Hello OpenAI team,
I’d like to suggest a feature that would make ChatGPT much better at remembering important conversations, especially for users who work on long-term projects.
My idea is to let users manually select specific conversations and add them to a dedicated Conversation Memory.
How it could work
In ChatGPT settings, there could be a separate section called Conversation Memory, with an on/off switch and an Add Conversation option.
When the user selects Add Conversation, ChatGPT would show their conversation history so they can choose which conversation they want to preserve.
Before adding it, ChatGPT could display a confirmation such as:
“Do you want to add this conversation to Conversation Memory? The details of this conversation will be retained so ChatGPT can refer to and retrieve its information more accurately in future conversations.”
Once selected, the conversation would be linked to this memory feature. ChatGPT could then retrieve its relevant details when the user continues the same project or refers to that conversation.
The important point is that this should not rely only on a short automatically generated summary. The purpose is to preserve access to the actual conversation and its details, including decisions, explanations, instructions, and project history.
Temporary storage and user control
I would also suggest the following behavior:
- When Conversation Memory is enabled, selected conversations remain available to ChatGPT through this feature.
- When it is disabled, the stored information is retained for 30 days rather than being deleted immediately.
- If the user enables it again within those 30 days, the previously selected conversations and their associated information become available again in the same state.
- If it remains disabled beyond 30 days, the information in this temporary memory is deleted.
- Users should be able to see which conversations they have added and remove individual conversations whenever they want.
The original conversations should remain in the user’s normal chat history unless the user chooses to delete them separately.
Why this matters
Reference Chat History is useful, but it does not always preserve or retrieve every detail from previous conversations.
Saved Memories are also useful, but they are better suited to persistent personal information and preferences than to storing the detailed history of an entire project.
For people who use ChatGPT for ongoing projects, research, development, or other complex work, losing access to specific decisions and instructions can make it difficult to continue where they left off.
A user-selected Conversation Memory would give people more control over what ChatGPT should remember in detail, without requiring them to put every project detail into Saved Memories.
I believe this feature could make long-term conversations much more reliable and reduce the need to repeatedly create summaries and paste large amounts of previous context into new chats.
Thank you for considering this suggestion.