I found it !! “Auto Classifier” Plugin in community plugin is what you are saying.
It is an Obsidian plugin that helps you automatically classify tag from your notes using the ChatGPT API. The plugin can analyze your note (It can be title, frontmatter, content or selected area) and suggest relevant tag based on the input with tags in your valut.
This can be used for various and specific purposes, for example, DDC classification for books, keyword recommendation, research paper categorization and so on. Save you time and improving your note organization.
This doesn’t seem to produce very good results. I set it to use tags from the vault but in only adds very few if any tags that I want it to add or that it should in my opinion add.
I didn’t have much luck with the community plugin ones as they can produce some odd results, thus they’re more of “tag suggestions” for the note you’re working on versus a tool to auto-tag a large archive of notes.
The key is that you need consistency for the tags to be relevant. So you have to be able to restrict the dictionary space that it distills down to, which introduces a the issue of it needing to be able to learn on your notes as you progress. This is really something that non-Obsidian tools would be best used for, such as BERTopic / BERT KeyPhrase.
Otherwise some form of pattern reduction UI may work. Where you supply comma separated list of patterns and a tag. If one of those patterns is seen in a note, it gets a tag added to the note.
Hi! I discovered an extension that enables you to save ChatGPT conversations directly to an Obsidian markdown file. It might be useful to you. Give it a try here.
Ideally the tags would be generated according to a controlled vocabulary such as that provided by the Library of Congress Subject Headings or the Universal Decimal Classification.
It would be great if Obsidian could work as Napkin app for auto-tagging notes.
Previously I used most of the time Google Keep to capture ideas and small fragments of the development of these ideas, to later elaborate them furthermore, restructuring the content, organising it, etc. in Obsidian. Google Keep has a manual tag system (as Obsidian’s and most of the note-taking app’s one), so I used more the search function to filter the notes.
I have almost replaced Keep with Obsidian with the Unique note creator core plugin of Obsidian, which allows me to create notes faster without needing to think about a title. I use in those notes the front matter to add tags, but I have the same problem that I previously had with Keep: the time I save by not having to think about the title of the note I dedicate to adding the tags to them. From that perspective, Napkin is a very great tool because the unique thing you have to do is taking the notes and filter them later to expanding or consulting them. The app will do the other work for you. But, it could be very useful if Obsidian could have this functionality, so you don’t have to take the notes in two different apps or systems.
To this day, there are 2 Obsidian plugins that work with ChatGPT API using the note’s content or part of the content as prompts that are given to the AI: Auto Classifier and Auto Tag. I haven’t tried them yet, so I can’t give my opinion, but it would be very useful to be able to use them with a local LLM using an Ollama local chatbot model. Or, to “simplify” the process and not to require a LLM running locally or with remote connection, a local algorithm to search in the note content words or similar words based on a dictionary, to then add the tags (previously created).