Progress
- Introductory call with Bill (Ton Hoang) (my mentor)
- Introductory meeting with all the mentors and other contributors
- Went through the plugin creation tutorial again (also referred to it while preparing the proposal)
- Tested the embedding model again
Discussion in the Introductory call with Bill (Ton Hoang)
- Weekly updates on Discourse
- Weekly meetings on Discord
- Push weekly updates on Discourse before the meeting
- Discussion regarding AI usage
Plan for Next Week
- Engage more actively with the community on Discourse, Discord, GitHub, etc.
- Finalize who will create and manage the plugin repo.
- Start coding, as I have already set up Transformers.js in the plugin during the proposal making.
Problems
While preparing the proposal, I set up the Joplin plugin environment with Transformers.js and tested different embedding models to compare their performance. I noticed that embedding a large number of notes was taking a lot of time, even without chunking. After testing it again recently, I observed the same issue. I think it would be better to first implement the core functionality and optimize the embedding strategy later.
GitHub repo: testing-embedding-model (readme of the repo contains all testing data)