383 results found

1. Floating widget

   Add the option to enable a floating Lumo widget that can help me writing better, answer chats, or summarize long texts.

   8 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

2. a switch to turn off adult content

   Recently I have encountered an unexpected adult content (sex-related) when I was asking Lumo. It was not my intention at all. The answer should have been totally unrelated to sex. And I want to avoid seeing such contents again.

   Should there be such a switch?

   5 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   1 comment  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

3. Transfer/Share chat between users

   It would be nice to be able to share chats between proton users. This may be restricted to users belonging to the same duo/family/pro/... account, in which case it may be interesting to allow the transfer of chats from one user to another.

   This would enable collaborative (but private to selected users) chats.

   6 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

4. Theme

   I understand that this isn’t a priority right now, but in the long term it would be great to let Lumo have the same themes that are offered in the other Proton products (Mail, Drive..). (Lumo with Ebony would be beautiful 😍​)

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   1 comment  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

5. Setting for switching between 'enter to ask' and clicking to ask.

   Japanese (and perhaps other languages) requires the enter button to convert text into characters, which ends up asking Lumo before the sentence is fully written. This is frustrating as theres no other way to ask Lumo without typing the sentence somewhere else first, then copying. A feature to turn off 'enter to ask' would be appreciated.

   3 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

6. Answers on Top instead of Bottom

   I would like to suggest, for answers to appear on top of a thread instead on the bottom. Instead of scrolling through the history from top to bottom, new answers appear on top. Should I want to reread the hlstory I can do so top to bottom Especially for longer threads that you would like to continue it would be more helpful.

   4 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   1 comment  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

7. Add Voxtral as an LLM in Lumo

   I’m an enthusiastic user of Lumo and appreciate the privacy‑first approach you’ve built into the assistant. I noticed that Lumo already incorporates several open‑source large language models, including Mistral, which demonstrates your willingness to adopt cutting‑edge, community‑driven models.

   Given the impressive performance and open‑source nature of Voxtral, I’d like to kindly request that you consider adding it to Lumo’s model roster. Integrating Voxtral would further diversify the options available to users, potentially improving multilingual capabilities and offering additional flexibility for specialized workloads—all while staying aligned with Lumo’s privacy and security standards.

   Thank you for continuously improving Lumo, and I look forward to seeing Voxtral (or other emerging models) become part of the platform.

   I’m an enthusiastic user of Lumo and appreciate the privacy‑first approach you’ve built into the assistant. I noticed that Lumo already incorporates several open‑source large language models, including Mistral, which demonstrates your willingness to adopt cutting‑edge, community‑driven models.

   Given the impressive performance and open‑source nature of Voxtral, I’d like to kindly request that you consider adding it to Lumo’s model roster. Integrating Voxtral would further diversify the options available to users, potentially improving multilingual capabilities and offering additional flexibility for specialized workloads—all while staying aligned with Lumo’s privacy and security standards.

   Thank you for continuously improving Lumo, and I look… more

   2 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

8. Sensitive‑Data Confirmation for Feedback

   Problem

   When users give a thumbs‑up/down on a response, the associated snippet (including any code or text shown) is sent for anonymous analysis. If the snippet unintentionally contains sensitive information (IP addresses, ports, encryption keys, hashes, credentials, etc.), that data could be transmitted without the user’s explicit consent.

   Proposed Solution

   Add a post‑feedback check that scans the rated snippet for patterns that look like sensitive data. If a match is found, display a prompt:

   “The content you just rated appears to contain potentially sensitive information (e.g., IP address, key, hash). Do you want to include this data in the feedback you’re sending?”
   

   Provide three options:

   Send anyway – proceed with the normal feedback submission.
   Redact & send – automatically replace detected values with placeholders (e.g., <IP>, <KEY>).
   Cancel – abort the feedback submission so the user can edit or remove the sensitive data.
   

   Benefits

   Reduces risk of accidental leakage of confidential information.
   Gives users explicit control over what data leaves the session.
   Improves overall trust in Lumo’s privacy guarantees, especially for developers and security‑focused users who often share code snippets.
   Aligns with Lumo’s “privacy‑first” positioning without hindering valuable feedback.
   

   Implementation Notes

   Use lightweight regex patterns for common sensitive data types (IPv4/IPv6, ports, hex strings of typical key lengths, SHA hashes, JWTs, etc.).
   Allow the detection sensitivity to be toggled in settings for power users.
   Log only the user’s decision (send/redact/cancel), not the raw snippet, unless the user opts to send it.
   

   Problem

   When users give a thumbs‑up/down on a response, the associated snippet (including any code or text shown) is sent for anonymous analysis. If the snippet unintentionally contains sensitive information (IP addresses, ports, encryption keys, hashes, credentials, etc.), that data could be transmitted without the user’s explicit consent.

   Proposed Solution

   Add a post‑feedback check that scans the rated snippet for patterns that look like sensitive data. If a match is found, display a prompt:

   “The content you just rated appears to contain potentially sensitive information (e.g., IP address, key, hash). Do you want to include this data in the feedback

   … more
   4 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

9. Add Concise/Think Longer option for all answers

   Add an option to choose to always be concise or think longer for all answers.

