dsh-auto-memory 从一个人的实验,长成了有人在用的工具。这页记录所有提过 issue、交过 PR、以及赞助过开发资源的人——名字和贡献都在这里。
dsh-auto-memory grew from a solo experiment into something people actually use. This page credits everyone who filed an issue, sent a pull request, or sponsored the development resources.
按贡献条目数排序。每条都指向原始的 issue / PR。
Sorted by number of contributions. Every entry links back to the original issue or PR.
负责新版前端(开发版 UI)的完整开发:参考图风格记忆工作台 —— 九页导航、列表与详情、四组设置、专注查看、深色适配、窄屏导航;以生成器方式接入(新皮肤版本从经典组件派生,上游结构漂移即明确报错),经典版与异常回退原样保留。此外贡献量最大,覆盖面也最广:语义索引同步卡死自愈、记忆搬包与安全导入;自动沉淀 100% 失效、水位虚高与接续时机、跨工作区与 per-runtime 隔离等稳定性根因;python 模型下载链路重写与坏产物清除;打包卫生、webhook 默认 fail-closed、CI 全量门禁恢复;并补齐数十条回归守卫。
Responsible for the complete development of the new frontend (the dev-preview UI): a reference-style memory workbench — nine-page navigation, list & detail views, four settings groups, focus view, dark-mode adaptation and narrow-screen navigation — integrated through a generator (new skin variants are derived from the classic components; structural drift fails loudly), with the classic skin and error fallbacks preserved unchanged. Also the most prolific contributor, and the broadest in scope: self-healing for a permanently stuck semantic-index sync, memory packing and safe import; root causes behind auto-consolidation failing 100% of the time, inflated context-watermark readings, continuation timing, and cross-workspace / per-runtime isolation; a rewritten python model-download chain with bad-artifact cleanup; packaging hygiene, webhook defaults switched to fail-closed, and restoration of the full CI gate — plus dozens of new regression guards.
持续的问题反馈者,覆盖面板行为、召回质量与配置边界等场景。
A persistent reporter covering panel behaviour, recall quality, and configuration edge cases.
关于记忆召回与注入行为的连续反馈。
Successive reports on memory recall and injection behaviour.
DSH Desktop 增强模式(透明 / Mica 材质)下的面板可读性修复,以及入口按钮防遮挡与外点 / Esc 关闭。
Panel readability fix for DSH Desktop enhanced mode (transparent / Mica materials), plus entry-button anti-occlusion and outside-click / Esc dismissal.
无人值守 / 批处理场景的加固反馈——直接推动了欢迎向导与逐功能开关化。
Unattended / batch-workflow hardening feedback that directly drove the welcome tour and per-feature switches.
早期问题反馈,帮助校准了最初的行为预期。
Early feedback that helped calibrate the initial behaviour expectations.
提出日文界面本地化([PR #143])。该 PR 虽被关闭,但内容已并入主线:在其基础上扩展为字典查表式 i18n 层 —— 界面文案改用 L(中文, English) 两参调用、第三语言按中文原串查词典,覆盖 602 条内联文案 + 583 条词典条目。此后新增语言只需加一份词典,无需改动任何调用点。
Proposed Japanese UI localisation ([PR #143]). The PR was closed, but its content landed in main and was extended into a dictionary-driven i18n layer — UI strings use L(Chinese, English) with the third language resolved from a Chinese-keyed dictionary, covering 602 inline strings + 583 dictionary keys. Adding a language now means adding one dictionary, with no call-site changes.
每一位的第一个反馈都算数——很多修复就是从这些 issue 开始的。
Every first report counts — many of these fixes started right here.
@Architectxz#1
@ALuoXue#2
@lhbsaa#3
@alexchenzl#6
@zjj871114037#7
@eclgo#13
@moonltppt#14
@swtseaman#21
@xiaochaZ#38
@jeffsui#39
@cuohua#40
@ADE150#102
本项目不接广告、不做付费墙。以下赞助方提供的是开发本身要烧掉的东西——API 额度。
No ads, no paywall. These sponsors provide what development actually burns: API quota.
API 中转站,为本项目的开发、回归测试,以及 M 系列语义引擎的选型与激活策略研究提供模型端点。没有这些额度,那些基准实验根本跑不起来。
An API relay station providing the model endpoints used for development, regression testing, and the M-series semantic-engine selection / activation research. Without that quota, those benchmark runs simply would not have happened.
不需要先成为"核心贡献者"。一个能复现的 issue 就是有效贡献。
You do not need to be a "core contributor" first. A reproducible issue is a real contribution.
带上 DSH 版本、插件版本、复现步骤和实际 vs 预期。这一步比什么都重要。
Include DSH version, plugin version, steps, and actual vs expected. This matters more than anything else.
小步、单一目的、带上验证结果。回归套件是绿的才算完成。
Small, single-purpose, with verification results. It counts as done only when the regression suite is green.
问题反馈、使用技巧交流,响应通常比 issue 更快。
Feedback and usage tips — usually faster than issues.
提供 API 额度或其他开发资源。开个 issue 说明来意即可。
Provide API quota or other development resources. Just open an issue.