I have taken a different approach (on Linux).
I use multiple AI features (ASR, local LLM chat interaction, text summarizations, etc.) for Joplin but outside of the application.
These AI tools target any open editable Linux window, thus, including Joplin , which happens to be one of the main beneficiaries of these tools.
Simmarization:
I use a GNOME extension called voluble that takes advantage of a local llama.cpp installation or a lamafile for the LLM summarization function. One selects a text anywhere and a summary is created ready to be pasted.
Speech Input (ASR):
This is provided by Blurt (GNOME extension) or more universally, for any Linux desktop, by BlahST, where speech recording starts and stops on a hotkey press and the recognized text is (auto)pasted into the window currently in focus, including Joplin.
LLM Interaction (Q&A, chat, etc.)
BlahST can also interact with local llama.cpp models, which is advantageous since many LLMs output their generated text directly into markdown. It also has a mode for replacing a selected text with a summary.
Forgot to mention language translation and input. There are a few demo videos with examples of multi-language LLM interaction and text input in the repository. One of the "roles" (prompts) for the local BlahST chatbot blahstbot is english speech to simplified-chinese text input into any window.
Thanks for Joplin. A great application. For me, it's especially important that I can easily feed an AI service like ChatGPT with parts of my Joplin notes archive, or conversely, that I can grant an AI service access to (parts of) my notes archive in Joplin. That way, I can use AI to obtain solid, relevant, and reliable knowledge.
@Coffee That would certainly be useful, but even with the absence of such a feature in the forum, you can still use AI tools externally to help you get to a solution.
I used the free ChatGPT AI Chatbot which is built into Firefox to have a conversation about your issue, and managed to get to the same answer as Laurent proposed (contact your Nextcloud provider) with a logical series of prompts
Prompt 1: For Joplin, how do I solve error "PUT info.json: “/…/info.json” is locked, existing lock on file"?
Prompt 2 (in response to a question): Nextcloud
Prompt 3: How I fix this if I am using a managed instance of Nextcloud?
Thank you for pointing me to external sources of information such as ChatGPT. However, I am looking for ‘priority support’ from Joplin Cloud Teams, for which I pay. I also use Joplin because I want to use a purely European service that is more compliant with European data protection law than ChatGPT, for example.
I'm not by any means saying AI is a suitable replacement for support. But in terms of using an AI tool for initial recommendations I'm just noting that even tools without full context can offer some good suggestions. If you're conscious about sharing confidential data with an external tool, you don't really need to share anything sensitive for straitforward queries like this, which is demonstrated by my minimal prompts
The priority for me with AI is transparency about whether/how the data exposed to AI is being used by others, and the option to easily avoid any exposure WITH CONFIDENCE. Privacy and security are important for me (and many users). I use Joplin in a way that ensures my data is not accessible by otherS, including at-rest local encryption. The introduction of AI in other software typically exposes user data to corporate capture. I would hate to see the security and privacy of Joplin diluted/destroyed by AI "innovations".
In that way, there's really nothing special about AI. For me this is just a tool like any other and it's sensible to see what we can do with it and how we can improve the user experience. If we can't that's fine too, we aren't going to force anything on the app for no reason.
With Joplin's focus on privacy and self-hosting, I'd imagine there would be interest in Joplin running llm's locally. With that in mind, most users wont's have a beefy computer to run giant 'do anything' llm's. So the focus would be on llms that are lighter on resources.
What would I like them to do in Joplin?
chat with notes / RAG. A hell of a lot can be done just with this
learn how I tag notes and auto-tag based on content (I think having a simple ux to confirm or remove an auto-tag would help train the AI and make it clear to users what is being done by the AI and what they did). The same for allocating notes into Notebooks. Optional extra: suggesting new tags or notebooks to the user
Of course there is lots more that could be done but these two are my top picks
