Built by Metorial, the integration platform for agentic AI.
Configure or retrieve custom entity types and edge types for knowledge graphs. Setting an ontology defines domain-specific graph structure that improves precision and relevance of agent memory. Can also list all currently configured entity and edge types.
Delete an episode from the knowledge graph by its UUID. Cascading delete removes edges created solely by this episode and nodes that are no longer referenced by any other episodes.
Retrieve assembled context for a thread. Returns a context block combining semantic search, full-text search, and breadth-first search results from the user's knowledge graph, optimized for relevance to the current conversation. Supports custom context templates.
Clone a user or group knowledge graph into a new standalone graph. Useful for creating copies of a user's knowledge for testing, migration, or sharing purposes.
Search a user's or group's knowledge graph using hybrid semantic and full-text search. Can search across nodes (entities), edges (facts/relationships), or episodes (messages/data chunks). Supports multiple reranking strategies and filtering options.
Add a structured fact triple to a user's or group's knowledge graph. A fact triple consists of a source entity, a relationship, and a target entity. Use this to ingest business data, user interactions, transactions, or any structured knowledge.
Browse nodes, edges, and episodes in a user's knowledge graph. Retrieve all nodes for a user, get edges connected to a specific node, or get episodes that mention a node. Useful for understanding the structure and contents of a graph.
Retrieve all conversation threads belonging to a specific user. Returns the complete list of threads associated with the user, useful for reviewing conversation history or managing user data.
Create, retrieve, list, or delete conversation threads. Threads represent conversations between a user and your agent, and serve as containers for chat messages that feed into the knowledge graph.
Add one or more chat messages to a thread. Messages are stored in thread history and automatically processed into the user's knowledge graph. Supports batch mode for concurrent processing of large message sets.
Create, retrieve, update, or delete a user in Zep. Users represent application end-users and are the foundation for user-specific knowledge graphs and conversation threads.
Configure custom instructions for how Zep generates summaries of user data in their knowledge graph. Use this to tailor summaries to emphasize information most relevant to your application.