Memory
Pia’s Memory feature lets you store structured personal knowledge that the AI assistant can reference during conversations.
Overview
Section titled “Overview”Memories are pieces of information — contacts, projects, preferences, notes — organized by type. The AI assistant uses them to give you more relevant, context-aware responses.
Browsing Memories
Section titled “Browsing Memories”Open Memory from the sidebar to see all your stored items, grouped by type.
- Search — use the search bar to find memories by keyword. If you have downloaded an embedding model, Pia uses hybrid search (semantic + keyword) for more accurate results.
- Details — click a memory to view its content in the detail panel on the right.
Editing
Section titled “Editing”Click the edit button on a memory to enter edit mode. Depending on the memory type, you have two editing modes:
- Form view — if the memory has a schema, Pia renders a structured form with validation. Fields are type-checked and required fields are enforced.
- JSON view — a raw JSON editor for direct editing. Use the toggle button to switch between form and JSON views.
@ Commands
Section titled “@ Commands”In the Assistant chat, type @ to reference memories (and other items) directly in your message:
| Syntax | Example | Description |
|---|---|---|
@Memory | @Memory | Reference the memory domain |
@Memory:Name | @Memory:ProjectAlpha | Reference a specific memory |
@Memory:"Multi Word" | @Memory:"My Project" | Quote titles with spaces |
You can also use @Todo and @Reminder to reference those items. An autocomplete popup appears as you type to help you find the right item.
Embedding Model
Section titled “Embedding Model”For enhanced search, you can download a local embedding model:
- Go to Memory view
- Click Download Embedding Model
- Once downloaded, search uses hybrid matching (semantic similarity + keyword)
The model runs locally — no data is sent to external services.
Maintenance
Section titled “Maintenance”- Export — export all memories as a file.
- Regenerate Embeddings — rebuild search embeddings after a model update.
- Review Stale — find and clean up memories not accessed in over 90 days.
Cloud Sync
Section titled “Cloud Sync”Memories sync automatically when Cloud Sync is enabled. With E2E encryption on Pro plans, memory content is encrypted before upload.