node-type-registry 0.28.0 → 0.29.0

package/data/data-chunks.d.ts

@@ -0,0 +1,18 @@
+import type { NodeTypeDefinition } from '../types';
+/**
+ * Standalone chunking node type.
+ *
+ * Creates an embedding_chunks record that provisions a chunks table with:
+ * - FK to parent table (CASCADE delete)
+ * - content text field
+ * - chunk_index integer field
+ * - embedding vector(N) field with HNSW index
+ * - metadata jsonb field
+ * - RLS policies inherited from parent
+ * - Optional job trigger for automatic chunking on INSERT/UPDATE
+ *
+ * This node is also composed internally by DataFileEmbedding (enabled by
+ * default in extract mode). Use it standalone when you want a chunks table
+ * without the full file-embedding pipeline.
+ */
+export declare const DataChunks: NodeTypeDefinition;

package/data/data-chunks.js

@@ -0,0 +1,101 @@
+"use strict";
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.DataChunks = void 0;
+/**
+ * Standalone chunking node type.
+ *
+ * Creates an embedding_chunks record that provisions a chunks table with:
+ * - FK to parent table (CASCADE delete)
+ * - content text field
+ * - chunk_index integer field
+ * - embedding vector(N) field with HNSW index
+ * - metadata jsonb field
+ * - RLS policies inherited from parent
+ * - Optional job trigger for automatic chunking on INSERT/UPDATE
+ *
+ * This node is also composed internally by DataFileEmbedding (enabled by
+ * default in extract mode). Use it standalone when you want a chunks table
+ * without the full file-embedding pipeline.
+ */
+exports.DataChunks = {
+    name: 'DataChunks',
+    slug: 'data_chunks',
+    category: 'data',
+    display_name: 'Chunks',
+    description: 'Creates a chunked-embedding child table for any parent table. ' +
+        'Provisions the chunks table with content, chunk_index, embedding vector, ' +
+        'metadata, HNSW index, inherited RLS, and optional job trigger for ' +
+        'automatic text splitting. Composed internally by DataFileEmbedding ' +
+        '(enabled by default in extract mode) but can also be used standalone.',
+    parameter_schema: {
+        type: 'object',
+        properties: {
+            // ── Content config ─────────────────────────────────────────────
+            content_field_name: {
+                type: 'string',
+                format: 'column-ref',
+                description: 'Name of the text content column in the chunks table',
+                default: 'content'
+            },
+            // ── Chunking strategy ──────────────────────────────────────────
+            chunk_size: {
+                type: 'integer',
+                description: 'Maximum number of characters per chunk',
+                default: 1000
+            },
+            chunk_overlap: {
+                type: 'integer',
+                description: 'Number of overlapping characters between consecutive chunks',
+                default: 200
+            },
+            chunk_strategy: {
+                type: 'string',
+                enum: ['fixed', 'sentence', 'paragraph', 'semantic'],
+                description: 'Strategy for splitting text into chunks',
+                default: 'paragraph'
+            },
+            // ── Embedding config ───────────────────────────────────────────
+            dimensions: {
+                type: 'integer',
+                description: 'Vector dimensions for per-chunk embeddings',
+                default: 768
+            },
+            metric: {
+                type: 'string',
+                enum: ['cosine', 'l2', 'ip'],
+                description: 'Distance metric for the HNSW index on chunk embeddings',
+                default: 'cosine'
+            },
+            // ── Table naming ───────────────────────────────────────────────
+            chunks_table_name: {
+                type: 'string',
+                description: 'Override the chunks table name. Defaults to {parent_table}_chunks.',
+            },
+            // ── Metadata ───────────────────────────────────────────────────
+            metadata_fields: {
+                type: 'array',
+                items: { type: 'string' },
+                description: 'Field names from the parent table to copy into chunk metadata'
+            },
+            // ── Job trigger ────────────────────────────────────────────────
+            enqueue_chunking_job: {
+                type: 'boolean',
+                description: 'Whether to create a job trigger that auto-enqueues chunking ' +
+                    'on parent INSERT/UPDATE',
+                default: true
+            },
+            chunking_task_name: {
+                type: 'string',
+                description: 'Task identifier for the chunking job queue',
+                default: 'generate_chunks'
+            }
+        }
+    },
+    tags: [
+        'embedding',
+        'chunks',
+        'vector',
+        'ai',
+        'rag'
+    ]
+};

