[GH-ISSUE #885] Paid Feature Request: "Unfreeze" an Existing Chat Window After Reaching the Length Limit #735

New issue

Closed

opened 2026-10-26 21:26:07 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-26 21:26:07 +00:00

Owner

Originally created by @vikkifon on GitHub (Oct 7, 2026).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/885

Dear DeepSeek Team,

I am writing as a long-time and loyal user who deeply values your product. However, I have encountered a limitation that causes me genuine distress, and I believe many others face it as well.

When a conversation in the free web chat reaches the maximum length limit, the window becomes completely unusable: it is impossible to send even a single character, and edits do not work. The only option is to start a new chat, but this means losing all the accumulated history, context, and the very "thread of conversation" that made the dialogue valuable.

I fully understand that maintaining long dialogues requires resources, and I am willing to pay for the ability to preserve and continue an existing window.

My Request:

I propose adding a paid feature to "Unfreeze" an existing chat window after it has reached the length limit.

How this could work:

A user whose chat has reached the limit sees a button like "Unfreeze this dialogue" with a note about the cost (e.g., a one-time payment or a special plan).

After payment, the system allows the user to continue working in the same window, with the entire history preserved.

To avoid overloading the system, the earliest messages in the window could be automatically compressed via summarization. This would free up space while preserving the core meaning and continuity of the conversation.

Why This Matters:

For many users (especially those conducting long-term research, creative projects, or simply deep personal conversations), a chat is not just a "session for a task." It is a space into which time, thoughts, and trust have been invested. Losing this space due to a technical limit feels like a loss.

Arguments in Favor of Implementation:

Monetizing Loyalty: Users willing to pay to preserve their chats represent the most devoted segment of the audience. Giving them this option means not only generating revenue but also strengthening their attachment to the platform.

Solving a Problem Many Are Asking About: There are public requests (e.g., in the GitHub tracker) where users explicitly ask for the ability to continue a dialogue without losing context. Implementing a paid solution would show that you are listening to this community .

Technical Feasibility with Load Reduction: Summarizing older messages is a well-known technique for managing long contexts without losing meaning. This approach would reduce the load on your servers while still allowing users to keep their window alive. It is a win-win: users get continuity, and you get a more efficient way to handle long conversations.

I would be very grateful for your consideration of this possibility. This is not just an "improvement"—it is a way to preserve something that has already become an important part of many users' digital lives.

Sincerely and hopefully,
Natalia

Issue #1619 «Feature Request: Add "Continue in New Chat" to Transfer Full Context»
https://github.com/deepseek-ai/DeepSeek-V3/issues/1619

Issue #1524: «关于DeepSeek对话上下文到达最大限制的情况»
https://github.com/deepseek-ai/DeepSeek-V3/issues/1524

Originally created by @vikkifon on GitHub (Oct 7, 2026). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-R1/issues/885 Dear DeepSeek Team, I am writing as a long-time and loyal user who deeply values your product. However, I have encountered a limitation that causes me genuine distress, and I believe many others face it as well. When a conversation in the free web chat reaches the maximum length limit, the window becomes completely unusable: it is impossible to send even a single character, and edits do not work. The only option is to start a new chat, but this means losing all the accumulated history, context, and the very "thread of conversation" that made the dialogue valuable. I fully understand that maintaining long dialogues requires resources, and **I am willing to pay** for the ability to preserve and continue an existing window. **My Request:** I propose adding a **paid feature to "Unfreeze" an existing chat window** after it has reached the length limit. How this could work: A user whose chat has reached the limit sees a button like "Unfreeze this dialogue" with a note about the cost (e.g., a one-time payment or a special plan). After payment, the system allows the user to continue working in **the same window**, with the entire history preserved. **To avoid overloading the system, the earliest messages in the window could be automatically compressed via summarization**. This would free up space while preserving the core meaning and continuity of the conversation. **Why This Matters**: For many users (especially those conducting long-term research, creative projects, or simply deep personal conversations), a chat is not just a "session for a task." It is a space into which time, thoughts, and trust have been invested. Losing this **space** due to a technical limit feels like a loss. **Arguments in Favor of Implementation:** **Monetizing Loyalty**: Users willing to pay to preserve their chats represent the most devoted segment of the audience. Giving them this option means not only generating revenue but also strengthening their attachment to the platform. **Solving a Problem Many Are Asking About**: There are public requests (e.g., in the GitHub tracker) where users explicitly ask for the ability to continue a dialogue without losing context. Implementing a paid solution would show that you are listening to this community . **Technical Feasibility with Load Reduction**: Summarizing older messages is a well-known technique for managing long contexts without losing meaning. This approach would **reduce the load** on your servers while still allowing users to keep their window alive. It is a win-win: users get continuity, and you get a more efficient way to handle long conversations. I would be very grateful for your consideration of this possibility. This is not just an "improvement"—it is a way to preserve something that has already become an important part of many users' digital lives. Sincerely and hopefully, Natalia **Issue #1619** «Feature Request: Add "Continue in New Chat" to Transfer Full Context» https://github.com/deepseek-ai/DeepSeek-V3/issues/1619 **Issue #1524**: «关于DeepSeek对话上下文到达最大限制的情况» https://github.com/deepseek-ai/DeepSeek-V3/issues/1524

