OpenScience v0.1.2

OpenScience v0.1.2

This is a bugfix release for the first OpenScience public builds. It focuses on improving the reliability of Science mode and MCP permission handling during multi-turn conversations.

Fixes

• Fixed a Science mode issue where built-in OpenScience MCP servers could ask for approval again in later turns even after the user selected full access or always allow.
• Fixed the approval flow so Codex/OpenScience correctly persists always allow choices for Science artifact and evidence MCP tools.
• Added a safe managed Codex MCP configuration sync for OpenScience's built-in evidence and artifact tools. This does not write PaperClip API keys or per-session secrets into Codex config.
• Hardened the permission dialog handling for both normal ACP approvals and the pet confirmation overlay.
• Bumped the desktop package version to 0.1.2 and rebuilt installers for macOS, Windows, and Linux.

Installers

• macOS Apple Silicon: OpenScience-0.1.2-mac-arm64.dmg
• macOS Intel: OpenScience-0.1.2-mac-x64.dmg
• Windows x64: OpenScience-0.1.2-win-x64.exe
• Linux x64: OpenScience-0.1.2-linux-amd64.deb
• Linux ARM64: OpenScience-0.1.2-linux-arm64.deb

OpenScience v0.1.2 中文说明

这是 OpenScience 首个公开构建后的 bugfix 版本，主要修复科学研究模式在多轮对话中 MCP 权限不能稳定延续的问题。

修复内容

• 修复科学研究模式中，即使用户选择了完全访问或总是允许，第二轮对话仍可能再次弹出 MCP 授权框的问题。
• 修复 always allow 选项没有正确传递和持久化到 Codex/OpenScience 权限系统的问题。
• 新增 OpenScience 内置证据与 artifact MCP 的安全配置同步逻辑，不会把 PaperClip API key 或单次会话密钥写入 Codex 配置。
• 加固普通 ACP 授权弹窗和桌宠确认弹窗的权限选项处理。
• 版本更新到 0.1.2，并重新构建 macOS、Windows 和 Linux 安装包。

安装包选择

• M 系列芯片 Mac：下载 OpenScience-0.1.2-mac-arm64.dmg
• Intel 芯片 Mac：下载 OpenScience-0.1.2-mac-x64.dmg
• Windows 电脑：下载 OpenScience-0.1.2-win-x64.exe
• 常见 x86 Linux：下载 OpenScience-0.1.2-linux-amd64.deb
• ARM Linux：下载 OpenScience-0.1.2-linux-arm64.deb

Full Changelog: v0.1.1...v0.1.2

OpenScience v0.1.1

OpenScience v0.1.1 is the first public desktop release of OpenScience.

This version brings a local AI research workspace, Science artifacts, medical evidence reports, and knowledge distillation workflows into an installable desktop app.

The goal of this first release is simple: make AI-assisted research feel like a real working environment, not just a chat box. You can ask questions, run analyses, inspect files, turn results into traceable reports, and reuse what you learned as skills for future projects.

Highlights

1. Four Work Modes

OpenScience now provides four entry points for different kinds of work:

• Science Research Mode
  Work inside a research project, run code, inspect data files, generate figures, and organize outputs as scientific artifacts with provenance. It is designed for computational biology, protein analysis, single-cell workflows, chemistry, data analysis, and other dry-lab scientific tasks.

• Medical Evidence Mode
  Build evidence-backed medical reports with indexed references, source traceability, and a report layout designed for clinical or biomedical evidence review.

• Goal Mode
  Use the normal local agent workspace for broader goals: reading a project, modifying code, organizing files, writing documents, or completing multi-step tasks.

• Knowledge Distillation Mode
  Turn conversations, lab procedures, project routines, and repeated workflows into reusable skills. The generated skill can be reviewed, edited, enabled, and reused later.

2. Science Artifacts

Science results are treated as research objects, not loose attachments.

A Science artifact can bring together:

• the visible report, figure, table, notebook, or file preview
• source code and generated files
• inputs, run logs, and environment notes
• evidence links and provenance records
• export paths for reports and downstream review

This is the beginning of the OpenScience artifact system: a place where a result can be opened, inspected, revised, and traced back to how it was produced.

