Repository navigation
Releases: pangyexing/vemry
Release list
vemry v1.0.3
vemry 1.0.3 lets you edit a table without typing its pipes, and smooths a few rough edges: quick capture on Windows, the cursor on a new task, and new items in the file tree.
What's new in 1.0.3
- Tables without typing pipes: with the cursor in a table, Tab and Shift-Tab step from cell to cell, and Tab past the last cell opens a new row. Enter opens an empty row below; Enter on an empty last row ends the table, the way it ends a list.
/tableinserts three rows by three columns, and a size can be typed onto it:/table4x5(or/4x5) is four rows by five columns, the header row counted among the rows. - Add bars and a Table menu: a rendered table shows two thin bars by the pointer, below it and to its right, that add a row or a column. Right-click a cell, or open the block handle's menu, for Table ›: insert a row above or below or a column left or right, align a column, delete a row or a column. The same operations are commands in the palette.
- Quick capture on Windows takes the keyboard as it opens: before, typing and Esc reached nothing until a click landed in the text. Esc now hides it from anywhere in the window.
- The cursor on a new task: after
- [ ]the cursor stands where the text will start, not against the checkbox. - New items in the file tree glide into place: after a new note, a new folder or a rename, the row slides from where you typed its name to where it sorts, instead of vanishing and reappearing elsewhere.
Downloads
| Platform | File |
|---|---|
| macOS, Apple silicon | vemry_1.0.3_aarch64.dmg |
| macOS, Intel | vemry_1.0.3_x64.dmg |
| Windows, x64 | vemry_1.0.3_x64-setup.exe |
| Linux, x86_64 | vemry_1.0.3_amd64.AppImage or vemry_1.0.3_amd64.deb |
- macOS 10.15 or later. The app is signed and notarized.
- Windows: the installer is not code-signed yet, so SmartScreen may show "Windows protected your PC" on first run. Choose More info → Run anyway.
- Updating from 1.0.2: Check for updates in the app finds 1.0.3; it installs only when you confirm.
Full changelog: https://vemry.app/changelog · Docs: SYNTAX · ARCHITECTURE
中文
vemry 1.0.3 让你不用手打竖线也能编辑表格,并打磨了几处细节:Windows 上的速记、新任务的光标、文件树里新建的条目。
1.0.3 的新内容
- 编辑表格不用再打竖线:光标在表格里时,Tab 和 Shift-Tab 在单元格之间移动,在最后一格按 Tab 会新开一行。回车在下方插入空行;在空的最后一行按回车结束表格,和结束列表一样。
/table插入三行三列,也可以直接带上尺寸:/table4x5(或/4x5)是四行五列,表头算在行数里。 - 添加条和「表格」菜单:渲染后的表格会在指针下方和右侧显示两条细条,点一下添加一行或一列。右键单元格,或打开块把手的菜单,选「表格 ›」:在上方或下方插入行、在左侧或右侧插入列、设置列对齐、删除行或列。命令面板里也有同样的命令。
- Windows 上速记窗口一打开就能输入:以前要先在文字区点一下,打字和 Esc 才有反应。现在在窗口里任何地方按 Esc 都能收起它。
- 新任务的光标:输入
- [ ]后,光标停在文字开始的位置,不再贴着复选框。 - 文件树里新建的条目会滑到位:新建笔记、新建文件夹或重命名之后,这一行从你输入名字的地方滑到它排序后的位置,不再一下子消失、又在别处出现。
下载:macOS Apple 芯片选 aarch64.dmg,Intel 选 x64.dmg;Windows 选 x64-setup.exe;Linux 选 .AppImage 或 .deb。
- macOS 需要 10.15 或更高版本,应用已签名并公证。
- Windows 安装包暂未代码签名,首次运行时 SmartScreen 可能提示"Windows 已保护你的电脑",点"更多信息"→"仍要运行"。
- 从 1.0.2 升级:在应用里点"检查更新"即可找到 1.0.3,你确认后才会安装。
vemry v1.0.2
vemry 1.0.2 sends your feedback to the developer from inside the app, and a first start offers Zhipu's current flagship.
