Persistent Memory¶
RagLeap's AI remembers important facts across conversations — not just within a single session, but permanently, across every channel.
How it works¶
- Store — a fact, note, or preference is saved as a memory entry, either automatically (from a conversation) or manually
- Embed — the memory is converted into a vector embedding for semantic search
- Retrieve — when a question comes in, relevant memories are found by meaning, not just keyword matching
- Use — retrieved memories are injected into the AI's context so it gives personalized, consistent answers
Memory scopes¶
| Scope | Visibility |
|---|---|
| User-level | Shared across all workspaces for one user — good for personal preferences |
| Workspace-level | Visible to everyone in that workspace — good for shared business facts (e.g. policies, escalation rules) |
Auto-compression¶
When a memory entry grows too large, it's automatically summarized to save space while keeping the key information intact.
Manage your memory¶
At /memory/ you can view every stored memory, its scope, token size, and embedding status — search, edit, or delete entries directly, export everything as JSONL, or purge all memory for a clean start.
Tip
Memory settings (default scope, token limits, retention period) are configurable per workspace, so you control exactly how long facts persist and how large a single memory entry can grow.