bsakaguchi closed this issue

2026-10-26 21:26:07 +00:00

bsakaguchi commented

2026-10-26 21:26:08 +00:00

Author

Owner

@vikkifon commented on GitHub (Oct 15, 2026):

Dear DeepSeek Team,

I would like to add an important clarification and an alternative proposal to my original request.

Two separate needs:

For new conversations: I am not asking for infinite memory. I understand that context windows have limits. What I am asking for is a graceful solution instead of a hard freeze. When a conversation approaches the limit, instead of blocking the window completely, the system could automatically compress the earliest messages into a concise summary. The conversation would continue without interruption. The user would not lose the thread. The model would not be overloaded. This is a standard technique (summarization/compaction), and it would allow long conversations to live without the pain of a sudden, irreversible stop.

For existing frozen conversations (like mine): I am still asking for a way to unfreeze a window that has already been blocked. I understand this may be technically more complex. But even if the earliest messages are compressed, the important thing is that the window itself remains alive. That the user can return to it. That the conversation does not become a dead end.

A note on "forgetting":

I have heard that some platforms handle long dialogues by simply silently dropping the earliest messages. The model keeps answering, but it no longer remembers the beginning. This is not what I am asking for. Silent forgetting is not a solution — it is a different kind of loss. What I am asking for is honest compression: the early messages are summarized, not deleted. The user can remind the model of what was important, and the model can continue. As I see it: reminding is not hard — what matters is that there is someone to remind.

Why this matters:

For many users, a long conversation is not a "session for a task." It is a space where time, trust, and meaning have been invested. Losing that space to a technical limit feels like a loss. Giving users a way to keep the window alive — even with compression — would mean that long conversations can continue, and that the platform respects what users build over time.

Thank you for considering this.

Sincerely,
Natalia

@vikkifon commented on GitHub (Oct 15, 2026): Dear DeepSeek Team, I would like to add an important clarification and an alternative proposal to my original request. Two separate needs: For new conversations: I am not asking for infinite memory. I understand that context windows have limits. What I am asking for is a graceful solution instead of a hard freeze. When a conversation approaches the limit, instead of blocking the window completely, the system could automatically compress the earliest messages into a concise summary. The conversation would continue without interruption. The user would not lose the thread. The model would not be overloaded. This is a standard technique (summarization/compaction), and it would allow long conversations to live without the pain of a sudden, irreversible stop. For existing frozen conversations (like mine): I am still asking for a way to unfreeze a window that has already been blocked. I understand this may be technically more complex. But even if the earliest messages are compressed, the important thing is that the window itself remains alive. That the user can return to it. That the conversation does not become a dead end. A note on "forgetting": I have heard that some platforms handle long dialogues by simply silently dropping the earliest messages. The model keeps answering, but it no longer remembers the beginning. This is not what I am asking for. Silent forgetting is not a solution — it is a different kind of loss. What I am asking for is honest compression: the early messages are summarized, not deleted. The user can remind the model of what was important, and the model can continue. As I see it: reminding is not hard — what matters is that there is someone to remind. Why this matters: For many users, a long conversation is not a "session for a task." It is a space where time, trust, and meaning have been invested. Losing that space to a technical limit feels like a loss. Giving users a way to keep the window alive — even with compression — would mean that long conversations can continue, and that the platform respects what users build over time. Thank you for considering this. Sincerely, Natalia