3. Evidence-First Reporting

OpenScience inherits and adapts the evidence-report style from the medical evidence workflow. Reports are designed to be readable first, while still keeping the supporting evidence reachable.

4. Skills and Workflow Memory

The Skills page can list available skills, highlight skills created through Knowledge Distillation, render skill Markdown, and support editing or downloading. This makes it easier for a lab or individual researcher to turn repeated work into an explicit SOP.

5. Local-First Desktop Experience

OpenScience runs as a desktop app and keeps the working experience close to local files, local projects, and local execution. The v0.1.1 release also removes the login gate so users can enter the workspace directly after installation.

Which File Should I Download?

Most users only need one installer. The latest*.yml files are auto-update metadata and are not meant to be opened manually.

Your system	Download	Example
macOS Apple Silicon	OpenScience-0.1.1-mac-arm64.dmg	M1, M2, M3, M4 MacBook / Mac mini / iMac
macOS Intel	OpenScience-0.1.1-mac-x64.dmg	Older Intel MacBook / iMac
Windows 10/11 64-bit	OpenScience-0.1.1-win-x64.exe	Most modern Windows laptops and desktops
Linux x86_64 / amd64	OpenScience-0.1.1-linux-amd64.deb	Ubuntu / Debian on Intel or AMD CPUs
Linux ARM64	OpenScience-0.1.1-linux-arm64.deb	ARM Linux machines or ARM servers
macOS zip fallback	OpenScience-0.1.1-mac-arm64.zip or OpenScience-0.1.1-mac-x64.zip	Use only if the DMG route is inconvenient

First Launch Notes

• macOS: if macOS blocks the app because it is from an unidentified developer, open System Settings → Privacy & Security, then choose Open Anyway for OpenScience. Only do this when the app was downloaded from this official release page.
• Windows: if SmartScreen shows a warning, choose More info → Run anyway only if the installer came from this release page.
• Linux: install the .deb package with your system package installer, or use a command such as sudo apt install ./OpenScience-0.1.1-linux-amd64.deb from the download folder.

Good First Tasks

You can start with natural language rather than strict commands:

• Science Research Mode
  “I have been looking at the human hemoglobin alpha chain and want a quick structural overview. Could you check HBA1 in UniProt, look at the matching AlphaFold structure, and make a clear figure showing the overall structure, pLDDT confidence, and a few important residues?”

• Medical Evidence Mode
  “I want to understand first-line management for adult anaphylaxis. Please find guidelines and systematic reviews, then organize the answer as an evidence-indexed report.”

• Goal Mode
  “Please read this local project, figure out how it starts, identify the core modules, and give me a minimal runnable check.”

• Knowledge Distillation Mode
  “Turn the workflow from this conversation into a reusable skill, keeping the SOP, trigger conditions, inputs, outputs, and cautions.”

Known Limitations

This is the first public release, so some edges are still being hardened.

• Scientific and medical conclusions still require human verification.
• Some evidence search or external-database features may require API keys or network access.
• Advanced scientific viewers, artifact review, and richer provenance workflows will continue to improve in later releases.
• Large computations still depend on the local environment, installed packages, available hardware, and user permissions.

Full Changelog

Full Changelog: https://github.com/ResearAI/OpenScience/commits/v0.1.1

---

中文说明

OpenScience v0.1.1 是 OpenScience 的第一个公开桌面版本。

这个版本把本地 AI 科研工作空间、Science artifact、医学循证报告和知识沉淀流程放进一个可以直接安装使用的桌面应用里。

这个首版的目标很直接：让 AI 辅助科研不只是聊天，而是一个可以真正工作的环境。你可以提出研究问题、运行分析、查看文件、把结果整理成可追踪报告，并把常用流程沉淀成以后可复用的技能。

版本亮点

1. 四种工作模式

OpenScience 现在提供四种入口，对应不同使用场景：

• 科学研究模式
  在研究项目中工作，运行代码、查看数据文件、生成图表，并把结果组织成带 provenance 的科研 artifact。适合计算生物学、蛋白分析、单细胞流程、化学、数据分析和其他干实验科研任务。

