Agent memory notes: a persistent folder per bot

Open
  • Is something wrong?
#39 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
45/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Active
Tech stack
javascript
Domain
backend

Research direction

Start by reading the existing memory module, the bot instruction setup, and how agents write persistent runner files; the issue does not name their paths. Build an off-by-default module with a per-bot folder that agents can read and write, and add one instruction line about using it across threads. Test whether Luna saves useful notes from thread 37315 and retrieves them in a fresh thread; defer offline review of notes.

Written by the indexing model from the issue text.

Description

Explore the ideas in Anthropic’s Components of agent memory.

The components and where we stand:

  • Transcripts: searchable event logs. We keep threads and agent conversations; threads reads and searches other threads.
  • Memory notes: observations an agent chooses to save across sessions. No dedicated notes store yet, though agents can write persistent runner files.
  • Memory representations: structures such as linked wikis. Our memory module has a summary tree and zoom, not a cross-thread knowledge wiki.
  • Injected context: memory loaded into a session. Our tree supplies catch-up and compaction context, not a curated cross-session notes index.
  • Dreaming: offline review that corrects notes, builds representations and promotes useful context. We do not have this.

Three passages worth keeping:

It’s also reasonable to consider transcript search as the foundation of Claude’s memory.

With frontier Claude models, we typically give simple guidance to create memory notes when a person asks them to or when something is clearly worth remembering, and we leave that judgment up to Claude.

Claude started each question in a fresh session, with one exception: it could retrieve files from a memory directory that persisted from one session to the next.

Try an off-by-default module that gives each bot a memory folder it reads and writes, plus one line in its instructions telling it the folder is available for memories across threads. Transcripts and the summary tree stay as they are: retaining everything and choosing useful notes are complementary.

Test whether Luna saves the right things unprompted (for example, which museum, from thread 37315), and whether it reads those notes in a fresh thread.

Later, maybe: dreaming, meaning an offline review that tidies the notes.

Dominant language
JavaScript
Stars
0
Forks
0
Avg merge
23m
Merged PRs (30d)
8

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from oskarrough/agent-ide

All issues in oskarrough/agent-ide

Similar issues

More JavaScript issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.