remdb 0.3.114__py3-none-any.whl → 0.3.127__py3-none-any.whl

rem/sql/migrations/002_install_models.sql

@@ -1,7 +1,7 @@
 -- REM Model Schema (install_models.sql)
 -- Generated from Pydantic models
 -- Source: model registry
--- Generated at: 2025-11-29T11:08:16.713884
+-- Generated at: 2025-11-29T18:45:11.372432
 --
 -- DO NOT EDIT MANUALLY - Regenerate with: rem db schema generate
 --
@@ -10,6 +10,7 @@
 -- 2. Embeddings tables (embeddings_<table>)
 -- 3. KV_STORE triggers for cache maintenance
 -- 4. Indexes (foreground only, background indexes separate)
+-- 5. Schema table entries (for agent-like table access)
 
 -- ============================================================================
 -- PREREQUISITES CHECK
@@ -1232,6 +1233,1868 @@ CREATE TRIGGER trg_users_kv_store
 AFTER INSERT OR UPDATE OR DELETE ON users
 FOR EACH ROW EXECUTE FUNCTION fn_users_kv_store_upsert();
 
+-- ============================================================================
+-- SCHEMA TABLE ENTRIES
+-- Every entity table gets a schemas entry for agent-like access
+-- ============================================================================
+
+-- Schema entry for Feedback (feedbacks)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'ae554853-e743-5d73-a2db-1ce20e7089fe'::uuid,
+    'system',
+    'Feedback',
+    '# Feedback
+
+
+    User feedback on a message or session.
+
+    Captures structured feedback including:
+    - Rating (1-5 scale or thumbs up/down)
+    - Categories (predefined or custom)
+    - Free-text comment
+    - Trace reference for OTEL/Phoenix integration
+
+    The feedback can be attached to:
+    - A specific message (message_id set)
+    - An entire session (session_id set, message_id null)
+    
+
+## Overview
+
+The `Feedback` entity is stored in the `feedbacks` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM feedbacks LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM feedbacks WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `feedbacks` |
+| Entity Key | `id` |
+| Embedding Fields | None |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `session_id`
+- **Type**: `<class ''str''>`
+- **Required**
+- Session ID this feedback relates to
+
+### `message_id`
+- **Type**: `str | None`
+- **Optional**
+- Specific message ID (null for session-level feedback)
+
+### `rating`
+- **Type**: `int | None`
+- **Optional**
+- Rating: -1 (thumbs down), 1 (thumbs up), or 1-5 scale
+
+### `categories`
+- **Type**: `list[str]`
+- **Optional**
+- Selected feedback categories (from FeedbackCategory or custom)
+
+### `comment`
+- **Type**: `str | None`
+- **Optional**
+- Optional free-text feedback comment
+
+### `trace_id`
+- **Type**: `str | None`
+- **Optional**
+- OTEL trace ID for linking to observability
+
+### `span_id`
+- **Type**: `str | None`
+- **Optional**
+- OTEL span ID for specific span feedback
+
+### `phoenix_synced`
+- **Type**: `<class ''bool''>`
+- **Optional**
+- Whether feedback has been synced to Phoenix as annotation
+
+### `phoenix_annotation_id`
+- **Type**: `str | None`
+- **Optional**
+- Phoenix annotation ID after sync
+
+### `annotator_kind`
+- **Type**: `<class ''str''>`
+- **Optional**
+- Annotator type: HUMAN, LLM, CODE
+
+',
+    '{"type": "object", "description": "\n    User feedback on a message or session.\n\n    Captures structured feedback including:\n    - Rating (1-5 scale or thumbs up/down)\n    - Categories (predefined or custom)\n    - Free-text comment\n    - Trace reference for OTEL/Phoenix integration\n\n    The feedback can be attached to:\n    - A specific message (message_id set)\n    - An entire session (session_id set, message_id null)\n    \n\nThis agent can search the `feedbacks` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "session_id": {"description": "Session ID this feedback relates to", "title": "Session Id", "type": "string"}, "message_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Specific message ID (null for session-level feedback)", "title": "Message Id"}, "rating": {"anyOf": [{"maximum": 5, "minimum": -1, "type": "integer"}, {"type": "null"}], "default": null, "description": "Rating: -1 (thumbs down), 1 (thumbs up), or 1-5 scale", "title": "Rating"}, "categories": {"description": "Selected feedback categories (from FeedbackCategory or custom)", "items": {"type": "string"}, "title": "Categories", "type": "array"}, "comment": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Optional free-text feedback comment", "title": "Comment"}, "trace_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "OTEL trace ID for linking to observability", "title": "Trace Id"}, "span_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "OTEL span ID for specific span feedback", "title": "Span Id"}, "phoenix_synced": {"default": false, "description": "Whether feedback has been synced to Phoenix as annotation", "title": "Phoenix Synced", "type": "boolean"}, "phoenix_annotation_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Phoenix annotation ID after sync", "title": "Phoenix Annotation Id"}, "annotator_kind": {"default": "HUMAN", "description": "Annotator type: HUMAN, LLM, CODE", "title": "Annotator Kind", "type": "string"}}, "required": ["session_id"], "json_schema_extra": {"table_name": "feedbacks", "entity_key_field": "id", "embedding_fields": [], "fully_qualified_name": "rem.models.entities.feedback.Feedback", "tools": ["search_rem"], "default_search_table": "feedbacks", "has_embeddings": false}}'::jsonb,
+    'entity',
+    '{"table_name": "feedbacks", "entity_key_field": "id", "embedding_fields": [], "fqn": "rem.models.entities.feedback.Feedback"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for File (files)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'c3b3ef33-59d4-57a1-81a3-cc6adc45b194'::uuid,
+    'system',
+    'File',
+    '# File
+
+
+    File metadata and tracking.
+
+    Represents files uploaded to or referenced by the REM system,
+    tracking their metadata and processing status. Tenant isolation
+    is provided via CoreModel.tenant_id field.
+    
+
+## Overview
+
+The `File` entity is stored in the `files` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM files LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM files WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `files` |
+| Entity Key | `id` |
+| Embedding Fields | `content` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+- File name
+
+### `uri`
+- **Type**: `<class ''str''>`
+- **Required**
+- File storage URI (S3, local path, etc.)
+
+### `content`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Extracted text content (if applicable)
+
+### `timestamp`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- File creation/modification timestamp
+
+### `size_bytes`
+- **Type**: `typing.Optional[int]`
+- **Optional**
+- File size in bytes
+
+### `mime_type`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- File MIME type
+
+### `processing_status`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- File processing status (pending, processing, completed, failed)
+
+',
+    '{"type": "object", "description": "\n    File metadata and tracking.\n\n    Represents files uploaded to or referenced by the REM system,\n    tracking their metadata and processing status. Tenant isolation\n    is provided via CoreModel.tenant_id field.\n    \n\nThis agent can search the `files` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"description": "File name", "title": "Name", "type": "string"}, "uri": {"description": "File storage URI (S3, local path, etc.)", "title": "Uri", "type": "string"}, "content": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Extracted text content (if applicable)", "title": "Content"}, "timestamp": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "File creation/modification timestamp", "title": "Timestamp"}, "size_bytes": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "File size in bytes", "title": "Size Bytes"}, "mime_type": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "File MIME type", "title": "Mime Type"}, "processing_status": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": "pending", "description": "File processing status (pending, processing, completed, failed)", "title": "Processing Status"}}, "required": ["name", "uri"], "json_schema_extra": {"table_name": "files", "entity_key_field": "id", "embedding_fields": ["content"], "fully_qualified_name": "rem.models.entities.file.File", "tools": ["search_rem"], "default_search_table": "files", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "files", "entity_key_field": "id", "embedding_fields": ["content"], "fqn": "rem.models.entities.file.File"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for ImageResource (image_resources)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'ab4bc90c-2cda-55b2-bd4b-e78e19f7d4a7'::uuid,
+    'system',
+    'ImageResource',
+    '# ImageResource
+
+
+    Image-specific resource with CLIP embeddings.
