I was a long time Evernote user and when they changed their pricing looked for an alternative. I am quite happy with Joplin. I sync to S3 and everything works well although I am still trying to figure out which plugins will really help me.
My problem is my own penchant to save everything. Whne I exported Evernote I relaized I have so much crap in my files and nothing is really indexed or cross indexed properly. I don't relish wding through thousands of notes manually though. I also have the same problem with photos, having about 77,000 in my library of which I think 10% are keepers.
I have been playing with ollama to try to keyword and tag my images and am making progress although there are a lot of weirdisms. For example a page of sheet music was recognized by llava as a bunch of people on a patio and by gemma3 as Fiordland National Park in New Zealand.
Anyone figured out an easy way to properly tag a bunch of joblin notes using one of the olama LLM's? I am hopeing that will speed up my progress in deleting old or un-necessary notes.
I know nothing about LLaVA (https://llava-vl.github.io/ ?) and ollama (https://ollama.com/ ?).
But If you need to manage thousands of images, you need ACDSEE photo studio.
It has his own database.
See also:
FASTSTONE is an interesting "ACDSEE like" free and powerfull software:
For text content, I'd be curious to hear your thoughts on the annotation features of the Jarvis plugin, which include tagging, summaries, re-titling of notes, and links to related notes. You can set it up with Ollama. I hope it might be useful, although I haven't tried it on such a large scale.
Other than annotation, you could use it, for example, to chat with your notes.
Except the text of all the notes would be fed to some process that would come up with the most popular tags, perhaps 20 or 30 based on both the content and a history of previous queries and searches.
Ot would be like re-creating the titles for the holes in the cards like the physical example in the link.
Since we go on to different research or topics the top 20 or 30 will change. For example in my clean water project there may be a lot of focus on specific chemicals like bifenthrin for a while but then we may focus on biological toxins like domoic acid and later on physical characteristics like salinity. Obviously one could just build a concordance or index or do a search but sometimes that yields too many hits to be useful.
What i am trying to day, but poorly, is for the system to present limited amount of tags and return a limited number of results based on what I have looked for most recently and most often. Maybe like computer cache management. I know i am not being clear because i am not exactly sure of what i want. Clairvoyance comes the closest.
If you could somehow grab statistics from Evernote in terms of last opened date time & amount of openings that would be great - I doubt they provide such though.
You could add them correspondinly to Joplin note properties which is going to help limit the number of tags.
Otherwise you're going to need to generate big list of tags and filter it manually. Here's how I would do it through some automated script.
Summary
Phase 1: Generate Initial Tag Suggestions
Fetch All Notes
Use the Joplin Data API to retrieve all existing notes.
Generate Tag Suggestions
For each note, send its content (or excerpt) to an LLM with a prompt like:
"Based on this note content, suggest a relevant tags. Return only 5 consice tags."
Collect Tags
Save the returned tag suggestions to a text file — one tag per line.
Manual Cleanup
Review the list of generated tags manually to:
Remove duplicates (e.g., “project” vs. “projects”)
Eliminate irrelevant or overly generic tags (like "note", "text", etc.)
Standardize capitalization or naming conventions (e.g., AI, ai, and A.I. become AI)
Phase 2: Assign Cleaned Tags to Notes
Load Clean Tag List
Load the cleaned and curated list of tags from the file.
Re-evaluate Each Note
For each note, send its content to the LLM with a prompt like:
"Based on this note content, which one of the following tags is most appropriate? Choose up to 5. [list of cleaned tags]"
Apply Tags via API
Use the Joplin API to assign the returned tag to the corresponding note.
Hmm, could be interesting to have an embedded tag that counts the number of times the note was opened and keeps the last date it was viewed (I'm toying with all kinds of tags). Then you could sort results by these tags.
Instead of tags, perhaps semantic search could help. Given some context (e.g., the note or section that you currently work on, or an explicit prompt), the search should provide the most similar sections in your note collection. In addition, you could let AI sift through the notes that come up, refine the search, and suggest what's relevant.
I don't think you're going to find a solution for the "too much" problem. I surely have not. I think the approach we need for Joplin, and any note taking app, is a powerful search that integrates a tuned LLM specifically for analyzing and tagging documents and figuring out their content.
This search/LLM will look at content, tags, datres, etc. to serve up what it thinks you might be looking for based on your search term, and will allow far better and more robust search terms.
An automatic deep archive function to pack away notes from a specified amount of time (month, year, etc.) with an option to search those if needed. Keep everything locked down with encryption when at rest.
There's simply no way to not keep adding to the pile as that's the whole purpose of having the data and to undermine that I think is the wrong approach.
Phase 1: Generate Initial Tag Suggestions
Phase 2: Assign Cleaned Tags to Notes