What's new in 1.0.2
- Send feedback from the app: Help › Send Feedback…, Send feedback… in the palette, or Settings › General. Pick a problem, an idea or a question, write it, and it goes to the developer, signed in or not. Signed in, replies go to your account's address; signed out, leave an address, or none. The diagnostics it attaches (version, system, interface language, current model) are listed line by line before you send, and can be unticked. Nothing from your vault is attached; what the server keeps is in the privacy policy. A draft survives closing the dialog, and if the server cannot be reached the same text opens as an e-mail in your mail program instead.
- A first start offers GLM-5.3: Zhipu's current flagship, waiting for your key. The Zhipu preset lists glm-5.3-flash and glm-5.3. A settings file from 1.0.x keeps its model.
- The image warning names your model: attaching an image to a model that cannot read it names the current model and suggests deepseek-flash or glm-5.3-flash.
Downloads
| Platform | File |
|---|---|
| macOS, Apple silicon | vemry_1.0.2_aarch64.dmg |
| macOS, Intel | vemry_1.0.2_x64.dmg |
| Windows, x64 | vemry_1.0.2_x64-setup.exe |
| Linux, x86_64 | vemry_1.0.2_amd64.AppImage or vemry_1.0.2_amd64.deb |
- macOS 10.15 or later. The app is signed and notarized.
- Windows: the installer is not code-signed yet, so SmartScreen may show "Windows protected your PC" on first run. Choose More info → Run anyway.
- Updating from 1.0.1: Check for updates in the app finds 1.0.2; it installs only when you confirm.
Full changelog: https://vemry.app/changelog · Docs: SYNTAX · ARCHITECTURE
中文
vemry 1.0.2 让你在应用里就能把反馈发给开发者;首次启动改为推荐智谱当前的旗舰模型。
1.0.2 的新内容
- 在应用里发送反馈:「帮助 › 发送反馈…」、命令面板的「发送反馈…」,或者设置 › 通用。选问题、建议或提问,写下来,直接发给开发者,登录与否都可以。登录时,回复发到你的账户邮箱;不登录时,可以留一个地址,也可以不留。附带的诊断信息(版本、系统、界面语言、当前模型)会在发送前逐行列出,可以取消勾选。知识库里的任何内容都不会附带;服务器存什么写在隐私政策里。关掉对话框草稿还在;连不上服务器时,可以改用邮件程序发出同样的内容。
- 首次启动推荐 GLM-5.3:智谱当前的旗舰,等你填 key。智谱预设列出 glm-5.3-flash 和 glm-5.3。1.0.x 的设置文件保留原来的模型。
- 附图提示写明当前模型:给读不了图片的模型附图时,提示会写明当前模型,并建议换成 deepseek-flash 或 glm-5.3-flash。
下载:macOS Apple 芯片选 aarch64.dmg,Intel 选 x64.dmg;Windows 选 x64-setup.exe;Linux 选 .AppImage 或 .deb。
- macOS 需要 10.15 或更高版本,应用已签名并公证。
- Windows 安装包暂未代码签名,首次运行时 SmartScreen 可能提示"Windows 已保护你的电脑",点"更多信息"→"仍要运行"。
- 从 1.0.1 升级:在应用里点"检查更新"即可找到 1.0.2,你确认后才会安装。
vemry v1.0.1
vemry 1.0.1 brings your own choice of AI vendor: several providers side by side, each vendor's current models spoken to in its own API, and one picker to switch between them.
What's new in 1.0.1
- Several providers, one picker — Settings › AI now holds a list of providers, each an endpoint, a key and the models it serves; vemry hosted is one more provider in the same list. The current model is picked in Settings, in the chat's model menu or from the palette (Switch model…), and holds from the next message in both chat windows. Translation can use a model of its own or follow the chat's.
- More vendors — presets for OpenAI, Anthropic, xAI, Meta, Moonshot and Z.ai join Zhipu, DeepSeek and Ollama, each with its current models. OpenAI, xAI and Meta are spoken to in their Responses APIs and Anthropic in its Messages API, so tools and thinking work together on their newest models.
