Jazzy
My feedback
56 results found
-
1 voteJazzy
shared this idea
·
-
76 votesJazzy
supported this idea
·
-
23 votesJazzy
supported this idea
·
-
351 votes
An error occurred while saving the comment Jazzy
supported this idea
·
-
190 votesJazzy
supported this idea
·
An error occurred while saving the comment Jazzy
commented
Extend Chat History for All – 7 Days is Too Short
The current 7-day chat history limit is simply too restrictive for practical use. Imagine being sick or away for two weeks; upon returning, you find everything gone—all your important notes, medical reminders, project ideas, and conversations have vanished. That is frustrating and defeats the purpose of having a personal assistant.
For paying customers with 500GB of encrypted storage (or Unlimited plans), there is no technical reason to limit history. "Unlimited" should mean unlimited retention as long as space permits.
My suggestion:
1. For Paid Users: Remove the limit entirely (unlimited history).
2. For Free Users: Extend the retention period (e.g., to 14 or 30 days) as a gesture of goodwill. This would help everyone manage their daily lives better without forcing an immediate upgrade, while still highlighting the added value of the Plus plan.
Please consider this adjustment. It would significantly improve the user experience for everyone and align Proton’s generosity with its privacy-first values.
-
1 voteJazzy
shared this idea
·
-
8 votesJazzy
supported this idea
·
-
3 votes
An error occurred while saving the comment Jazzy
commented
Great suggestion! Being able to jump between prompts would definitely solve the scrolling issue.
I'd love to add one complementary thought: A built-in search function (like Ctrl+F) specifically for keywords within a chat would be the perfect partner to this. Sometimes I don't remember the "topic block" to jump to, but I remember a specific word (e.g., "Schuh" or "birthday").
Imagine having both: A navigation list to jump between topics, AND a search bar to find specific words instantly. That would make managing long conversations truly effortless!
Thanks for starting this discussion!Jazzy
supported this idea
·
-
8 votesJazzy
supported this idea
·
-
4 votes
An error occurred while saving the comment Jazzy
commented
This is exactly what I need! Currently, when a chat hits its limit, I lose the context and have to start a new thread for the same topic. A feature that lets me 'pull' the context from an old chat into a new one (like an 'Amnesia Bridge') would solve the fragmentation problem perfectly. Please prioritize this!
Jazzy
supported this idea
·
-
25 votesJazzy
supported this idea
·
-
9 votesJazzy
supported this idea
·
-
9 votesJazzy
supported this idea
·
-
9 votesJazzy
supported this idea
·
-
41 votesJazzy
supported this idea
·
-
8 votesJazzy
supported this idea
·
-
557 votes
An error occurred while saving the comment Jazzy
commented
This is exactly what I need! Currently, even with the 'compaction' feature, there's a hard limit where the conversation context is lost, and I have to start a new chat for the same topic. This breaks the flow for long-term projects or deep analysis.
A true 'Memory' function that allows Lumo to remember key facts across sessions (encrypted, of course!) would be far more valuable than a simple price bundle. Please prioritize improving the continuity and context retention so we don't lose the thread of our conversationsJazzy
supported this idea
·
-
6 votesJazzy
supported this idea
·
-
6 votesJazzy
supported this idea
·
-
9 votesJazzy
supported this idea
·
Implement a 3-Position Web Search Switch for Speed vs. Accuracy Control
The current web search toggle is unreliable because the system tries to guess whether a search is needed to optimize for speed. This often results in outdated answers from the training data (pre-April 2024) even when the user expects current information, or unnecessary delays when quick answers are sufficient.
Users report that they often have to explicitly type "search the web" or add keywords like "latest" to force a search, which defeats the purpose of a simple toggle. The current behavior creates a "guessing game" where users never know if the answer is real-time or historical until it is too late.
I propose replacing the binary toggle with a 3-POSITION SWITCH to give users full control:
OPTION 1: OFFLINE / TRAINING ONLY (SPEED MODE)
Disables web search completely.
Uses only the model's internal knowledge (fastest response time).
Ideal for creative writing, coding, summarizing old texts, or when current events are irrelevant.
OPTION 2: AUTO (SMART MODE)
The AI decides if a search is needed based on the query.
This is the current intended behavior but needs to be fixed to be deterministic and reliable.
Ideal for general questions where the user is unsure if the info is current.
OPTION 3: FORCE WEB SEARCH (ACCURACY MODE)
Always performs a live web search, ignoring internal training data for facts.
Guarantees up-to-date information (slower response time due to search latency).
Essential for news, stock prices, weather, sports scores, and current events.
WHY THIS MATTERS: This solves the conflict between speed and accuracy. Users who need instant answers can choose "Offline," while those needing verified current data can choose "Force Search." It eliminates the frustration of guessing and makes the tool predictable. Currently, the inconsistency drives users to competitors like ChatGPT who handle this more transparently.
TECHNICAL FEASIBILITY: This is a simple UI and logic change. It involves passing a specific flag to the model router: "Ignore internal knowledge for facts" (Force) vs. "Use internal knowledge only" (Offline). No new backend infrastructure is required, just a clearer user interface and stricter logic for the "Auto" mode.