   5 votes

   Vote Vote Vote Vote

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Vote

   We’ll send you updates on this idea

   0 comments  ·  New feature  ·  Delete…  ·  Admin →
   How important is this to you?

   Not at all You must login first! Important You must login first! Critical You must login first!

   We're glad you're here

   Please sign in to leave feedback

   Signed in as (Sign out)

   Close

   Close

   Submit Rating

10. lumo.ai for quick access > lumo.proton.me

    Also: https://lumo.ai/search?q=%s site search support

    5 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

11. New chat shortcut

    It would be nice to have a new chat keyboard shortcut to avoid potential hallucinations without reaching for the mouse

    5 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    1 comment  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

12. Let Lumo reference past chats & have memory

    Hello Proton Team,
    I would like to suggest adding a new feature to Lumo that allows it to collect personal information and interact with users in a manner similar to ChatGPT, but with a strong focus on privacy. The idea is to enable Lumo to understand and assist users by learning from their information, preferences, and interactions — all while ensuring that this data is fully private and protected via end-to-end encryption.
    This would combine the convenience and intelligence of an AI assistant with Proton’s commitment to user privacy and security. It could open up new possibilities for personal productivity, smart assistance, and secure communication without compromising sensitive data.
    Thank you for considering this innovative enhancement for Lumo.

    Hello Proton Team,
    I would like to suggest adding a new feature to Lumo that allows it to collect personal information and interact with users in a manner similar to ChatGPT, but with a strong focus on privacy. The idea is to enable Lumo to understand and assist users by learning from their information, preferences, and interactions — all while ensuring that this data is fully private and protected via end-to-end encryption.
    This would combine the convenience and intelligence of an AI assistant with Proton’s commitment to user privacy and security. It could open up new possibilities for personal productivity,… more

    5 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

13. Hierarchical Multi-Task Constraint Optimization Engine

    I am writing to propose a mission-critical feature to address severe hallucination and performance degradation when processing complex, multi-project prompts with nested constraints.

    Problem Statement

    Our system currently exhibits fundamental breakdowns when handling prompts containing multiple distinct projects with interdependent constraints and domain-specific requirements:

    1. Critical Performance Drop: Benchmarks show accuracy falling below 10% compared to single-task prompts. When processing 32+ nested hierarchical tasks (e.g., "Project Z's Project B's Project D"), performance degrades to GPT-3 equivalence levels.

    2. Constraint Isolation Failure: The system cannot reconcile project-specific constraints with global constraints simultaneously. Complex nesting ("Project A's Project B's Project 9") causes complete context loss from parent-level backgrounds and requirements.

    3. Format Contamination: Long-format examples (e.g., Q Project's 10,000-character engineering specs) incorrectly propagate to unrelated projects, overriding their designated output formats.

    4. Cross-Domain Interference: Performance degrades exponentially when tasks span disparate domains (mathematics, image generation, genetics, history). Similar projects (G/J) get merged despite explicit separation instructions.

    5. Execution Planning Deficit: No mechanism exists for task weighting, dependency mapping, or sequential vs. parallel execution determination across multiple projects.

    Specific Failure Scenarios

    • Multi-tier References: When Project Z's Project B's Project D requires referencing constraints from Project B's Project A background, the system fails to trace and apply hierarchical dependencies.
    • Background-Only Tasks: Projects like P (life philosophy) that provide only contextual information contaminate active task execution.
    • Unprecedented Generation: Z Project's requirement to generate entirely novel combinations (e.g., merging A's B's 9) triggers hallucination cascades.

    Proposed Feature: Hierarchical Multi-Task Constraint Optimization Engine

    This architectural enhancement would deliver:

    1. Constraint Inheritance & Isolation: Maintain separate constraint namespaces per task/node with explicit inheritance rules from parent levels
    2. Format Locking: Enforce project-specific output schemas without cross-contamination, regardless of example length
    3. Dynamic Weighting & Dependency Graph: Allow user-defined priorities, execution sequences, and dependency mapping across task trees
    4. Domain Siloing: Prevent knowledge transfer between disparate fields (e.g., D Project's history exams vs. Q Project's engineering specs)
    5. Reference Resolution Engine: Intelligent handling of inter-project citations without context merging
    6. Performance Scaling: Maintain GPT-4-level quality for up to 50+ hierarchical tasks

    High-Value Application Domains

    • Mathematical/Genetic Optimization: Simultaneous constraint satisfaction across multiple algorithmic problems
    • Technical Documentation: Assembling complex specs from modular components while preserving format integrity
    • Academic Assessment: Processing multi-domain exam hierarchies (e.g., Project D's fill-in-the-blank history tests) without domain bleeding
    • Creative Synthesis: Executing Z Project's "unprecedented generation" requirements through controlled combinatorial logic

    Business Impact

    • Restore enterprise-grade reliability for complex workflows
    • Enable new revenue streams in technical fields (biotech, engineering, finance)
    • Reduce API costs from hallucination-induced retry loops
    • Establish competitive moat in compound AI task handling

    I have detailed benchmark data and failure pattern analysis available for review. I would appreciate the opportunity to discuss this proposal in our next roadmap planning session.