+
+    Stored in separate `image_resources` table with CLIP embeddings
+    instead of text embeddings. This enables:
+    - Multimodal search (text-to-image, image-to-image)
+    - Proper dimensionality (512/768 for CLIP vs 1536 for text)
+    - Cost tracking (CLIP tokens separate from text tokens)
+
+    Embedding Strategy:
+    - Default (when JINA_API_KEY set): Jina CLIP API (jina-clip-v2)
+    - Future: Self-hosted OpenCLIP models via KEDA-scaled pods
+    - Fallback: No embeddings (images searchable by metadata only)
+
+    Vision LLM Strategy (tier/sampling gated):
+    - Gold tier: Always get vision descriptions
+    - Silver/Free: Probabilistic sampling (IMAGE_VLLM_SAMPLE_RATE)
+    - Fallback: Basic metadata only
+
+    Tenant isolation provided via CoreModel.tenant_id field.
+    
+
+## Overview
+
+The `ImageResource` entity is stored in the `image_resources` table. Each record is uniquely
+identified by its `name` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by name (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM image_resources LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM image_resources WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `image_resources` |
+| Entity Key | `name` |
+| Embedding Fields | `content` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Human-readable resource name (used as graph label). Auto-generated from uri+ordinal if not provided.
+
+### `uri`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Content URI or identifier (file path, URL, etc.)
+
+### `ordinal`
+- **Type**: `<class ''int''>`
+- **Optional**
+- Chunk ordinal for splitting large documents (0 for single-chunk resources)
+
+### `content`
+- **Type**: `<class ''str''>`
+- **Optional**
+- Resource content text
+
+### `timestamp`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Resource timestamp (content creation/publication time)
+
+### `category`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Resource category (document, conversation, artifact, etc.)
+
+### `related_entities`
+- **Type**: `list[dict]`
+- **Optional**
+- Extracted entities (people, projects, concepts) with metadata
+
+### `image_width`
+- **Type**: `typing.Optional[int]`
+- **Optional**
+- Image width in pixels
+
+### `image_height`
+- **Type**: `typing.Optional[int]`
+- **Optional**
+- Image height in pixels
+
+### `image_format`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Image format (PNG, JPEG, GIF, WebP)
+
+### `vision_description`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Vision LLM generated description (markdown, only for gold tier or sampled images)
+
+### `vision_provider`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Vision provider used (anthropic, gemini, openai)
+
+### `vision_model`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Vision model used for description
+
+### `clip_embedding`
+- **Type**: `typing.Optional[list[float]]`
+- **Optional**
+- CLIP embedding vector (512 or 768 dimensions, from Jina AI or self-hosted)
+
+### `clip_dimensions`
+- **Type**: `typing.Optional[int]`
+- **Optional**
+- CLIP embedding dimensionality (512 for jina-clip-v2, 768 for jina-clip-v1)
+
+',
+    '{"type": "object", "description": "\n    Image-specific resource with CLIP embeddings.\n\n    Stored in separate `image_resources` table with CLIP embeddings\n    instead of text embeddings. This enables:\n    - Multimodal search (text-to-image, image-to-image)\n    - Proper dimensionality (512/768 for CLIP vs 1536 for text)\n    - Cost tracking (CLIP tokens separate from text tokens)\n\n    Embedding Strategy:\n    - Default (when JINA_API_KEY set): Jina CLIP API (jina-clip-v2)\n    - Future: Self-hosted OpenCLIP models via KEDA-scaled pods\n    - Fallback: No embeddings (images searchable by metadata only)\n\n    Vision LLM Strategy (tier/sampling gated):\n    - Gold tier: Always get vision descriptions\n    - Silver/Free: Probabilistic sampling (IMAGE_VLLM_SAMPLE_RATE)\n    - Fallback: Basic metadata only\n\n    Tenant isolation provided via CoreModel.tenant_id field.\n    \n\nThis agent can search the `image_resources` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Human-readable resource name (used as graph label). Auto-generated from uri+ordinal if not provided.", "entity_key": true, "title": "Name"}, "uri": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Content URI or identifier (file path, URL, etc.)", "title": "Uri"}, "ordinal": {"composite_key": true, "default": 0, "description": "Chunk ordinal for splitting large documents (0 for single-chunk resources)", "title": "Ordinal", "type": "integer"}, "content": {"default": "", "description": "Resource content text", "title": "Content", "type": "string"}, "timestamp": {"description": "Resource timestamp (content creation/publication time)", "format": "date-time", "title": "Timestamp", "type": "string"}, "category": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Resource category (document, conversation, artifact, etc.)", "title": "Category"}, "related_entities": {"description": "Extracted entities (people, projects, concepts) with metadata", "items": {"additionalProperties": true, "type": "object"}, "title": "Related Entities", "type": "array"}, "image_width": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "Image width in pixels", "title": "Image Width"}, "image_height": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "Image height in pixels", "title": "Image Height"}, "image_format": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Image format (PNG, JPEG, GIF, WebP)", "title": "Image Format"}, "vision_description": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Vision LLM generated description (markdown, only for gold tier or sampled images)", "title": "Vision Description"}, "vision_provider": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Vision provider used (anthropic, gemini, openai)", "title": "Vision Provider"}, "vision_model": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Vision model used for description", "title": "Vision Model"}, "clip_embedding": {"anyOf": [{"items": {"type": "number"}, "type": "array"}, {"type": "null"}], "default": null, "description": "CLIP embedding vector (512 or 768 dimensions, from Jina AI or self-hosted)", "title": "Clip Embedding"}, "clip_dimensions": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "CLIP embedding dimensionality (512 for jina-clip-v2, 768 for jina-clip-v1)", "title": "Clip Dimensions"}}, "required": [], "json_schema_extra": {"table_name": "image_resources", "entity_key_field": "name", "embedding_fields": ["content"], "fully_qualified_name": "rem.models.entities.image_resource.ImageResource", "tools": ["search_rem"], "default_search_table": "image_resources", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "image_resources", "entity_key_field": "name", "embedding_fields": ["content"], "fqn": "rem.models.entities.image_resource.ImageResource"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Message (messages)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'be36f9da-6df4-51ba-bb41-bf51246ecec1'::uuid,
+    'system',
+    'Message',
+    '# Message
+
+
+    Communication content unit.
