I want to make it clear that Lumo is no where close to ChatGPT, Lumo is special in how it has a "ghost mode" button which deletes chats the second you close them, as I have been told by Proton.

I think that this is like a situation with the tor browser for example, the absolute lack of any personalization shields you from threats, what your proposing would be like making tor collect data for search personalization. Its like replacing privacy for that ounce of convenience so what, it takes 5 minutes to explain context to Lumo, if you want a convenient AI then ChatGPT already exists, we don't need another ChatGPT.

I’m not clear on why the existing ‘Anything else Lumo should know about you?’ personalization is insufficient to meet this requirement. I wouldn’t want profiling that isn’t transparent, explicit and on an opt‑in basis. Perhaps Lumo could allow adding to the personalization from within the chat, but I would expect to have to give explicit consent each time a save is proposed.

This doesn't seem possible without Lumo breaking some of its current tenets. Storing encrypted chats after the fact is one thing. Lumo doesn't really have access outside your profile settings and what you say in a single session. The value in a memory is clear and obvious, no argument there. But I don't see how it can be possible without opening up access to customer data on the server side of things. That's always been kept to the minimum amount possible and I applaud them for that. I don't like the idea of "just this one time though it'll be ok."

While I appreciate the enthusiasm for improving Lumo, I personally don’t want a memory or persistent-knowledge feature. One of the main reasons I trust Proton is because their tools don’t retain long-term personal data unless absolutely necessary. Even if the feature were opt-in, adding persistent memory feels like a step toward more data collection than I’m comfortable with.

I prefer Lumo to stay stateless and privacy-focused, without storing information about my preferences, past conversations, or behavior. If I need Lumo to know something, I’d rather tell it each time than have the system quietly accumulate personal details—even with controls in place.

For me, Proton’s strength is minimal data retention, strong boundaries, and tools that don’t try to “learn” about me. I hope Lumo keeps that philosophy intact.

This should be totally doable (and awesome) seeing as though Proton stores your chats anyway.

No guys it's not implemented, yet😑

Supported. I think with the release of Lumo 1.2 (is it really 1.2?) we got a bit closer to fulfillilng the request. At least now we can tell Lumo how we want to be called and other preferences. That is great!

I want to add something, though: It would be super nice to have some sort of profiles that Lumo can work with. So a personal one, one for work, one for ... That way, Lumo could be used in a more granular fashion. You know, so that in one profile it keeps certain important information that are related to a job or a project or something like that withouth "mixing" it up with inputs from other profiles.

To put it bluntly, all the features that OpenAI's ChatGPT has and adds (like memory, Agent/Research mode, Tasks, automatic thinking mode, etc.) are just super great and useful - and Lumo lacks them all. So it's hard to stick with Lumo, as much as I'd like to. I appreciate it's really hard to keep up with how fast OpenAI churns out features, but I'd hope that Proton can do a bit better than it currently does and up the pace a bit - it's been July since this was originally posted, and I believe ChatGPT has had Memory mode for longer already.
Thank you Proton team! :)

We urge the Lumo Team to work on it.

I still think it's critical!

More comments make it more visible

Guys, we should comment more on this suggestion to make it more visible to the Lumo team.

What Rutger van der Eijk said
(*^ー^)ノ♪

Being able to set some preferences would be very usefull to have.

There are certain things I need to repeat in conversations, like the locale I use in Excel. I can do this with
"Please format all Excel formulas using ; as the separator."
It would be usefull not to have to repeat this every time.

A full memory of all conversations (as is stated in this Feature request) is maybe difficult to implement. But a more lightweight Feature for having some preferences saved should be easier I expect.

I know that this is a very complicated issue, and it's one of the reasons I switched to lumo. I didn't know *which* things chatgpt was keeping in memory. The only thing I can think of is to have a transparency screen or tab where you can actually see what is in the saved memory and delete anything you don't want lumo to remember (or delete everything if you like). Alternatively, it could be made into an affirmative feature where you have to open the "memory" tab and tell it what you want it to remember. I'm not a developer, so these may be untenable ideas, but as a frequent user who likes to explore the bounds of what is possible with AI, that is how I would be the ideal situation in terms of ease of use and transparency.

Yeah, even just quality of life stuff like remembering I live in a country using metric would save a lot of faff

Specifically a system prompt box like LM Studio Settings->Context

This is critical, but there are very effective strategies for doing this now. They are also very flexible. Documenting those would be an excellent idea. "Memory" and "Knowledge" are extremely difficult and hideously expensive for any LLM, and privacy requirements raise difficulty and price exponentially.

At least Lumo should have an option to save text about you. This will prevent Lumo from giving advice for other usage scenarios. For example, operating system/platform, location, language, etc.