Repository navigation
Releases
WhyJ Quant Desktop 1.4.0
WhyJ Quant Desktop 1.4.0
This Windows release updates the packaged research terminal from 1.3.1 with Pi 1.2, a configurable installation directory, and the shared JQ visual identity.
Highlights
- Choose where to install. The per-user Windows installer supports other drives and paths containing spaces or Chinese characters, while research data stays in its existing location. (#16)
- A consistent JQ identity. The Desktop shell, terminal and newly generated reports share the JQ mark and warm color roles. Existing report files retain their original appearance. (#28)
- Clearer research controls. Interrupt research with Ctrl+G, retry failed quotes with Ctrl+R, and find configuration and report guidance from the terminal. (#15)
What's Changed
Runtime and Research
- Adopt Pi 1.2 and consolidate research task, evidence and report workflows. This change landed directly without a PR. (d00a856)
- Reuse Pi APIs for explicitly selected community skills and MCP configuration. Skills are not downloaded on startup; MCP configuration does not start a JQ MCP client. (#35)
- Add research data contracts and price-volume calculations. (#6)
Fixes
- Correct backtest equity and compound annual return calculations. (#4)
- Improve Chinese answer qualification checks and reuse unchanged message layouts during long conversations. (#15)
Documentation and Maintenance
- Consolidate product scope, requirements and usage into the current Quant Harness guide. (#66)
- Stabilize CI policy and test lifecycles, including deterministic overview recovery and skill cancellation tests. These are verification improvements, not new application features. (#41, #42, #45)
Installation and Upgrade
Download WhyJ-Quant-Setup-1.4.0-win-x64.exe from this Release's Assets and compare its SHA-256 with SHA256SUMS.txt. Windows 10 1809+ and Windows 11 x64 are supported. Bun, Python 3.13.12, AKShare 1.18.63, AKQuant and Git are bundled.
Reinstallation keeps the existing installation directory by default. Settings, sessions and reports remain in %APPDATA%/WhyJ Quant/workspace/.ohquant, independently of the application directory. AKShare is the default data source; configure an AI model connection with Ctrl+P.
Known Limitations
- The installer is not Authenticode-signed. Windows may show an unknown-publisher warning, and strict application-control policies may block startup.
- Model and market-data services require network access and any credentials required by the selected provider. Live provider behavior is not certified by installer checks.
- The packaged Desktop still hosts the terminal interface. The separate Web research workspace introduced in #27 is not the Desktop entry; Electron workspace integration and packaging remain tracked by #47 and #48.
Verification
The published installer was built from 4a83606. Publication requires both platform CI checks and Windows installation acceptance. Automated coverage includes startup without developer runtimes on PATH, bundled Python/Git, settings persistence, Unicode multiline paste, resizing, restart/exit, custom installation paths, shortcuts, reinstallation and retained research data. These checks do not replace interactive acceptance on a user's Windows machine.
Full Changelog
Verification: Windows build and installation acceptance.
WhyJ Quant Desktop 1.3.1