+
+    Represents individual messages in conversations, chats, or other
+    communication contexts. Tenant isolation is provided via CoreModel.tenant_id field.
+
+    Trace fields (trace_id, span_id) enable integration with OTEL/Phoenix
+    for observability and feedback annotation.
+    
+
+## Overview
+
+The `Message` entity is stored in the `messages` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM messages LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM messages WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `messages` |
+| Entity Key | `id` |
+| Embedding Fields | `content` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `content`
+- **Type**: `<class ''str''>`
+- **Required**
+- Message content text
+
+### `message_type`
+- **Type**: `str | None`
+- **Optional**
+- Message type e.g. role: ''user'', ''assistant'', ''system'', ''tool''
+
+### `session_id`
+- **Type**: `str | None`
+- **Optional**
+- Session identifier for tracking message context
+
+### `prompt`
+- **Type**: `str | None`
+- **Optional**
+- Custom prompt used for this message (if overridden from default)
+
+### `model`
+- **Type**: `str | None`
+- **Optional**
+- Model used for generating this message (provider:model format)
+
+### `token_count`
+- **Type**: `int | None`
+- **Optional**
+- Token count for this message
+
+### `trace_id`
+- **Type**: `str | None`
+- **Optional**
+- OTEL trace ID for observability integration
+
+### `span_id`
+- **Type**: `str | None`
+- **Optional**
+- OTEL span ID for specific span reference
+
+',
+    '{"type": "object", "description": "\n    Communication content unit.\n\n    Represents individual messages in conversations, chats, or other\n    communication contexts. Tenant isolation is provided via CoreModel.tenant_id field.\n\n    Trace fields (trace_id, span_id) enable integration with OTEL/Phoenix\n    for observability and feedback annotation.\n    \n\nThis agent can search the `messages` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "content": {"description": "Message content text", "title": "Content", "type": "string"}, "message_type": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Message type e.g. role: ''user'', ''assistant'', ''system'', ''tool''", "title": "Message Type"}, "session_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Session identifier for tracking message context", "title": "Session Id"}, "prompt": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Custom prompt used for this message (if overridden from default)", "title": "Prompt"}, "model": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Model used for generating this message (provider:model format)", "title": "Model"}, "token_count": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "Token count for this message", "title": "Token Count"}, "trace_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "OTEL trace ID for observability integration", "title": "Trace Id"}, "span_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "OTEL span ID for specific span reference", "title": "Span Id"}}, "required": ["content"], "json_schema_extra": {"table_name": "messages", "entity_key_field": "id", "embedding_fields": ["content"], "fully_qualified_name": "rem.models.entities.message.Message", "tools": ["search_rem"], "default_search_table": "messages", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "messages", "entity_key_field": "id", "embedding_fields": ["content"], "fqn": "rem.models.entities.message.Message"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Moment (moments)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'a08f0a8c-5bab-5bf5-9760-0e67bc69bd74'::uuid,
+    'system',
+    'Moment',
+    '# Moment
+
+
+    Temporal narrative extracted from resources.
+
+    Moments provide temporal structure and context for the REM graph,
+    enabling time-based queries and understanding of when events occurred.
+    Tenant isolation is provided via CoreModel.tenant_id field.
+    
+
+## Overview
+
+The `Moment` entity is stored in the `moments` table. Each record is uniquely
+identified by its `name` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by name (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on summary (e.g., `SEARCH "concept" FROM moments LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM moments WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `moments` |
+| Entity Key | `name` |
+| Embedding Fields | `summary` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Human-readable moment name (used as graph label). Auto-generated from starts_timestamp+moment_type if not provided.
+
+### `moment_type`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Moment classification (meeting, coding-session, conversation, etc.)
+
+### `category`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Moment category for grouping and filtering
+
+### `starts_timestamp`
+- **Type**: `<class ''datetime.datetime''>`
+- **Required**
+- Moment start time
+
+### `ends_timestamp`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Moment end time
+
+### `present_persons`
+- **Type**: `list[rem.models.entities.moment.Person]`
+- **Optional**
+- People present in the moment
+
+### `emotion_tags`
+- **Type**: `list[str]`
+- **Optional**
+- Emotion/sentiment tags (happy, frustrated, focused, etc.)
+
+### `topic_tags`
+- **Type**: `list[str]`
+- **Optional**
+- Topic/concept tags (project names, technologies, etc.)
+
+### `summary`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Natural language summary of the moment
+
+### `source_resource_ids`
+- **Type**: `list[str]`
+- **Optional**
+- Resource IDs used to construct this moment
+
+',
+    '{"type": "object", "description": "\n    Temporal narrative extracted from resources.\n\n    Moments provide temporal structure and context for the REM graph,\n    enabling time-based queries and understanding of when events occurred.\n    Tenant isolation is provided via CoreModel.tenant_id field.\n    \n\nThis agent can search the `moments` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Human-readable moment name (used as graph label). Auto-generated from starts_timestamp+moment_type if not provided.", "entity_key": true, "title": "Name"}, "moment_type": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Moment classification (meeting, coding-session, conversation, etc.)", "title": "Moment Type"}, "category": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Moment category for grouping and filtering", "title": "Category"}, "starts_timestamp": {"description": "Moment start time", "format": "date-time", "title": "Starts Timestamp", "type": "string"}, "ends_timestamp": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Moment end time", "title": "Ends Timestamp"}, "present_persons": {"description": "People present in the moment", "items": {"$ref": "#/$defs/Person"}, "title": "Present Persons", "type": "array"}, "emotion_tags": {"description": "Emotion/sentiment tags (happy, frustrated, focused, etc.)", "items": {"type": "string"}, "title": "Emotion Tags", "type": "array"}, "topic_tags": {"description": "Topic/concept tags (project names, technologies, etc.)", "items": {"type": "string"}, "title": "Topic Tags", "type": "array"}, "summary": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Natural language summary of the moment", "title": "Summary"}, "source_resource_ids": {"description": "Resource IDs used to construct this moment", "items": {"type": "string"}, "title": "Source Resource Ids", "type": "array"}}, "required": ["starts_timestamp"], "json_schema_extra": {"table_name": "moments", "entity_key_field": "name", "embedding_fields": ["summary"], "fully_qualified_name": "rem.models.entities.moment.Moment", "tools": ["search_rem"], "default_search_table": "moments", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "moments", "entity_key_field": "name", "embedding_fields": ["summary"], "fqn": "rem.models.entities.moment.Moment"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Ontology (ontologies)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'a702ed74-8988-534a-9917-2977349777c1'::uuid,
+    'system',
+    'Ontology',
+    '# Ontology
+
+Domain-specific knowledge extracted from files using custom agents.