package/data/data-file-embedding.js

@@ -9,9 +9,10 @@ exports.DataFileEmbedding = {
     description: 'Generic, MIME-scoped embedding node for file tables. Supports two modes: ' +
         'direct (whole-file to single vector, e.g. CLIP for images) when extraction ' +
         'is omitted, or extract (file to text to chunks to per-chunk vectors) when ' +
-        'extraction config is provided. Composes SearchVector + DataJobTrigger ' +
-        'internally. Multiple instances can coexist on the same table with different ' +
-        'MIME scopes, field names, and embedding strategies.',
+        'extraction config is provided. Composes SearchVector + DataJobTrigger + ' +
+        'DataChunks (enabled by default in extract mode) internally. Multiple ' +
+        'instances can coexist on the same table with different MIME scopes, field ' +
+        'names, and embedding strategies.',
     parameter_schema: {
         type: 'object',
         properties: {
@@ -114,12 +115,18 @@ exports.DataFileEmbedding = {
                     }
                 }
             },
-            // ── Chunking config (optional — creates embedding_chunks) ──────
+            // ── Chunking (enabled by default in extract mode) ──────────────
+            include_chunks: {
+                type: 'boolean',
+                description: 'Whether to create a chunks table via DataChunks. Defaults to true ' +
+                    'when extraction is provided, false in direct mode. Set explicitly ' +
+                    'to override.',
+            },
             chunks: {
                 type: 'object',
-                description: 'Chunking configuration. Creates an embedding_chunks record that drives ' +
-                    'automatic text splitting and per-chunk embedding. Only meaningful when ' +
-                    'extraction is also provided.',
+                description: 'Chunking configuration passed through to DataChunks. When ' +
+                    'include_chunks is true (or defaults to true in extract mode), these ' +
+                    'params configure the chunks table, embedding dimensions, strategy, etc.',
                 properties: {
                     content_field_name: {
                         type: 'string',
@@ -144,8 +151,9 @@ exports.DataFileEmbedding = {
                         default: 'paragraph'
                     },
                     metadata_fields: {
-                        type: 'object',
-                        description: 'Metadata fields from parent to copy into chunks'
+                        type: 'array',
+                        items: { type: 'string' },
+                        description: 'Field names from parent to copy into chunk metadata'
                     },
                     enqueue_chunking_job: {
                         type: 'boolean',

package/data/index.d.ts

@@ -1,3 +1,4 @@
+export { DataChunks } from './data-chunks';
 export { DataCompositeField } from './data-composite-field';
 export { DataDirectOwner } from './data-direct-owner';
 export { DataEntityMembership } from './data-entity-membership';

package/data/index.js

@@ -1,6 +1,8 @@
 "use strict";
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.TableUserSettings = exports.TableUserProfiles = exports.TableOrganizationSettings = exports.SearchVector = exports.SearchUnified = exports.SearchTrgm = exports.SearchSpatialAggregate = exports.SearchSpatial = exports.SearchFullText = exports.SearchBm25 = exports.DataTimestamps = exports.DataTags = exports.DataStatusField = exports.DataSoftDelete = exports.DataSlug = exports.DataPublishable = exports.DataPeoplestamps = exports.DataOwnershipInEntity = exports.DataOwnedFields = exports.DataJsonb = exports.DataLimitCounter = exports.DataJobTrigger = exports.DataInheritFromParent = exports.DataInflection = exports.DataImmutableFields = exports.DataImageEmbedding = exports.DataId = exports.DataForceCurrentUser = exports.DataFeatureFlag = exports.DataFileEmbedding = exports.DataEntityMembership = exports.DataDirectOwner = exports.DataCompositeField = void 0;
+exports.TableUserSettings = exports.TableUserProfiles = exports.TableOrganizationSettings = exports.SearchVector = exports.SearchUnified = exports.SearchTrgm = exports.SearchSpatialAggregate = exports.SearchSpatial = exports.SearchFullText = exports.SearchBm25 = exports.DataTimestamps = exports.DataTags = exports.DataStatusField = exports.DataSoftDelete = exports.DataSlug = exports.DataPublishable = exports.DataPeoplestamps = exports.DataOwnershipInEntity = exports.DataOwnedFields = exports.DataJsonb = exports.DataLimitCounter = exports.DataJobTrigger = exports.DataInheritFromParent = exports.DataInflection = exports.DataImmutableFields = exports.DataImageEmbedding = exports.DataId = exports.DataForceCurrentUser = exports.DataFeatureFlag = exports.DataFileEmbedding = exports.DataEntityMembership = exports.DataDirectOwner = exports.DataCompositeField = exports.DataChunks = void 0;
+var data_chunks_1 = require("./data-chunks");
+Object.defineProperty(exports, "DataChunks", { enumerable: true, get: function () { return data_chunks_1.DataChunks; } });
 var data_composite_field_1 = require("./data-composite-field");
 Object.defineProperty(exports, "DataCompositeField", { enumerable: true, get: function () { return data_composite_field_1.DataCompositeField; } });
 var data_direct_owner_1 = require("./data-direct-owner");