- Thinking depth on every current model — GPT-6, Claude 5, Grok 4.7 and 4.3, Muse Spark 1.3, Kimi K3, GLM-5.3 and DeepSeek each take the level in their own terms; a model that always thinks gets its lightest level for Off.
- Web search on more vendors — DeepSeek, Anthropic, OpenAI, xAI, Meta and Zhipu's GLM-5.3. Search stays off until you turn it on, and each search is billed by the vendor.
- Switch models mid-conversation — even across vendors, between any two messages. Each answer in the session log names the model that gave it, and Session Replay shows where the model changed.
- Steadier sessions — a conversation that outgrows a model's context compacts itself on every vendor, an overloaded server is retried, and a declined answer no longer leaves a fragment in the history.
ZHIPU_API_KEYfrom the environment now fills only a blank Zhipu key, so it is never sent to another host.
Downloads
| Platform | File |
|---|---|
| macOS, Apple silicon | vemry_1.0.1_aarch64.dmg |
| macOS, Intel | vemry_1.0.1_x64.dmg |
| Windows, x64 | vemry_1.0.1_x64-setup.exe |
| Linux, x86_64 | vemry_1.0.1_amd64.AppImage or vemry_1.0.1_amd64.deb |
- macOS 10.15 or later. The app is signed and notarized.
- Windows: the installer is not code-signed yet, so SmartScreen may show "Windows protected your PC" on first run — choose More info → Run anyway.
- Updating from 1.0.0: Check for updates in the app finds 1.0.1; it installs only when you confirm.
Full changelog: https://vemry.app/changelog · Docs: SYNTAX · ARCHITECTURE
中文
vemry 1.0.1 让你自己选 AI 厂商:多个提供方并列,每家的当前模型都走它自己的 API,一个选择器就能切换。
1.0.1 的新内容
- 多个提供方,一个选择器:设置 › AI 现在是一份提供方列表,每个提供方是一个端点、一个 key 和它提供的模型;vemry 托管也是列表里的一个提供方。当前模型可以在设置、对话的模型菜单或命令面板(「切换模型…」)里选,从下一条消息起在两个对话窗口里都生效。翻译可以用单独的模型,也可以跟随对话。
- 更多厂商:在智谱、DeepSeek、Ollama 之外,新增 OpenAI、Anthropic、xAI、Meta、月之暗面和 Z.ai 的预设,带上各家当前的模型。OpenAI、xAI、Meta 走它们的 Responses API,Anthropic 走 Messages API,最新模型上工具调用和思考可以同时进行。
- 每个当前模型都能调思考深度:GPT-6、Claude 5、Grok 4.7 与 4.3、Muse Spark 1.3、Kimi K3、GLM-5.3 和 DeepSeek 都按各自的档位接收;总是在思考的模型,「关」对应它最轻的一档。
- 更多厂商能联网搜索:DeepSeek、Anthropic、OpenAI、xAI、Meta 和智谱 GLM-5.3。搜索默认关闭,开了才用,每次搜索由厂商计费。
- 对话中途换模型:任意两条消息之间都能换,跨厂商也行。会话日志里每个回答都标明是哪个模型给的,会话回放会显示模型在哪里换了。
- 会话更稳:对话超出模型上下文时,各家都会自动压缩;服务器过载会自动重试;模型拒答时不再在历史里留下半截回答。
- 环境变量
ZHIPU_API_KEY现在只填补空着的智谱 key,绝不会发给别的主机。
下载:macOS Apple 芯片选 aarch64.dmg,Intel 选 x64.dmg;Windows 选 x64-setup.exe;Linux 选 .AppImage 或 .deb。
- macOS 需要 10.15 或更高版本,应用已签名并公证。
- Windows 安装包暂未代码签名,首次运行时 SmartScreen 可能提示"Windows 已保护你的电脑",点"更多信息"→"仍要运行"。
- 从 1.0.0 升级:在应用里点"检查更新"即可找到 1.0.1,你确认后才会安装。
vemry v1.0.0
The first release of vemry: a local-first Markdown workspace for study and research, with an AI agent that works on your notes behind a permission gate. Your notes are plain Markdown files in a folder you choose; everything else in the app is a projection of them.