+
+    Attributes:
+        name: Human-readable label for this ontology instance
+        file_id: Foreign key to File entity that was processed
+        agent_schema_id: Foreign key to Schema entity that performed extraction
+        provider_name: LLM provider used for extraction (e.g., "anthropic", "openai")
+        model_name: Specific model used (e.g., "claude-sonnet-4-5")
+        extracted_data: Structured data extracted by agent (arbitrary JSON)
+        confidence_score: Optional confidence score from extraction (0.0-1.0)
+        extraction_timestamp: When extraction was performed
+        embedding_text: Text used for generating embedding (derived from extracted_data)
+
+    Inherited from CoreModel:
+        id: UUID or string identifier
+        created_at: Entity creation timestamp
+        updated_at: Last update timestamp
+        deleted_at: Soft deletion timestamp
+        tenant_id: Multi-tenancy isolation
+        user_id: Ownership
+        graph_edges: Relationships to other entities
+        metadata: Flexible metadata storage
+        tags: Classification tags
+        column: Database schema metadata
+
+    Example Usage:
+        # CV extraction
+        cv_ontology = Ontology(
+            name="john-doe-cv-2024",
+            file_id="file-uuid-123",
+            agent_schema_id="cv-parser-v1",
+            provider_name="anthropic",
+            model_name="claude-sonnet-4-5-20250929",
+            extracted_data={
+                "candidate_name": "John Doe",
+                "email": "john@example.com",
+                "skills": ["Python", "PostgreSQL", "Kubernetes"],
+                "experience": [
+                    {
+                        "company": "TechCorp",
+                        "role": "Senior Engineer",
+                        "years": 3,
+                        "achievements": ["Led migration to k8s", "Reduced costs 40%"]
+                    }
+                ],
+                "education": [
+                    {"degree": "BS Computer Science", "institution": "MIT", "year": 2018}
+                ]
+            },
+            confidence_score=0.95,
+            tags=["cv", "engineering", "senior-level"]
+        )
+
+        # Contract extraction
+        contract_ontology = Ontology(
+            name="acme-supplier-agreement-2024",
+            file_id="file-uuid-456",
+            agent_schema_id="contract-parser-v2",
+            provider_name="openai",
+            model_name="gpt-4o",
+            extracted_data={
+                "contract_type": "supplier_agreement",
+                "parties": [
+                    {"name": "ACME Corp", "role": "buyer"},
+                    {"name": "SupplyChain Inc", "role": "supplier"}
+                ],
+                "effective_date": "2024-01-01",
+                "termination_date": "2026-12-31",
+                "payment_terms": {
+                    "amount": 500000,
+                    "currency": "USD",
+                    "frequency": "quarterly"
+                },
+                "key_obligations": [
+                    "Supplier must deliver within 30 days",
+                    "Buyer must pay within 60 days of invoice"
+                ]
+            },
+            confidence_score=0.92,
+            tags=["contract", "supplier", "procurement"]
+        )
+    
+
+## Overview
+
+The `Ontology` entity is stored in the `ontologies` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM ontologies LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM ontologies WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `ontologies` |
+| Entity Key | `id` |
+| Embedding Fields | None |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `file_id`
+- **Type**: `uuid.UUID | str`
+- **Required**
+
+### `agent_schema_id`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `provider_name`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `model_name`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `extracted_data`
+- **Type**: `dict[str, typing.Any]`
+- **Required**
+
+### `confidence_score`
+- **Type**: `typing.Optional[float]`
+- **Optional**
+
+### `extraction_timestamp`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+### `embedding_text`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+',
+    '{"type": "object", "description": "Domain-specific knowledge extracted from files using custom agents.\n\n    Attributes:\n        name: Human-readable label for this ontology instance\n        file_id: Foreign key to File entity that was processed\n        agent_schema_id: Foreign key to Schema entity that performed extraction\n        provider_name: LLM provider used for extraction (e.g., \"anthropic\", \"openai\")\n        model_name: Specific model used (e.g., \"claude-sonnet-4-5\")\n        extracted_data: Structured data extracted by agent (arbitrary JSON)\n        confidence_score: Optional confidence score from extraction (0.0-1.0)\n        extraction_timestamp: When extraction was performed\n        embedding_text: Text used for generating embedding (derived from extracted_data)\n\n    Inherited from CoreModel:\n        id: UUID or string identifier\n        created_at: Entity creation timestamp\n        updated_at: Last update timestamp\n        deleted_at: Soft deletion timestamp\n        tenant_id: Multi-tenancy isolation\n        user_id: Ownership\n        graph_edges: Relationships to other entities\n        metadata: Flexible metadata storage\n        tags: Classification tags\n        column: Database schema metadata\n\n    Example Usage:\n        # CV extraction\n        cv_ontology = Ontology(\n            name=\"john-doe-cv-2024\",\n            file_id=\"file-uuid-123\",\n            agent_schema_id=\"cv-parser-v1\",\n            provider_name=\"anthropic\",\n            model_name=\"claude-sonnet-4-5-20250929\",\n            extracted_data={\n                \"candidate_name\": \"John Doe\",\n                \"email\": \"john@example.com\",\n                \"skills\": [\"Python\", \"PostgreSQL\", \"Kubernetes\"],\n                \"experience\": [\n                    {\n                        \"company\": \"TechCorp\",\n                        \"role\": \"Senior Engineer\",\n                        \"years\": 3,\n                        \"achievements\": [\"Led migration to k8s\", \"Reduced costs 40%\"]\n                    }\n                ],\n                \"education\": [\n                    {\"degree\": \"BS Computer Science\", \"institution\": \"MIT\", \"year\": 2018}\n                ]\n            },\n            confidence_score=0.95,\n            tags=[\"cv\", \"engineering\", \"senior-level\"]\n        )\n\n        # Contract extraction\n        contract_ontology = Ontology(\n            name=\"acme-supplier-agreement-2024\",\n            file_id=\"file-uuid-456\",\n            agent_schema_id=\"contract-parser-v2\",\n            provider_name=\"openai\",\n            model_name=\"gpt-4o\",\n            extracted_data={\n                \"contract_type\": \"supplier_agreement\",\n                \"parties\": [\n                    {\"name\": \"ACME Corp\", \"role\": \"buyer\"},\n                    {\"name\": \"SupplyChain Inc\", \"role\": \"supplier\"}\n                ],\n                \"effective_date\": \"2024-01-01\",\n                \"termination_date\": \"2026-12-31\",\n                \"payment_terms\": {\n                    \"amount\": 500000,\n                    \"currency\": \"USD\",\n                    \"frequency\": \"quarterly\"\n                },\n                \"key_obligations\": [\n                    \"Supplier must deliver within 30 days\",\n                    \"Buyer must pay within 60 days of invoice\"\n                ]\n            },\n            confidence_score=0.92,\n            tags=[\"contract\", \"supplier\", \"procurement\"]\n        )\n    \n\nThis agent can search the `ontologies` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"title": "Name", "type": "string"}, "file_id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}], "title": "File Id"}, "agent_schema_id": {"title": "Agent Schema Id", "type": "string"}, "provider_name": {"title": "Provider Name", "type": "string"}, "model_name": {"title": "Model Name", "type": "string"}, "extracted_data": {"additionalProperties": true, "title": "Extracted Data", "type": "object"}, "confidence_score": {"anyOf": [{"type": "number"}, {"type": "null"}], "default": null, "title": "Confidence Score"}, "extraction_timestamp": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Extraction Timestamp"}, "embedding_text": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Embedding Text"}}, "required": ["name", "file_id", "agent_schema_id", "provider_name", "model_name", "extracted_data"], "json_schema_extra": {"table_name": "ontologies", "entity_key_field": "id", "embedding_fields": [], "fully_qualified_name": "rem.models.entities.ontology.Ontology", "tools": ["search_rem"], "default_search_table": "ontologies", "has_embeddings": false}}'::jsonb,
+    'entity',
+    '{"table_name": "ontologies", "entity_key_field": "id", "embedding_fields": [], "fqn": "rem.models.entities.ontology.Ontology"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for OntologyConfig (ontology_configs)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    '9a7e50d0-ef3a-5641-9ff4-b2be5a77053b'::uuid,
+    'system',
+    'OntologyConfig',
+    '# OntologyConfig
+
+User configuration for automatic ontology extraction.
