Types / Schemas
Define the structure and validation rules for your organizational knowledge. Schemas are typed contracts that describe what data looks like before it enters memory.
Memory & Knowledge
Memory
Knowledge
Memory is a semantic storage layer for structured business data that your agents can query during conversations. Instead of relying on static prompts or external databases, agents discover context incrementally by searching typed objects and traversing a knowledge graph — giving them access to the right information at the right time.
Core building blocks
Section titled “Core building blocks”Objects
Store actual business data as instances of schemas. Objects contain the content your agents read and reference — product specs, troubleshooting guides, FAQ entries, and more.
Knowledge Graph
Connect objects through typed relationships (edges) so agents can traverse related information. A troubleshooting guide links to the symptoms it resolves; a product links to its documentation.
Memory Buckets
Execution-scoped storage for agent pass-by-reference data. Buckets are created per agent execution and store intermediate results that tools can reference during the same run.
Explorer
Browse and organize memory objects in a visual folder hierarchy. The Explorer provides a file-system-like interface for navigating large knowledge bases.
How data flows into memory
Section titled “How data flows into memory”-
Define types (schemas). Create schemas that describe the structure of your organizational knowledge — fields, property types, and validation rules.
-
Create objects conforming to types. Add objects manually or import them from documents. Each object is an instance of a schema and contains actual business data.
-
Connect objects via edges in the graph. Link related objects with typed relationships. A FAQ entry can reference a product; a troubleshooting guide can link to known symptoms.
-
Agents query memory at runtime. During conversations, agents use semantic search and graph traversal to find relevant context — no pre-built prompts required.
Common use cases
Section titled “Common use cases”| Use case | Description |
|---|---|
| Troubleshooting guides | Step-by-step resolution procedures linked to symptoms and error codes |
| Product catalogs | Structured product information with specifications, pricing, and related documentation |
| FAQ databases | Question-answer pairs organized by topic, searchable by agents during support conversations |
| Process documentation | Standard operating procedures and workflows that agents can reference and walk users through |
| Customer information | Account details, service history, and preferences that provide personalized context |
How agents use memory
Section titled “How agents use memory”Agents interact with memory through dedicated tools during conversations:
- queryMemory — semantic search across all memory objects, filtered by type and content. Agents use this to find relevant information based on the user’s question.
- neighborhoodRag — retrieval-augmented generation that walks the knowledge graph from a known object, finding contextually related information by traversing edges.
- storeObject — save new knowledge into memory during a conversation, allowing agents to capture information for future reference.
Next steps
Section titled “Next steps” Schemas Define data types and schemas to structure your organizational knowledge.
Explorer Browse and manage memory objects in a visual tree structure.
Objects Create, import, and manage memory objects that store your knowledge.
Knowledge Graph Connect memory objects through typed relationships in a knowledge graph.