingestion/connectors/crm/hubspot/dbt/hubspot__crm_activities.sql

@@ -70,6 +70,8 @@ WITH calls AS (
             'direction',      coalesce(toString(properties_hs_call_direction), ''),
             'archived',       toString(coalesce(archived, false))
         ))                                              AS metadata,
+        -- Envelope parity with salesforce__crm_* (no HubSpot custom-fields blob).
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         data_source,
         greatest(
@@ -112,6 +114,7 @@ emails AS (
             'direction',      coalesce(toString(properties_hs_email_direction), ''),
             'archived',       toString(coalesce(archived, false))
         ))                                              AS metadata,
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         data_source,
         greatest(
@@ -167,6 +170,7 @@ meetings AS (
             'location',       coalesce(toString(properties_hs_meeting_location), ''),
             'archived',       toString(coalesce(archived, false))
         ))                                              AS metadata,
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         data_source,
         coalesce(
@@ -208,6 +212,7 @@ tasks AS (
             'type',           coalesce(toString(properties_hs_task_type), ''),
             'archived',       toString(coalesce(archived, false))
         ))                                              AS metadata,
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         data_source,
         greatest(

src/ingestion/connectors/crm/hubspot/dbt/hubspot__crm_contacts.sql

@@ -48,6 +48,8 @@ WITH src AS (
             'hs_analytics_source', coalesce(toString(properties_hs_analytics_source), ''),
             'archived',         toString(coalesce(archived, false))
         ))                                              AS metadata,
+        -- Envelope parity with salesforce__crm_* (no HubSpot custom-fields blob).
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         updatedAt                                       AS updated_at,
         data_source,

src/ingestion/connectors/crm/hubspot/dbt/hubspot__crm_deals.sql

@@ -70,6 +70,8 @@ WITH src AS (
             'deal_type',      coalesce(toString(properties_dealtype), ''),
             'archived',       toString(coalesce(archived, false))
         ))                                              AS metadata,
+        -- Envelope parity with salesforce__crm_* (no HubSpot custom-fields blob).
+        '{}'                                            AS custom_fields,
         createdAt                                       AS created_at,
         updatedAt                                       AS updated_at,
         data_source,

src/ingestion/connectors/crm/hubspot/dbt/hubspot__crm_users.sql

@@ -54,6 +54,8 @@ WITH src AS (
             'userId',   coalesce(toString(userId), ''),
             'archived', toString(coalesce(archived, false))
         ))                                              AS metadata,
+        -- Envelope parity with salesforce__crm_* (no HubSpot custom-fields blob).
+        '{}'                                            AS custom_fields,
         collected_at,
         data_source,
         greatest(

src/ingestion/dbt/macros/ai_labels.sql

@@ -0,0 +1,27 @@
+{# Display labels for the AI `tool` and `surface` dimension codes — static
+   product vocabulary, same rationale as git_source_label: computed at gold
+   read time, never denormalized into silver rows. An unmapped code falls
+   through as itself. #}
+
+{% macro ai_tool_label(tool_expr) %}
+multiIf(
+    {{ tool_expr }} = 'cursor', 'Cursor',
+    {{ tool_expr }} = 'claude_code', 'Claude Code',
+    {{ tool_expr }} = 'copilot', 'GitHub Copilot',
+    {{ tool_expr }} = 'codex', 'Codex',
+    {{ tool_expr }} = 'claude', 'Claude',
+    {{ tool_expr }} = 'chatgpt', 'ChatGPT',
+    {{ tool_expr }}
+)
+{% endmacro %}
+
+{% macro ai_surface_label(surface_expr) %}
+multiIf(
+    {{ surface_expr }} = 'chat', 'Chat',
+    {{ surface_expr }} = 'excel', 'Excel',
+    {{ surface_expr }} = 'powerpoint', 'PowerPoint',
+    {{ surface_expr }} = 'cowork', 'Cowork',
+    {{ surface_expr }} = 'cross', 'Cross',
+    {{ surface_expr }}
+)
+{% endmacro %}