package/esm/data/data-chunks.d.ts

@@ -0,0 +1,18 @@
+import type { NodeTypeDefinition } from '../types';
+/**
+ * Standalone chunking node type.
+ *
+ * Creates an embedding_chunks record that provisions a chunks table with:
+ * - FK to parent table (CASCADE delete)
+ * - content text field
+ * - chunk_index integer field
+ * - embedding vector(N) field with HNSW index
+ * - metadata jsonb field
+ * - RLS policies inherited from parent
+ * - Optional job trigger for automatic chunking on INSERT/UPDATE
+ *
+ * This node is also composed internally by DataFileEmbedding (enabled by
+ * default in extract mode). Use it standalone when you want a chunks table
+ * without the full file-embedding pipeline.
+ */
+export declare const DataChunks: NodeTypeDefinition;

package/esm/data/data-chunks.js

@@ -0,0 +1,98 @@
+/**
+ * Standalone chunking node type.
+ *
+ * Creates an embedding_chunks record that provisions a chunks table with:
+ * - FK to parent table (CASCADE delete)
+ * - content text field
+ * - chunk_index integer field
+ * - embedding vector(N) field with HNSW index
+ * - metadata jsonb field
+ * - RLS policies inherited from parent
+ * - Optional job trigger for automatic chunking on INSERT/UPDATE
+ *
+ * This node is also composed internally by DataFileEmbedding (enabled by
+ * default in extract mode). Use it standalone when you want a chunks table
+ * without the full file-embedding pipeline.
+ */
+export const DataChunks = {
+    name: 'DataChunks',
+    slug: 'data_chunks',
+    category: 'data',
+    display_name: 'Chunks',
+    description: 'Creates a chunked-embedding child table for any parent table. ' +
+        'Provisions the chunks table with content, chunk_index, embedding vector, ' +
+        'metadata, HNSW index, inherited RLS, and optional job trigger for ' +
+        'automatic text splitting. Composed internally by DataFileEmbedding ' +
+        '(enabled by default in extract mode) but can also be used standalone.',
+    parameter_schema: {
+        type: 'object',
+        properties: {
+            // ── Content config ─────────────────────────────────────────────
+            content_field_name: {
+                type: 'string',
+                format: 'column-ref',
+                description: 'Name of the text content column in the chunks table',
+                default: 'content'
+            },
+            // ── Chunking strategy ──────────────────────────────────────────
+            chunk_size: {
+                type: 'integer',
+                description: 'Maximum number of characters per chunk',
+                default: 1000
+            },
+            chunk_overlap: {
+                type: 'integer',
+                description: 'Number of overlapping characters between consecutive chunks',
+                default: 200
+            },
+            chunk_strategy: {
+                type: 'string',
+                enum: ['fixed', 'sentence', 'paragraph', 'semantic'],
+                description: 'Strategy for splitting text into chunks',
+                default: 'paragraph'
+            },
+            // ── Embedding config ───────────────────────────────────────────
+            dimensions: {
+                type: 'integer',
+                description: 'Vector dimensions for per-chunk embeddings',
+                default: 768
+            },
+            metric: {
+                type: 'string',
+                enum: ['cosine', 'l2', 'ip'],
+                description: 'Distance metric for the HNSW index on chunk embeddings',
+                default: 'cosine'
+            },
+            // ── Table naming ───────────────────────────────────────────────
+            chunks_table_name: {
+                type: 'string',
+                description: 'Override the chunks table name. Defaults to {parent_table}_chunks.',
+            },
+            // ── Metadata ───────────────────────────────────────────────────
+            metadata_fields: {
+                type: 'array',
+                items: { type: 'string' },
+                description: 'Field names from the parent table to copy into chunk metadata'
+            },
+            // ── Job trigger ────────────────────────────────────────────────
+            enqueue_chunking_job: {
+                type: 'boolean',
+                description: 'Whether to create a job trigger that auto-enqueues chunking ' +
+                    'on parent INSERT/UPDATE',
+                default: true
+            },
+            chunking_task_name: {
+                type: 'string',
+                description: 'Task identifier for the chunking job queue',
+                default: 'generate_chunks'
+            }
+        }
+    },
+    tags: [
+        'embedding',
+        'chunks',
+        'vector',
+        'ai',
+        'rag'
+    ]
+};

