Getting Started
Key Features
Use MemorySync to store useful context, retrieve it by meaning, and keep each request inside the correct project and end-user scope.
Memory operations
| Capability | Use it to |
|---|---|
| Add | Store a durable fact with an optional source, tags, importance, and metadata. |
| Query | Retrieve relevant records with a natural-language question and optional filters. |
| Update | Change editable metadata, tags, source, event type, or importance on a record. |
| Forget | Remove selected memory records from the active memory set. |
| Summarize | Create a condensed record from selected memory IDs. |
| Export | Retrieve the memory data available to the authenticated scope. |
Relevant results, not full history
Send the question your application is answering. MemorySync returns candidate records from the same authenticated scope, ordered for relevance. Your application decides which results belong in the final prompt or experience.
Project and end-user separation
- An API key authenticates your trusted server.
- A project identifier selects the application or workload.
- An end-user identifier selects the person your server is acting for.
- The same project and end-user scope must be used when writing and retrieving a user’s memory.
Application-defined context
- Use
sourceto identify where a memory came from, such aschatorprofile. - Use metadata for non-secret identifiers your application needs for tracing or filtering.
- Use tags only when they have a stable meaning in your product.
- Keep permissions, credentials, and authoritative business records in their source systems.
Try the core flow
Was this page helpful?