src/ingestion/gold/ai_metric_observations.sql

@@ -8,14 +8,16 @@
 -- Source measure observations for the unified metrics runtime. Reads only
 -- class-contract fields: activity is row existence (the class contract
 -- guarantees rows exist only for real activity — see silver/ai/schema.yml),
--- display labels come from tool_label / surface_label, and conversation
+-- display labels derive from the tool / surface discriminator codes via the
+-- shared vocabulary macros (macros/ai_labels.sql — static product
+-- vocabulary, never denormalized into silver rows), and conversation
 -- semantics come from data presence (conversation_count is NULL for sources
--- without a conversation concept). Discriminator and label columns are
--- non-null non-empty by the class contract (enforced by silver schema tests);
--- this model consumes them as-is. No vendor-specific columns, tool names, or
--- label mappings may appear in this model. Every measure is emitted through
--- the shape macros in macros/metric_observation_measures.sql; filter
--- predicates may reference only class-contract dimension values.
+-- without a conversation concept). Discriminator columns are non-null
+-- non-empty by the class contract (enforced by silver schema tests); this
+-- model consumes them as-is. No vendor mapping is inlined in this model —
+-- labels go through the macros only. Every measure is emitted through the
+-- shape macros in macros/metric_observation_measures.sql; filter predicates
+-- may reference only class-contract dimension values.
 
 WITH
 ai_dev_usage_source AS (
@@ -24,7 +26,7 @@ ai_dev_usage_source AS (
         lower(email) AS entity_id,
         day AS metric_date,
         CAST(
-            [tuple('tool', tool, tool_label)]
+            [tuple('tool', tool, {{ ai_tool_label('tool') }})]
             AS Array(Tuple(key String, value String, label Nullable(String)))
         ) AS tool_dimensions,
         conversation_count,
@@ -44,13 +46,13 @@ ai_assistant_usage_source AS (
         day AS metric_date,
         surface,
         CAST(
-            [tuple('tool', tool, tool_label)]
+            [tuple('tool', tool, {{ ai_tool_label('tool') }})]
             AS Array(Tuple(key String, value String, label Nullable(String)))
         ) AS tool_dimensions,
         CAST(
             [
-                tuple('tool', tool, tool_label),
-                tuple('surface', surface, surface_label)
+                tuple('tool', tool, {{ ai_tool_label('tool') }}),
+                tuple('surface', surface, {{ ai_surface_label('surface') }})
             ] AS Array(Tuple(key String, value String, label Nullable(String)))
         ) AS tool_surface_dimensions,
         conversation_count,

src/ingestion/gold/collab_metric_observations.sql

@@ -4,7 +4,13 @@
     order_by=['source_key', 'measure_key', 'entity_id', 'metric_date'],
     schema='insight',
     alias='collab_metric_observations',
-    tags=['gold']
+    tags=['gold'],
+    query_settings={
+        'max_memory_usage': 1610612736,
+        'max_threads': 4,
+        'max_bytes_before_external_group_by': 805306368,
+        'max_bytes_before_external_sort': 805306368
+    }
 ) }}
 
 -- Source measure observations for the unified metrics runtime, collaboration
@@ -22,6 +28,10 @@
 -- (source_key, measure_key, entity_id, metric_date), so single-measure
 -- queries read index-pruned ranges rather than the whole relation.
 --
+-- query_settings bound the CREATE-AS-SELECT for every runner: an
+-- over-limit build spills aggregation/sort state to disk instead of
+-- failing on the server memory tracker.
+--
 -- The `tool` dimension value is the `data_source` discriminator with the
 -- `insight_` prefix stripped (m365, slack, zoom, zulip_proxy). All Microsoft
 -- 365 surfaces (Teams, Outlook, OneDrive, SharePoint) share the m365 tool.

src/ingestion/gold/git_metric_observations.sql

@@ -4,7 +4,13 @@
     order_by=['source_key', 'measure_key', 'entity_id', 'metric_date'],
     schema='insight',
     alias='git_metric_observations',
-    tags=['gold']
+    tags=['gold'],
+    query_settings={
+        'max_memory_usage': 1610612736,
+        'max_threads': 4,
+        'max_bytes_before_external_group_by': 805306368,
+        'max_bytes_before_external_sort': 805306368
+    }
 ) }}
 
 -- Source measure observations for the unified metrics runtime, git family.
@@ -22,6 +28,10 @@
 -- (source_key, measure_key, entity_id, metric_date), so single-measure
 -- queries read index-pruned ranges rather than the whole relation.
 --
+-- query_settings bound the CREATE-AS-SELECT for every runner: an
+-- over-limit build spills aggregation/sort state to disk instead of
+-- failing on the server memory tracker.
+--
 -- Grain per measure:
 --   day-grain sums:  commit_count, code_lines_added, lines_added,
 --                    pr_created, pr_merged