• 医学循证模式
  生成带证据编号、来源追踪和专门报告版式的医学循证报告，适合临床或生物医学证据综述。

• 目标模式
  使用普通本地 Agent 工作区完成更通用的任务，例如阅读项目、修改代码、整理文件、写文档或处理多步骤目标。

• 知识沉淀模式
  把对话、实验室流程、项目习惯和重复工作沉淀成可复用的 skills。生成后的 skill 可以预览、编辑、启用，并在之后继续使用。

2. 科研 Artifact

科研结果不再只是“附件”，而是一个带上下文的科研对象。

一个 Science artifact 可以把这些内容组织在一起：

• 可见的报告、图、表、notebook 或文件预览
• 源代码和生成文件
• 输入、运行日志和环境说明
• 证据链接与 provenance 记录
• 报告导出和后续复核路径

这是 OpenScience artifact 系统的起点：一个结果可以被打开、检查、修改，并追踪到它是怎样产生的。

3. 证据优先的报告体验

OpenScience 复用了并改造了医学循证模式里的报告风格。报告优先保证可读性，同时保留证据、来源和文件链路，方便之后复核。

4. 技能与流程记忆

技能页面可以展示当前可用的 skills，并把通过知识沉淀生成的技能放在更显眼的位置。你可以阅读、编辑和下载这些技能，让个人或实验室的常用流程逐渐变成明确的 SOP。

5. 本地优先的桌面体验

OpenScience 以桌面应用形式运行，围绕本地文件、本地项目和本地执行来组织体验。v0.1.1 也去除了登录门槛，安装后可以直接进入工作区。

应该下载哪个文件？

大多数用户只需要下载一个安装包。latest*.yml 是自动更新用的元数据文件，不需要手动下载或打开。

你的系统	下载文件	举例
macOS Apple Silicon	OpenScience-0.1.1-mac-arm64.dmg	M1、M2、M3、M4 MacBook / Mac mini / iMac
macOS Intel	OpenScience-0.1.1-mac-x64.dmg	较早的 Intel MacBook / iMac
Windows 10/11 64 位	OpenScience-0.1.1-win-x64.exe	大多数现代 Windows 笔记本和台式机
Linux x86_64 / amd64	OpenScience-0.1.1-linux-amd64.deb	Intel 或 AMD CPU 的 Ubuntu / Debian
Linux ARM64	OpenScience-0.1.1-linux-arm64.deb	ARM Linux 机器或 ARM 服务器
macOS zip 备用包	OpenScience-0.1.1-mac-arm64.zip 或 OpenScience-0.1.1-mac-x64.zip	仅在不方便使用 DMG 时选择

首次启动提示

• macOS：如果系统提示“无法验证开发者”，请进入 系统设置 → 隐私与安全性，找到 OpenScience 并选择 仍要打开。请只对从本 Release 页面下载的安装包这样操作。
• Windows：如果 SmartScreen 提示风险，请确认安装包来自本页面后，再选择 更多信息 → 仍要运行。
• Linux：可以用系统安装器安装 .deb，也可以在下载目录运行类似 sudo apt install ./OpenScience-0.1.1-linux-amd64.deb 的命令。

可以先试试这些任务

你可以直接用自然语言开始，不需要写成标准命令：

• 科学研究模式
  “我最近在看人血红蛋白 alpha 链，想先快速了解一下它的结构特点。你能帮我查一下 UniProt 里的 HBA1，对应的 AlphaFold 结构也一起看看吗？最好顺手做一张清楚一点的图，把整体结构、pLDDT 可信度和几个比较重要的残基标出来。”

• 医学循证模式
  “我想了解成人过敏性休克的一线处理方案。请帮我找指南和系统综述，整理成一份有证据编号的循证报告。”

• 目标模式
  “帮我阅读这个本地项目，先判断它怎么启动、核心模块在哪里，然后给我一个最小可运行检查。”

• 知识沉淀模式
  “把这次对话里形成的实验流程整理成一个以后可复用的技能，保留 SOP、触发条件、输入输出和注意事项。”

当前限制

这是第一个公开版本，因此仍有一些边界在继续打磨：

• 科学与医学结论仍然需要人工复核。
• 部分证据搜索或外部数据库能力可能需要 API key 或网络环境。
• 科研 viewer、artifact review 和更完整的 provenance 流程会在后续版本继续增强。
• 大型计算仍然取决于本机环境、软件包、硬件和用户授权。

完整变更

Full Changelog: https://github.com/ResearAI/OpenScience/commits/v0.1.1