    Best regards,

    I am writing to propose a mission-critical feature to address severe hallucination and performance degradation when processing complex, multi-project prompts with nested constraints.

    Problem Statement

    Our system currently exhibits fundamental breakdowns when handling prompts containing multiple distinct projects with interdependent constraints and domain-specific requirements:

    1. Critical Performance Drop: Benchmarks show accuracy falling below 10% compared to single-task prompts. When processing 32+ nested hierarchical tasks (e.g., "Project Z's Project B's Project D"), performance degrades to GPT-3 equivalence levels.

    2. Constraint Isolation Failure: The system cannot reconcile project-specific constraints with global constraints simultaneously. Complex nesting ("Project A's Project B's Project 9") causes complete context… more

    2 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    1 comment  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

14. Integrate into proton family

    Make the plus plan be available to proton family subscribers

    5 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

15. SynthID or similar feature

    Hi !
    With the aim of having an ethical minimum for the internet and in view of the announcement of an image generation option soon added to Lumi, I think it would be good to have something like 'SynthID' implemented in Lumi to be able to detect that the generated images are indeed generated by AI, and thus make possible malicious uses of AI more difficult
    Maybe its aleready in lumni but i didn't find information about it, sorry if it the case.

    2 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

16. Customize with URL parameters

    Can Lumo recognize URL parameters?
    It's convenient for Firefox user if Lumo can adhere to OpenWebUI convention.

    https://docs.openwebui.com/features/chat-features/url-params/
    https://docs.openwebui.com/tutorials/integrations/firefox-sidebar/#url-parameters-for-open-webui

    Related to this request:

    https://protonmail.uservoice.com/forums/932842-lumo/suggestions/50256153-add-lumo-to-firefox-ai-chatbots

    It solves most of the following ideas that request permanent customization.

    https://protonmail.uservoice.com/forums/932842-lumo/suggestions/50449944-ghost-mode-and-or-web-search-saved
    https://protonmail.uservoice.com/forums/932842-lumo/suggestions/50680241-enable-https-lumo-proton-me-search-q-s-as-a-dir
    https://protonmail.uservoice.com/forums/932842-lumo/suggestions/50413809-public-url-parameter-that-pre-loads-a-specific-pro

    3 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

17. Integrate with Windsurf

    Have Lumo be an available Model via Windsurf like Claude, ChatGPT, Gemini, etc.
    I've found Lumo to be very good with coding example and questions. Having Lumo be able to directly see and edit the code files would be a huge win.

    6 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

18. Lumo system prompt

    When I ask Lumo for help, I have to always give instructions how it should respond, otherwise it may generate an undesired huge response for rather simple questions. Having an ability to predefine a system prompt to influence this behaviour is in my opinion a required feature. This feature could be also expanded to have several system prompts to choose from at the beginning of a conversation.

    4 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

19. Personal Assistants / Preconfigured Prompts in Lumo

    I’d like Lumo to support personal assistants or preconfigured prompts, similar to what Kagi Assistant and other AI tools already offer.

    This would make it easier to switch between different roles (e.g., researcher, writer, teacher, developer) without having to retype or remember long instructions each time. It would also give users more control and flexibility to adapt Lumo to their own workflows.

    4 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating

20. The ability to set a permanent but editable preamble that is always considered by Lumo+ at the start and during each chat.

    I have recently moved from ChatGBT over to Lumo+ and I noticed some differences. Lumo+ under it's current configuration isn't nearly as accurate as Chat. Yes, Chat can get you up the wrong tree at times, but not as much as Lumo. One thing that I think that helped with Chat is the the ability to create a general header for how I wanted Chat to operate which included instructions about accuracy, depth, reply style, and whether to resort to the web to get latest information about softwre updates on things like Home Assistant and Raspberry Pi. With Chat I could have a standing request to go deeper and consider more references before making an answer giving links to where it got it's information.
    I made the mistake of asking Lumo how to send feedback and it got it all wrong. I really don't want an AI that just makes things up.

    I have recently moved from ChatGBT over to Lumo+ and I noticed some differences. Lumo+ under it's current configuration isn't nearly as accurate as Chat. Yes, Chat can get you up the wrong tree at times, but not as much as Lumo. One thing that I think that helped with Chat is the the ability to create a general header for how I wanted Chat to operate which included instructions about accuracy, depth, reply style, and whether to resort to the web to get latest information about softwre updates on things like Home Assistant and Raspberry Pi. With Chat I… more

    3 votes

    Vote Vote Vote Vote

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Vote

    We’ll send you updates on this idea

    0 comments  ·  New feature  ·  Delete…  ·  Admin →
    How important is this to you?

    Not at all You must login first! Important You must login first! Critical You must login first!

    We're glad you're here

    Please sign in to leave feedback

    Signed in as (Sign out)

    Close

    Close

    Submit Rating