mirror of
https://github.com/mudler/LocalAI.git
synced 2026-11-07 21:13:27 -05:00
Ettore Di Giacinto
7079666185
Declare the save-mode user setting (ask, auto, never; default ask) and edit it on the Short memory page with the settings kit, a line for each option, and the rule that messaging apps and background runs always ask. The page reads and writes through sdk.userSettings and keeps the earlier choice on screen when a save fails. The memory skill now says what to tell the user for each answer of propose_change. Assisted-by: Claude:claude-sonnet-5-5 node
92 lines
4.5 KiB
Markdown
92 lines
4.5 KiB
Markdown
---
|
|
title: "Short memory"
|
|
weight: 36
|
|
---
|
|
|
|
Short memory is an add-on that gives your chats a few notes about you. The model reads
|
|
the note that matches a request before it answers, and by default it saves a new note only
|
|
after you confirm it. You can change that on the add-on page. Install `memory` from the add-on library. Its visible name is **Short
|
|
memory**.
|
|
|
|
It is for the facts that change what a good answer is: where you live, what you eat, the
|
|
language and units you use, how you want answers. It is not a store for large documents.
|
|
Use [Collections](/features/collections/) to search many files.
|
|
|
|
## How it works
|
|
|
|
Each turn, the host puts an index of your notes in the prompt, under a fixed header that
|
|
asks the model to read the matching topic before it answers. The index is a list of lines
|
|
such as `- home: lives in Rome -> home.md`. The model opens a topic with the `context_read`
|
|
tool. The add-on ships two skills, `memory` and `memory-check-own-files`, that tell the
|
|
model to consult memory for almost every request and to say what it can and cannot read.
|
|
|
|
There are three kinds of notes, each in the Memory folder of your files:
|
|
|
|
| Kind | File | Applies to |
|
|
|---|---|---|
|
|
| Yours | `Memory/index.md` | every chat |
|
|
| One saved agent | `Memory/agents/<agent>/index.md` | the chats of that agent |
|
|
| One conversation | `Memory/channels/<binding>-<hash>/index.md` | one conversation that came from a messaging app |
|
|
|
|
An index has at most 40 lines and 2 KiB. All indexes of a turn together have at most 6 KiB.
|
|
The details of a topic sit in a file next to the index.
|
|
|
|
## Saving and forgetting
|
|
|
|
When you ask the model to remember something, it calls `propose_change`. By default
|
|
nothing is written at once. You confirm the proposal on the card in the chat or in the
|
|
inbox, and it expires after 24 hours. A confirmed change is one change set in your files,
|
|
so you can undo it. To forget a note you just saved, the model calls `revert_change`. For
|
|
an older note it proposes a shorter file, and you confirm that too.
|
|
|
|
A file over its cap is refused, and the answer lists the current entries so the model
|
|
merges them.
|
|
|
|
### Choose how a note is saved
|
|
|
|
On the page of the add-on, under **Saving a note**, you choose what happens when the
|
|
assistant wants to save something to your memory:
|
|
|
|
| Choice | What happens |
|
|
|---|---|
|
|
| **Ask me first** (default) | The assistant proposes the note and you confirm it. |
|
|
| **Save it and tell me** | In your own chats the note is saved at once. The chat shows "Saved to memory" with an **Undo**, and no confirmation card. |
|
|
| **Never** | The assistant cannot propose a note. It says that saving is off if you ask it to remember something. |
|
|
|
|
The choice is yours alone and the server keeps it, not the add-on. These rules hold
|
|
whatever you choose:
|
|
|
|
- Chats that come from a messaging app always ask first. The person who wrote the
|
|
message is not always you, so a message from outside can never save to your memory
|
|
without your confirmation. Scheduled and background runs cannot save notes at all.
|
|
- A saved agent cannot widen your choice. If you choose to be asked, it asks, even when
|
|
you let it write to memory.
|
|
- A reply saves at most 5 notes at once. A long note, or one that looks like it holds a
|
|
secret such as a key or a password, waits for your confirmation instead.
|
|
- An automatic note is one change set like any other. The Undo on the chat, the model's
|
|
`revert_change` and the Files history restore the earlier text. Every automatic note
|
|
is in the audit of changes (`GET /api/chats/changes/audit`).
|
|
|
|
If the server cannot read your choice, it asks first.
|
|
|
|
## Saved agents and messaging apps
|
|
|
|
A saved agent reads notes only when you allow the Memory add-on for it in the agent
|
|
editor. A chat that came from a messaging app reads your notes and the notes of its own
|
|
conversation, and never the notes of another conversation. A saved agent can only
|
|
propose a change when the owner granted it the right.
|
|
|
|
## Your control
|
|
|
|
The notes are plain files. Read, edit and delete them on the **Memory** tab of your account
|
|
or on the Files page. Stopping or removing the add-on stops the model from reading them; it
|
|
does not delete them.
|
|
|
|
## For add-on authors
|
|
|
|
The add-on is a package like the others: a manifest, two skills and an executable that
|
|
serves no tool. It declares `capabilities.contextFiles`, so the host loads the index files.
|
|
See [Context files](/features/addons/#context-files). It also declares one user setting,
|
|
`save-mode` (`ask`, `auto`, `never`), which the page edits through `sdk.userSettings`; see
|
|
[User settings](/features/addons/#user-settings). The manifest needs `api` 2.
|