+
+    Attributes:
+        name: Human-readable config name
+        agent_schema_id: Foreign key to Schema entity to use for extraction
+        description: Purpose and scope of this config
+
+        # File matching rules (ANY matching rule triggers extraction)
+        mime_type_pattern: Regex pattern for file MIME types (e.g., "application/pdf")
+        uri_pattern: Regex pattern for file URIs (e.g., "s3://bucket/resumes/.*")
+        tag_filter: List of tags (file must have ALL tags to match)
+
+        # Execution control
+        priority: Execution order (higher = earlier, default 100)
+        enabled: Whether this config is active (default True)
+
+        # LLM provider configuration
+        provider_name: Optional LLM provider override (defaults to settings)
+        model_name: Optional model override (defaults to settings)
+
+    Inherited from CoreModel:
+        id, created_at, updated_at, deleted_at, tenant_id, user_id,
+        graph_edges, metadata, tags, column
+
+    Example Usage:
+        # CV extraction for recruitment
+        cv_config = OntologyConfig(
+            name="recruitment-cv-parser",
+            agent_schema_id="cv-parser-v1",
+            description="Extract candidate information from resumes",
+            mime_type_pattern="application/pdf",
+            uri_pattern=".*/resumes/.*",
+            tag_filter=["cv", "candidate"],
+            priority=100,
+            enabled=True,
+            tenant_id="acme-corp",
+            tags=["recruitment", "hr"]
+        )
+
+        # Contract analysis for legal team
+        contract_config = OntologyConfig(
+            name="legal-contract-analyzer",
+            agent_schema_id="contract-parser-v2",
+            description="Extract key terms from supplier contracts",
+            mime_type_pattern="application/(pdf|msword|vnd.openxmlformats.*)",
+            tag_filter=["legal", "contract"],
+            priority=200,  # Higher priority = runs first
+            enabled=True,
+            provider_name="openai",  # Override default provider
+            model_name="gpt-4o",
+            tenant_id="acme-corp",
+            tags=["legal", "procurement"]
+        )
+
+        # Medical records for healthcare
+        medical_config = OntologyConfig(
+            name="medical-records-extractor",
+            agent_schema_id="medical-parser-v1",
+            description="Extract diagnoses and treatments from medical records",
+            mime_type_pattern="application/pdf",
+            tag_filter=["medical", "patient-record"],
+            priority=50,
+            enabled=True,
+            tenant_id="healthsystem",
+            tags=["medical", "hipaa-compliant"]
+        )
+    
+
+## Overview
+
+The `OntologyConfig` entity is stored in the `ontology_configs` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on description (e.g., `SEARCH "concept" FROM ontology_configs LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM ontology_configs WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `ontology_configs` |
+| Entity Key | `id` |
+| Embedding Fields | `description` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `agent_schema_id`
+- **Type**: `<class ''str''>`
+- **Required**
+
+### `description`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+### `mime_type_pattern`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+### `uri_pattern`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+### `tag_filter`
+- **Type**: `list[str]`
+- **Optional**
+
+### `priority`
+- **Type**: `<class ''int''>`
+- **Optional**
+
+### `enabled`
+- **Type**: `<class ''bool''>`
+- **Optional**
+
+### `provider_name`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+### `model_name`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+
+',
+    '{"type": "object", "description": "User configuration for automatic ontology extraction.\n\n    Attributes:\n        name: Human-readable config name\n        agent_schema_id: Foreign key to Schema entity to use for extraction\n        description: Purpose and scope of this config\n\n        # File matching rules (ANY matching rule triggers extraction)\n        mime_type_pattern: Regex pattern for file MIME types (e.g., \"application/pdf\")\n        uri_pattern: Regex pattern for file URIs (e.g., \"s3://bucket/resumes/.*\")\n        tag_filter: List of tags (file must have ALL tags to match)\n\n        # Execution control\n        priority: Execution order (higher = earlier, default 100)\n        enabled: Whether this config is active (default True)\n\n        # LLM provider configuration\n        provider_name: Optional LLM provider override (defaults to settings)\n        model_name: Optional model override (defaults to settings)\n\n    Inherited from CoreModel:\n        id, created_at, updated_at, deleted_at, tenant_id, user_id,\n        graph_edges, metadata, tags, column\n\n    Example Usage:\n        # CV extraction for recruitment\n        cv_config = OntologyConfig(\n            name=\"recruitment-cv-parser\",\n            agent_schema_id=\"cv-parser-v1\",\n            description=\"Extract candidate information from resumes\",\n            mime_type_pattern=\"application/pdf\",\n            uri_pattern=\".*/resumes/.*\",\n            tag_filter=[\"cv\", \"candidate\"],\n            priority=100,\n            enabled=True,\n            tenant_id=\"acme-corp\",\n            tags=[\"recruitment\", \"hr\"]\n        )\n\n        # Contract analysis for legal team\n        contract_config = OntologyConfig(\n            name=\"legal-contract-analyzer\",\n            agent_schema_id=\"contract-parser-v2\",\n            description=\"Extract key terms from supplier contracts\",\n            mime_type_pattern=\"application/(pdf|msword|vnd.openxmlformats.*)\",\n            tag_filter=[\"legal\", \"contract\"],\n            priority=200,  # Higher priority = runs first\n            enabled=True,\n            provider_name=\"openai\",  # Override default provider\n            model_name=\"gpt-4o\",\n            tenant_id=\"acme-corp\",\n            tags=[\"legal\", \"procurement\"]\n        )\n\n        # Medical records for healthcare\n        medical_config = OntologyConfig(\n            name=\"medical-records-extractor\",\n            agent_schema_id=\"medical-parser-v1\",\n            description=\"Extract diagnoses and treatments from medical records\",\n            mime_type_pattern=\"application/pdf\",\n            tag_filter=[\"medical\", \"patient-record\"],\n            priority=50,\n            enabled=True,\n            tenant_id=\"healthsystem\",\n            tags=[\"medical\", \"hipaa-compliant\"]\n        )\n    \n\nThis agent can search the `ontology_configs` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"title": "Name", "type": "string"}, "agent_schema_id": {"title": "Agent Schema Id", "type": "string"}, "description": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Description"}, "mime_type_pattern": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Mime Type Pattern"}, "uri_pattern": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Uri Pattern"}, "tag_filter": {"default": [], "items": {"type": "string"}, "title": "Tag Filter", "type": "array"}, "priority": {"default": 100, "title": "Priority", "type": "integer"}, "enabled": {"default": true, "title": "Enabled", "type": "boolean"}, "provider_name": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Provider Name"}, "model_name": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "title": "Model Name"}}, "required": ["name", "agent_schema_id"], "json_schema_extra": {"table_name": "ontology_configs", "entity_key_field": "id", "embedding_fields": ["description"], "fully_qualified_name": "rem.models.entities.ontology_config.OntologyConfig", "tools": ["search_rem"], "default_search_table": "ontology_configs", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "ontology_configs", "entity_key_field": "id", "embedding_fields": ["description"], "fqn": "rem.models.entities.ontology_config.OntologyConfig"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Resource (resources)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'a579f379-4f1c-5414-8ff4-1382d0f783b7'::uuid,
+    'system',
+    'Resource',
+    '# Resource
+
+
+    Base content unit in REM.
