Provision a new knowledge graph project. Creates a Neo4j graph database with a knowledge-optimized schema (Topics, Articles, Authors, Concepts) and deploys it to staging. May take 30-60 seconds.

List all graph projects. Shows project IDs, names, codes, and status.

Delete a graph project and ALL its data. DESTRUCTIVE — cannot be undone.

Bulk create entities in the knowledge graph. The LLM extracts entities from content and provides them here. Each entity needs an entity_id (kebab-case), entity_type (matching schema — e.g., 'Topic', 'Article', 'Author', 'Concept'), and properties dict matching that type's schema fields.

Use get_knowledge_schema first to see available entity types and their fields.

Bulk create relationships between entities in the knowledge graph. Each relationship needs from_id, to_id (matching existing entity_ids), rel_type (matching schema — e.g., 'AUTHORED', 'COVERS', 'REFERENCES'), and optional properties.

Use get_knowledge_schema first to see available relationship types.

Full-text search across all string properties in the knowledge graph. Returns matching nodes with their types, properties, and relevance scores.

Get the full schema — entity types with fields, relationship types with from/to mappings. CALL THIS FIRST before adding nodes or relationships to understand what types and fields are available.

Get node/relationship counts broken down by type. Useful for understanding the graph's size and composition.

Walk the graph from a starting entity, following relationships up to a specified depth. Returns connected nodes and relationships.

List entities of a specific type. Use get_knowledge_schema first to see available entity types.

Get a single entity by type and ID, with all properties.

BATCH INGESTION — the fast way to populate a knowledge graph.

Provide a large block of text (up to 500k chars) and the project code. This tool fetches the graph schema and returns structured extraction instructions. Then call add_knowledge_nodes and add_knowledge_relationships with the extracted data.

3-CALL WORKFLOW:

1. ingest_text_content(project_code, text) → schema + instructions

2. add_knowledge_nodes(project_code, entities) → bulk create nodes

3. add_knowledge_relationships(project_code, relationships) → bulk create edges

This replaces per-entity approach. Extract EVERYTHING from the text in one pass, then write it all in two bulk calls.

Scrape a URL and extract clean text content. Returns the text for the LLM to read, extract entities from, and then call add_knowledge_nodes/relationships. Also returns metadata (title, author, date).