What's in 1.0.0
- Writing — a live-preview Markdown editor with a
/insert menu, callouts, folding and inline fields; a command palette (⌘P / Ctrl+P); quick capture from any app (⌥⇧Space / Alt+Shift+Space) and from a paired phone. - Reading and study — import a paper from an arXiv link or a PDF, side-by-side translation of a note, a web page or a PDF, flashcards with a Review view, and a built-in browser that clips pages into notes.
- Projections — Agenda, Board, Timeline, Graph, Claims, Evidence, Review and Session Replay, all querying the same Markdown, plus a
vemry-queryblock that embeds one in a note. - The agent — typed tools over your vault; every write shows a diff and waits for your approval unless you have allowed it; every session is logged.
- Privacy — privacy zones make marked notes and blocks nonexistent to the model. Opening a note makes no network request: remote images wait for a click. The network exits are a closed, documented list, and there is no telemetry.
- Scripting — user scripts in a sandbox with no file, network or process access; a script's first write asks you.
- Account (optional) — sign in by email code for hosted AI credit, billed per call at the vendor's price × 1.3 and listed on the account page. Your own API key works without an account.
- Languages — English, Chinese, Japanese, Korean, French, German and Spanish.
Downloads
| Platform | File |
|---|---|
| macOS, Apple silicon | vemry_1.0.0_aarch64.dmg |
| macOS, Intel | vemry_1.0.0_x64.dmg |
| Windows, x64 | vemry_1.0.0_x64-setup.exe |
| Linux, x86_64 | vemry_1.0.0_amd64.AppImage or vemry_1.0.0_amd64.deb |
- macOS 10.15 or later. The app is signed and notarized.
- Windows: the installer is not code-signed yet, so SmartScreen may show "Windows protected your PC" on first run — choose More info → Run anyway.
- Updates: the app checks only when you click Check for updates (or at launch, if you turn that on), and installs only when you confirm.
Full changelog: https://vemry.app/changelog · Docs: SYNTAX · ARCHITECTURE
中文
vemry 的第一个版本:一个本地优先、面向学习和研究的 Markdown 工作区,内置一个在权限门后面操作笔记的 AI agent。笔记就是你选定文件夹里的普通 Markdown 文件,应用里的其余一切都是它们的投影视图。
1.0.0 包含
- 写作:实时预览的 Markdown 编辑器,
/插入菜单、callout、折叠、行内字段;命令面板(⌘P / Ctrl+P);在任何应用里速记(⌥⇧Space / Alt+Shift+Space),也可以从配对的手机速记。 - 阅读与学习:从 arXiv 链接或 PDF 导入论文;笔记、网页、PDF 的双语对照翻译;闪卡与复习视图;内置浏览器,可把网页剪藏成笔记。
- 投影视图:日程、看板、时间线、图谱、断言、证据、复习、会话回放,全部查询同一份 Markdown;
vemry-query代码块可把投影视图嵌进笔记。 - Agent:基于知识库的类型化工具;每次写入都先展示 diff,等你批准(除非你已授权);每个会话都有记录。
- 隐私:隐私区里的笔记和块对模型而言根本不存在。打开笔记不联网,远程图片点了才加载。网络出口是一份封闭、有文档的清单,没有遥测。
- 脚本:用户脚本运行在沙箱里,不能访问文件、网络和进程;脚本第一次写入会询问你。
- 账户(可选):邮箱验证码登录,使用托管的 AI 额度,按厂商价格 × 1.3 逐次计费,每次调用都列在账户页。用自己的 API key 无需账户。
- 界面语言:英文、中文、日文、韩文、法文、德文、西班牙文。
下载:macOS Apple 芯片选 aarch64.dmg,Intel 选 x64.dmg;Windows 选 x64-setup.exe;Linux 选 .AppImage 或 .deb。
- macOS 需要 10.15 或更高版本,应用已签名并公证。
- Windows 安装包暂未代码签名,首次运行时 SmartScreen 可能提示"Windows 已保护你的电脑",点"更多信息"→"仍要运行"。
- 更新:只有你点"检查更新"时才会检查(或在你打开"启动时检查"之后),并且要你确认才会安装。