+
+    Resources are content units that feed into dreaming workflows for moment
+    extraction and affinity graph construction. Tenant isolation is provided
+    via CoreModel.tenant_id field.
+    
+
+## Overview
+
+The `Resource` entity is stored in the `resources` table. Each record is uniquely
+identified by its `name` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by name (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM resources LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM resources WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `resources` |
+| Entity Key | `name` |
+| Embedding Fields | `content` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Human-readable resource name (used as graph label). Auto-generated from uri+ordinal if not provided.
+
+### `uri`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Content URI or identifier (file path, URL, etc.)
+
+### `ordinal`
+- **Type**: `<class ''int''>`
+- **Optional**
+- Chunk ordinal for splitting large documents (0 for single-chunk resources)
+
+### `content`
+- **Type**: `<class ''str''>`
+- **Optional**
+- Resource content text
+
+### `timestamp`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Resource timestamp (content creation/publication time)
+
+### `category`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Resource category (document, conversation, artifact, etc.)
+
+### `related_entities`
+- **Type**: `list[dict]`
+- **Optional**
+- Extracted entities (people, projects, concepts) with metadata
+
+',
+    '{"type": "object", "description": "\n    Base content unit in REM.\n\n    Resources are content units that feed into dreaming workflows for moment\n    extraction and affinity graph construction. Tenant isolation is provided\n    via CoreModel.tenant_id field.\n    \n\nThis agent can search the `resources` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Human-readable resource name (used as graph label). Auto-generated from uri+ordinal if not provided.", "entity_key": true, "title": "Name"}, "uri": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Content URI or identifier (file path, URL, etc.)", "title": "Uri"}, "ordinal": {"composite_key": true, "default": 0, "description": "Chunk ordinal for splitting large documents (0 for single-chunk resources)", "title": "Ordinal", "type": "integer"}, "content": {"default": "", "description": "Resource content text", "title": "Content", "type": "string"}, "timestamp": {"description": "Resource timestamp (content creation/publication time)", "format": "date-time", "title": "Timestamp", "type": "string"}, "category": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Resource category (document, conversation, artifact, etc.)", "title": "Category"}, "related_entities": {"description": "Extracted entities (people, projects, concepts) with metadata", "items": {"additionalProperties": true, "type": "object"}, "title": "Related Entities", "type": "array"}}, "required": [], "json_schema_extra": {"table_name": "resources", "entity_key_field": "name", "embedding_fields": ["content"], "fully_qualified_name": "rem.models.entities.resource.Resource", "tools": ["search_rem"], "default_search_table": "resources", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "resources", "entity_key_field": "name", "embedding_fields": ["content"], "fqn": "rem.models.entities.resource.Resource"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Schema (schemas)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    '2372e956-add6-58b8-a638-758a91a2b6c4'::uuid,
+    'system',
+    'Schema',
+    '# Schema
+
+
+    Agent schema definition.
+
+    Schemas define agents that can be dynamically loaded into Pydantic AI.
+    They store JsonSchema specifications with embedded metadata for tools,
+    resources, and system prompts.
+
+    For ontology extraction agents:
+    - `provider_configs` enables multi-provider support (test across Anthropic, OpenAI, etc.)
+    - `embedding_fields` specifies which output fields should be embedded for semantic search
+
+    Tenant isolation is provided via CoreModel.tenant_id field.
+    
+
+## Overview
+
+The `Schema` entity is stored in the `schemas` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM schemas LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM schemas WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `schemas` |
+| Entity Key | `id` |
+| Embedding Fields | `content` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+- Human-readable schema name (used as identifier)
+
+### `content`
+- **Type**: `<class ''str''>`
+- **Optional**
+- Markdown documentation and instructions for the schema
+
+### `spec`
+- **Type**: `<class ''dict''>`
+- **Required**
+- JsonSchema specification defining the agent structure and capabilities
+
+### `category`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Schema category distinguishing schema types. Values: ''agent'' (AI agents), ''evaluator'' (LLM-as-a-Judge evaluators). Maps directly from json_schema_extra.kind field during ingestion.
+
+### `provider_configs`
+- **Type**: `list[dict]`
+- **Optional**
+- Optional provider configurations for multi-provider testing. Each dict has ''provider_name'' and ''model_name''. Example: [{''provider_name'': ''anthropic'', ''model_name'': ''claude-sonnet-4-5''}]
+
+### `embedding_fields`
+- **Type**: `list[str]`
+- **Optional**
+- JSON paths in extracted_data to embed for semantic search. Example: [''summary'', ''candidate_name'', ''skills''] for CV extraction. Values will be concatenated and embedded using configured embedding provider.
+
+',
+    '{"type": "object", "description": "\n    Agent schema definition.\n\n    Schemas define agents that can be dynamically loaded into Pydantic AI.\n    They store JsonSchema specifications with embedded metadata for tools,\n    resources, and system prompts.\n\n    For ontology extraction agents:\n    - `provider_configs` enables multi-provider support (test across Anthropic, OpenAI, etc.)\n    - `embedding_fields` specifies which output fields should be embedded for semantic search\n\n    Tenant isolation is provided via CoreModel.tenant_id field.\n    \n\nThis agent can search the `schemas` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"description": "Human-readable schema name (used as identifier)", "title": "Name", "type": "string"}, "content": {"default": "", "description": "Markdown documentation and instructions for the schema", "title": "Content", "type": "string"}, "spec": {"additionalProperties": true, "description": "JsonSchema specification defining the agent structure and capabilities", "title": "Spec", "type": "object"}, "category": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Schema category distinguishing schema types. Values: ''agent'' (AI agents), ''evaluator'' (LLM-as-a-Judge evaluators). Maps directly from json_schema_extra.kind field during ingestion.", "title": "Category"}, "provider_configs": {"description": "Optional provider configurations for multi-provider testing. Each dict has ''provider_name'' and ''model_name''. Example: [{''provider_name'': ''anthropic'', ''model_name'': ''claude-sonnet-4-5''}]", "items": {"additionalProperties": true, "type": "object"}, "title": "Provider Configs", "type": "array"}, "embedding_fields": {"description": "JSON paths in extracted_data to embed for semantic search. Example: [''summary'', ''candidate_name'', ''skills''] for CV extraction. Values will be concatenated and embedded using configured embedding provider.", "items": {"type": "string"}, "title": "Embedding Fields", "type": "array"}}, "required": ["name", "spec"], "json_schema_extra": {"table_name": "schemas", "entity_key_field": "id", "embedding_fields": ["content"], "fully_qualified_name": "rem.models.entities.schema.Schema", "tools": ["search_rem"], "default_search_table": "schemas", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "schemas", "entity_key_field": "id", "embedding_fields": ["content"], "fqn": "rem.models.entities.schema.Schema"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for Session (sessions)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    '5893fbca-2d8e-5402-ac41-7bac2c0c472a'::uuid,
+    'system',
+    'Session',
+    '# Session
+
+
+    Conversation session container.
