Use Chroma to augment and make LocalCache memory/embeddings pluggable

[swyxio]

Background

The LocalCache memory backend implemented in #372 used a handrolled implementation of a storage engine, and the embedding strategy was pinned to text-embedding-ada-002 which we may want to change/does cost money/latency

Changes

• reimplemented LocalCache using Chroma - which also runs either in-memory or persisted, but is well tested and managed.
• made embeddings swappable to ANY openai embedding model - and uses the Chroma default SentenceTransformers which costs no money

Documentation

in README

Test Plan

testing it locally

no new tests needed since this is a strict upgrade of LocalCache.

PR Quality Checklist

• My pull request is atomic and focuses on a single change.
• I have thoroughly tested my changes with multiple different prompts.
• I have considered potential risks and mitigations for my changes.
• I have documented my changes clearly and comprehensively.
• I have not snuck in any "extra" small tweaks changes

[swyxio]

its ready for review :)

[aosan]

great addition @sw-yx ! I hope it gets merged soon.

[nponeccop]

@sw-yx There are conflicts

[swyxio]

whaaat i just made this PR yesterday and there are conflicts haha. will fix!

[nponeccop]

We move fast, new conflicts can happen every 5 minutes here :)

[swyxio]

@nponeccop i think i fixed the conflict!

[swyxio]

re-fixed new conflicts! @nponeccop what typically happens next?

[nponeccop]

@sw-yx There are conflicts now

[swyxio]

@nponeccop as far as i can tell i fixed the conflicts as of commit a6928c2. i've just updated my fork to current master but still not seeing any conflicts as of right now!

[nponeccop]

@sw-yx There are both conflicts and CI failure now

[github-actions]

Conflicts have been resolved! 🎉 A maintainer will review the pull request shortly.

[github-actions]

This PR exceeds the recommended size of 1000 lines. Please make sure you are NOT addressing multiple issues with one PR. Note this PR might be rejected due to its size.

[swyxio]

This PR exceeds the recommended size of 1000 lines.

no, it only has ~100 lines changed...?

[github-actions]

This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.

[github-actions]

Conflicts have been resolved! 🎉 A maintainer will review the pull request shortly.

[github-actions]

This pull request has conflicts with the base branch, please resolve those so we can evaluate the pull request.

[p-i-]

This is a mass message from the AutoGPT core team.
Our apologies for the ongoing delay in processing PRs.
This is because we are re-architecting the AutoGPT core!

For more details (and for infor on joining our Discord), please refer to:
https://github.com/Significant-Gravitas/Auto-GPT/wiki/Architecting

[Pwuts]

First of all: sorry for leaving this unresolved for so long. This is an interesting PR and unfortunately I'm only seeing it now for the first time.

Since this PR, we have rebuilt the memory storage system (#4208) and removed all third-party memory stores, to make room for redeveloping the entire memory system without having to worry about maintaining multiple memory stores.
Once we have a PoC of Auto-GPT working with vector memory again, I'd be interested to see a comparison with a simpler implementation using Chroma.

For now, I can only close this, as the PR is based on a very much outdated version of the codebase and because we want to have only one memory backend until the memory system is in better shape. We would still be happy to have your involvement; feel free to join in on the action via discord!