package/esm/data/data-file-embedding.js

@@ -6,9 +6,10 @@ export const DataFileEmbedding = {
     description: 'Generic, MIME-scoped embedding node for file tables. Supports two modes: ' +
         'direct (whole-file to single vector, e.g. CLIP for images) when extraction ' +
         'is omitted, or extract (file to text to chunks to per-chunk vectors) when ' +
-        'extraction config is provided. Composes SearchVector + DataJobTrigger ' +
-        'internally. Multiple instances can coexist on the same table with different ' +
-        'MIME scopes, field names, and embedding strategies.',
+        'extraction config is provided. Composes SearchVector + DataJobTrigger + ' +
+        'DataChunks (enabled by default in extract mode) internally. Multiple ' +
+        'instances can coexist on the same table with different MIME scopes, field ' +
+        'names, and embedding strategies.',
     parameter_schema: {
         type: 'object',
         properties: {
@@ -111,12 +112,18 @@ export const DataFileEmbedding = {
                     }
                 }
             },
-            // ── Chunking config (optional — creates embedding_chunks) ──────
+            // ── Chunking (enabled by default in extract mode) ──────────────
+            include_chunks: {
+                type: 'boolean',
+                description: 'Whether to create a chunks table via DataChunks. Defaults to true ' +
+                    'when extraction is provided, false in direct mode. Set explicitly ' +
+                    'to override.',
+            },
             chunks: {
                 type: 'object',
-                description: 'Chunking configuration. Creates an embedding_chunks record that drives ' +
-                    'automatic text splitting and per-chunk embedding. Only meaningful when ' +
-                    'extraction is also provided.',
+                description: 'Chunking configuration passed through to DataChunks. When ' +
+                    'include_chunks is true (or defaults to true in extract mode), these ' +
+                    'params configure the chunks table, embedding dimensions, strategy, etc.',
                 properties: {
                     content_field_name: {
                         type: 'string',
@@ -141,8 +148,9 @@ export const DataFileEmbedding = {
                         default: 'paragraph'
                     },
                     metadata_fields: {
-                        type: 'object',
-                        description: 'Metadata fields from parent to copy into chunks'
+                        type: 'array',
+                        items: { type: 'string' },
+                        description: 'Field names from parent to copy into chunk metadata'
                     },
                     enqueue_chunking_job: {
                         type: 'boolean',

package/esm/data/index.d.ts

@@ -1,3 +1,4 @@
+export { DataChunks } from './data-chunks';
 export { DataCompositeField } from './data-composite-field';
 export { DataDirectOwner } from './data-direct-owner';
 export { DataEntityMembership } from './data-entity-membership';

package/esm/data/index.js

@@ -1,3 +1,4 @@
+export { DataChunks } from './data-chunks';
 export { DataCompositeField } from './data-composite-field';
 export { DataDirectOwner } from './data-direct-owner';
 export { DataEntityMembership } from './data-entity-membership';

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "node-type-registry",
-  "version": "0.28.0",
+  "version": "0.29.0",
   "description": "Node type definitions for the Constructive blueprint system. Single source of truth for all Authz*, Data*, Relation*, and View* node types.",
   "author": "Constructive <developers@constructive.io>",
   "main": "index.js",
@@ -47,5 +47,5 @@
     "registry",
     "graphile"
   ],
-  "gitHead": "0538fe39630e6bafd228cb3f2b114fff4f9a61aa"
+  "gitHead": "44e6712bd8a37e2089418a69d801d67651c89350"
 }