+
+    Groups messages together and supports different modes for normal conversations
+    and evaluation/experimentation scenarios.
+
+    For evaluation sessions, stores:
+    - original_trace_id: Reference to the original session being evaluated
+    - settings_overrides: Model, temperature, prompt overrides
+    - prompt: Custom prompt being tested
+
+    Default sessions are lightweight - just a session_id on messages.
+    Special sessions store additional metadata for experiments.
+    
+
+## Overview
+
+The `Session` entity is stored in the `sessions` table. Each record is uniquely
+identified by its `name` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by name (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on description (e.g., `SEARCH "concept" FROM sessions LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM sessions WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `sessions` |
+| Entity Key | `name` |
+| Embedding Fields | `description` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+- Session name/identifier
+
+### `mode`
+- **Type**: `<enum ''SessionMode''>`
+- **Optional**
+- Session mode: ''normal'' or ''evaluation''
+
+### `description`
+- **Type**: `str | None`
+- **Optional**
+- Optional session description
+
+### `original_trace_id`
+- **Type**: `str | None`
+- **Optional**
+- For evaluation mode: ID of the original session/trace being evaluated
+
+### `settings_overrides`
+- **Type**: `dict | None`
+- **Optional**
+- Settings overrides (model, temperature, max_tokens, system_prompt)
+
+### `prompt`
+- **Type**: `str | None`
+- **Optional**
+- Custom prompt for this session (can override agent prompt)
+
+### `agent_schema_uri`
+- **Type**: `str | None`
+- **Optional**
+- Agent schema used for this session
+
+### `message_count`
+- **Type**: `<class ''int''>`
+- **Optional**
+- Number of messages in this session
+
+### `total_tokens`
+- **Type**: `int | None`
+- **Optional**
+- Total tokens used in this session
+
+',
+    '{"type": "object", "description": "\n    Conversation session container.\n\n    Groups messages together and supports different modes for normal conversations\n    and evaluation/experimentation scenarios.\n\n    For evaluation sessions, stores:\n    - original_trace_id: Reference to the original session being evaluated\n    - settings_overrides: Model, temperature, prompt overrides\n    - prompt: Custom prompt being tested\n\n    Default sessions are lightweight - just a session_id on messages.\n    Special sessions store additional metadata for experiments.\n    \n\nThis agent can search the `sessions` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"description": "Session name/identifier", "entity_key": true, "title": "Name", "type": "string"}, "mode": {"$ref": "#/$defs/SessionMode", "default": "normal", "description": "Session mode: ''normal'' or ''evaluation''"}, "description": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Optional session description", "title": "Description"}, "original_trace_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "For evaluation mode: ID of the original session/trace being evaluated", "title": "Original Trace Id"}, "settings_overrides": {"anyOf": [{"additionalProperties": true, "type": "object"}, {"type": "null"}], "default": null, "description": "Settings overrides (model, temperature, max_tokens, system_prompt)", "title": "Settings Overrides"}, "prompt": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Custom prompt for this session (can override agent prompt)", "title": "Prompt"}, "agent_schema_uri": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Agent schema used for this session", "title": "Agent Schema Uri"}, "message_count": {"default": 0, "description": "Number of messages in this session", "title": "Message Count", "type": "integer"}, "total_tokens": {"anyOf": [{"type": "integer"}, {"type": "null"}], "default": null, "description": "Total tokens used in this session", "title": "Total Tokens"}}, "required": ["name"], "json_schema_extra": {"table_name": "sessions", "entity_key_field": "name", "embedding_fields": ["description"], "fully_qualified_name": "rem.models.entities.session.Session", "tools": ["search_rem"], "default_search_table": "sessions", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "sessions", "entity_key_field": "name", "embedding_fields": ["description"], "fqn": "rem.models.entities.session.Session"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for SharedSession (shared_sessions)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    'be5c5711-6c45-5fc4-9cd1-e076599261c7'::uuid,
+    'system',
+    'SharedSession',
+    '# SharedSession
+
+
+    Session sharing record between users.
+
+    Links a session (identified by session_id from Message records) to a
+    recipient user, enabling collaborative access to conversation history.
+    
+
+## Overview
+
+The `SharedSession` entity is stored in the `shared_sessions` table. Each record is uniquely
+identified by its `id` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by id (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on content (e.g., `SEARCH "concept" FROM shared_sessions LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM shared_sessions WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `shared_sessions` |
+| Entity Key | `id` |
+| Embedding Fields | None |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `session_id`
+- **Type**: `<class ''str''>`
+- **Required**
+- The session being shared (matches Message.session_id)
+
+### `owner_user_id`
+- **Type**: `<class ''str''>`
+- **Required**
+- User ID of the session owner (the sharer)
+
+### `shared_with_user_id`
+- **Type**: `<class ''str''>`
+- **Required**
+- User ID of the recipient (who can now view the session)
+
+',
+    '{"type": "object", "description": "\n    Session sharing record between users.\n\n    Links a session (identified by session_id from Message records) to a\n    recipient user, enabling collaborative access to conversation history.\n    \n\nThis agent can search the `shared_sessions` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "session_id": {"description": "The session being shared (matches Message.session_id)", "title": "Session Id", "type": "string"}, "owner_user_id": {"description": "User ID of the session owner (the sharer)", "title": "Owner User Id", "type": "string"}, "shared_with_user_id": {"description": "User ID of the recipient (who can now view the session)", "title": "Shared With User Id", "type": "string"}}, "required": ["session_id", "owner_user_id", "shared_with_user_id"], "json_schema_extra": {"table_name": "shared_sessions", "entity_key_field": "id", "embedding_fields": [], "fully_qualified_name": "rem.models.entities.shared_session.SharedSession", "tools": ["search_rem"], "default_search_table": "shared_sessions", "has_embeddings": false}}'::jsonb,
+    'entity',
+    '{"table_name": "shared_sessions", "entity_key_field": "id", "embedding_fields": [], "fqn": "rem.models.entities.shared_session.SharedSession"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
+-- Schema entry for User (users)
+INSERT INTO schemas (id, tenant_id, name, content, spec, category, metadata)
+VALUES (
+    '1ad3d95e-32e9-54d6-ad7d-e39b9ed5018b'::uuid,
+    'system',
+    'User',
+    '# User
+
+
+    User entity.
+
+    Represents people in the REM system, either as active users
+    or entities extracted from content. Tenant isolation is provided
+    via CoreModel.tenant_id field.
+
+    Enhanced by dreaming worker:
+    - summary: Generated from activity analysis
+    - interests: Extracted from resources and sessions
+    - activity_level: Computed from recent engagement
+    - preferred_topics: Extracted from moment/resource topics
+    
+
+## Overview
+
+The `User` entity is stored in the `users` table. Each record is uniquely
+identified by its `name` field for lookups and graph traversal.
