Feature Request: Persistent Memory System for Users/功能建议 添加“用户记忆”功能以提升个性化体验

Description
I suggest adding a persistent user memory system to DeepSeek, similar to the way ChatGPT implements user memory. This would allow DeepSeek to remember important information users want it to keep, such as:

User preferences (e.g. preferred language, coding style, model behaviors)

Project context or ongoing tasks

Specific knowledge, usernames, terms, or shorthand frequently referenced

Features

Users can explicitly tell DeepSeek what to remember (e.g., “remember I use VS Code and prefer Python 3.11”)

Users can view, edit, or delete memory items at any time

Memory would improve long-term productivity, especially for developers who return frequently

Benefits

Stronger personalization for individual users

Better long-term assistance across sessions

More competitive experience compared to models with memory (e.g., GPT-4 Turbo, Claude)

Looking forward to seeing this amazing feature added. Thank you for the great work you're doing with DeepSeek!
建议内容：
我希望 DeepSeek 可以加入类似 ChatGPT 的 “用户记忆系统”，它可以：

让用户主动告诉它记住一些内容（例如“我偏好 Python 3.11”或“我每次写代码都用 VS Code”）

可以在设置中随时查看、编辑、删除这些记忆内容

实现用户多次使用时的上下文连贯和偏好一致性，提高交互效率

为什么重要：

这将极大提升 DeepSeek 的个性化能力

对于日常使用的开发者来说能提升长期效率

对比 GPT 系列等已有“记忆”模型，这一功能会提升 DeepSeek 的用户粘性和满意度

非常期待你们能考虑这个功能，再次感谢你们团队的努力和贡献！

The issue names no files, tests, or entry points. Start by reviewing the repository's existing user, session, and conversation-context handling, then define how memory creation, viewing, editing, deletion, storage, and retrieval should work. Done requires an agreed design and an end-to-end implementation.

Written by the indexing model from the issue text.