+
+## Search Capabilities
+
+This schema includes the `search_rem` tool which supports:
+- **LOOKUP**: O(1) exact match by name (e.g., `LOOKUP "entity-name"`)
+- **FUZZY**: Typo-tolerant search (e.g., `FUZZY "partial" THRESHOLD 0.3`)
+- **SEARCH**: Semantic vector search on summary (e.g., `SEARCH "concept" FROM users LIMIT 10`)
+- **SQL**: Complex queries (e.g., `SELECT * FROM users WHERE ...`)
+
+## Table Info
+
+| Property | Value |
+|----------|-------|
+| Table | `users` |
+| Entity Key | `name` |
+| Embedding Fields | `summary` |
+| Tools | `search_rem` |
+
+## Fields
+
+### `id`
+- **Type**: `typing.Union[uuid.UUID, str, NoneType]`
+- **Optional**
+- Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.
+
+### `created_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Entity creation timestamp
+
+### `updated_at`
+- **Type**: `<class ''datetime.datetime''>`
+- **Optional**
+- Last update timestamp
+
+### `deleted_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Soft deletion timestamp
+
+### `tenant_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Tenant identifier for multi-tenancy isolation
+
+### `user_id`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.
+
+### `graph_edges`
+- **Type**: `list[dict]`
+- **Optional**
+- Knowledge graph edges stored as InlineEdge dicts
+
+### `metadata`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Flexible metadata storage
+
+### `tags`
+- **Type**: `list[str]`
+- **Optional**
+- Entity tags
+
+### `name`
+- **Type**: `<class ''str''>`
+- **Required**
+- User name (human-readable, used as graph label)
+
+### `email`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- User email address
+
+### `role`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- User role (employee, contractor, external, etc.)
+
+### `tier`
+- **Type**: `<enum ''UserTier''>`
+- **Optional**
+- User subscription tier (free, basic, pro) for feature gating
+
+### `anonymous_ids`
+- **Type**: `list[str]`
+- **Optional**
+- Linked anonymous session IDs used for merging history
+
+### `sec_policy`
+- **Type**: `<class ''dict''>`
+- **Optional**
+- Security policy configuration (JSON, extensible for custom policies)
+
+### `summary`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- LLM-generated user profile summary (updated by dreaming worker)
+
+### `interests`
+- **Type**: `list[str]`
+- **Optional**
+- User interests extracted from activity
+
+### `preferred_topics`
+- **Type**: `list[str]`
+- **Optional**
+- Frequently discussed topics in kebab-case
+
+### `activity_level`
+- **Type**: `typing.Optional[str]`
+- **Optional**
+- Activity level: active, moderate, inactive
+
+### `last_active_at`
+- **Type**: `typing.Optional[datetime.datetime]`
+- **Optional**
+- Last activity timestamp
+
+',
+    '{"type": "object", "description": "\n    User entity.\n\n    Represents people in the REM system, either as active users\n    or entities extracted from content. Tenant isolation is provided\n    via CoreModel.tenant_id field.\n\n    Enhanced by dreaming worker:\n    - summary: Generated from activity analysis\n    - interests: Extracted from resources and sessions\n    - activity_level: Computed from recent engagement\n    - preferred_topics: Extracted from moment/resource topics\n    \n\nThis agent can search the `users` table using the `search_rem` tool. Use REM query syntax: LOOKUP for exact match, FUZZY for typo-tolerant search, SEARCH for semantic similarity, or SQL for complex queries.", "properties": {"id": {"anyOf": [{"format": "uuid", "type": "string"}, {"type": "string"}, {"type": "null"}], "default": null, "description": "Unique identifier (UUID or string, generated per model type). Generated automatically if not provided.", "title": "Id"}, "created_at": {"description": "Entity creation timestamp", "format": "date-time", "title": "Created At", "type": "string"}, "updated_at": {"description": "Last update timestamp", "format": "date-time", "title": "Updated At", "type": "string"}, "deleted_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Soft deletion timestamp", "title": "Deleted At"}, "tenant_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Tenant identifier for multi-tenancy isolation", "title": "Tenant Id"}, "user_id": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Owner user identifier (tenant-scoped). This is a VARCHAR(256), not a UUID, to allow flexibility for external identity providers. Typically generated as a hash of the user''s email address. In future, other strong unique claims (e.g., OAuth sub, verified phone) could also be used for generation.", "title": "User Id"}, "graph_edges": {"description": "Knowledge graph edges stored as InlineEdge dicts", "items": {"additionalProperties": true, "type": "object"}, "title": "Graph Edges", "type": "array"}, "metadata": {"additionalProperties": true, "description": "Flexible metadata storage", "title": "Metadata", "type": "object"}, "tags": {"description": "Entity tags", "items": {"type": "string"}, "title": "Tags", "type": "array"}, "name": {"description": "User name (human-readable, used as graph label)", "entity_key": true, "title": "Name", "type": "string"}, "email": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "User email address", "title": "Email"}, "role": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "User role (employee, contractor, external, etc.)", "title": "Role"}, "tier": {"$ref": "#/$defs/UserTier", "default": "free", "description": "User subscription tier (free, basic, pro) for feature gating"}, "anonymous_ids": {"description": "Linked anonymous session IDs used for merging history", "items": {"type": "string"}, "title": "Anonymous Ids", "type": "array"}, "sec_policy": {"additionalProperties": true, "description": "Security policy configuration (JSON, extensible for custom policies)", "title": "Sec Policy", "type": "object"}, "summary": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "LLM-generated user profile summary (updated by dreaming worker)", "title": "Summary"}, "interests": {"description": "User interests extracted from activity", "items": {"type": "string"}, "title": "Interests", "type": "array"}, "preferred_topics": {"description": "Frequently discussed topics in kebab-case", "items": {"type": "string"}, "title": "Preferred Topics", "type": "array"}, "activity_level": {"anyOf": [{"type": "string"}, {"type": "null"}], "default": null, "description": "Activity level: active, moderate, inactive", "title": "Activity Level"}, "last_active_at": {"anyOf": [{"format": "date-time", "type": "string"}, {"type": "null"}], "default": null, "description": "Last activity timestamp", "title": "Last Active At"}}, "required": ["name"], "json_schema_extra": {"table_name": "users", "entity_key_field": "name", "embedding_fields": ["summary"], "fully_qualified_name": "rem.models.entities.user.User", "tools": ["search_rem"], "default_search_table": "users", "has_embeddings": true}}'::jsonb,
+    'entity',
+    '{"table_name": "users", "entity_key_field": "name", "embedding_fields": ["summary"], "fqn": "rem.models.entities.user.User"}'::jsonb
+)
+ON CONFLICT (id) DO UPDATE SET
+    name = EXCLUDED.name,
+    content = EXCLUDED.content,
+    spec = EXCLUDED.spec,
+    category = EXCLUDED.category,
+    metadata = EXCLUDED.metadata,
+    updated_at = CURRENT_TIMESTAMP;
+
 -- ============================================================================
 -- RECORD MIGRATION
 -- ============================================================================