Windows 桌面版提供一键安装包。下载 WhyJ-Quant-Setup-1.3.1-win-x64.exe,双击安装后从桌面或开始菜单启动。本次修复了小窗口及调整字号后终端底部状态栏可能被裁切的问题。
- Electron 内直接使用原 CLI 界面,命令、设置、组合、历史会话与报告保持一致。
- 内置 Bun、Python 3.13.12、AKShare 1.18.63、AKQuant 和 Git,无需另装开发环境。
- 首次使用默认选择免 Token 的 AKShare;按 Ctrl+P 配置自己的 AI 模型连接。
- 支持中文输入、复制、多行安全粘贴、字号调整和窗口状态保存。
- 数据保存在
%APPDATA%/WhyJ Quant/workspace/.ohquant,升级保留设置、会话和报告。
支持 Windows 10 1809+ / Windows 11 x64。模型和行情服务需要网络及相应账号权限。此版本未作 Authenticode 签名,Windows 可能提示未知发布者;启用严格应用控制的设备可能阻止运行,本机已观察到这种限制。
安装包经干净 Windows 环境实际安装,并验证了无开发运行时 PATH 启动、CLI 面板、设置持久化、中文多行粘贴、窗口缩放、随包 Python/Git、重启与退出。此验收不代表所有外部模型和行情供应商均可用。可用同页 SHA256SUMS.txt 核对下载完整性。
验证记录:Windows 构建与安装验收。
WhyJ Quant Desktop 1.3.0
Windows 桌面版正式提供一键安装包。下载 WhyJ-Quant-Setup-1.3.0-win-x64.exe,双击安装后从桌面或开始菜单启动。
- Electron 内直接使用原 CLI 界面,命令、设置、组合、历史会话与报告保持一致。
- 内置 Bun、Python 3.13.12、AKShare 1.18.63、AKQuant 和 Git,无需另装开发环境。
- 首次使用默认选择免 Token 的 AKShare;按 Ctrl+P 配置自己的 AI 模型连接。
- 支持中文输入、复制、多行安全粘贴、字号调整和窗口状态保存。
- 数据保存在
%APPDATA%/WhyJ Quant/workspace/.ohquant,升级保留设置、会话和报告。
支持 Windows 10 1809+ / Windows 11 x64。模型和行情服务需要网络及相应账号权限。此版本未作 Authenticode 签名,Windows 可能提示未知发布者;启用严格应用控制的设备可能阻止运行,本机已观察到这种限制。
安装包经干净 Windows 环境实际安装,并验证了无开发运行时 PATH 启动、CLI 面板、设置持久化、中文多行粘贴、窗口缩放、随包 Python/Git、重启与退出。此验收不代表所有外部模型和行情供应商均可用。可用同页 SHA256SUMS.txt 核对下载完整性。
验证记录:Windows 构建与安装验收。
WhyJ Quant 1.2.0
Added
- 新增量化研究 Dashboard:内置 K 线/成交量/MACD、NAV 与收益曲线、风险收益指标、因子图、数据覆盖度和会话统计;图表颜色与 WhyJ TUI 主题 token 保持一致。
- 新增完整 Trajectory 复原:按 Turn 统计 calls、returns、skills 和 outputs,保留用户/模型输入、assistant 输出、thinking、compact、Tool/Bash/Skill 名称、Call ID、调用参数、完整返回与源 JSONL 定位信息。
- 新增 Artifact 自动生成配置。
/config中开启Auto Artifacts后,量化分析 turn 在agent_end持久化完成后生成 Dashboard + Trajectory,并自动打开浏览器预览。 - 新增研究对象标题推断,报告标题优先使用名称与证券/基金代码;无法识别时使用中性的“量化研究报告”。
- 新增 Artifact 恢复索引和上下文 note,使恢复后的 agent 能继续理解“这个报告”或“刚才的 artifact”。
Changed
/artifact收敛为无参数的一级命令,只负责浏览、恢复和打开已有报告;生成模式不再作为 Slash Command 的二级选项。- Settings 按
Agent Runtime、Research Data、Workspace Experience三组规整;面板说明与操作提示进一步精简并增加可读间距。 - Codex、Claude、Agents 与 pi 的用户级 skill 集成合并为单一
External Skills开关,项目量化 skills 始终可用。 - Slash command 目录、Composer 建议、Help 与 CLI 帮助共享统一 catalog,公开描述保持短而一致。
- Agent 的证券、基金与组合研究流程内置技术因子和风险指标要求;有有效收益序列时保留波动率、VaR/CVaR、最大回撤、偏度、峰度与风险调整收益,供 Dashboard 解析。
- Source provider 改为可注入 adapter 和分层 fallback;AKShare 普通 A 股日线依次尝试东方财富、新浪、腾讯,上层仍可继续降级到 Tushare。
Fixed
- 修复 Overview 刷新失败或超时后长期停留在
refreshing的问题;刷新状态可恢复,过期请求结果不会覆盖较新的数据,并保留最近一次成功面板。 - 修复组合中的开放式基金被误判为交易所证券、只得到全零行情的问题;基金代码会保留
.OFprovider 语义,无有效数据时显示--。 - 修复 Artifact 恢复时来源 session 已不存在仍继续打开的问题;报告列表和恢复前都会校验并删除孤立 HTML/metadata。
- 修复 resume 后 TUI 结构化表格和工具输出丢失,以及 Artifact Trajectory 未完整展示 Tool、Bash、Skill 调用及返回细节的问题。
- 修复报告标题沿用用户第一句话、K 线缺失或误用 Trajectory 数据生成金融图表等问题。
Verification
- 覆盖 Artifact 生成/恢复/孤立报告清理、Trajectory schema 兼容、K 线渲染、开放式基金报价、Overview 超时恢复与竞态保护的回归测试。
- 发布前执行
bun run test、bun run typecheck、bun run check:architecture、bun run build与 npm pack dry-run。
v0.2.0
Summary
This release reduces oh-my-quant to a smaller runnable core focused on the commands that actually work end to end.
Highlights
- Keep only the runnable CLI surface:
data,factor,backtest,risk,benchmark,validate - Remove placeholder and non-CLI modules such as
research,intel, andconsensus - Remove the benchmark design layer: HTML dashboard pages, K-line pages, metric pages, and
docs/DESIGN.md - Simplify data loading to stable runtime paths centered on
yfinanceandakshare - Fix benchmark/backtest runtime issues and package the runtime directories into the wheel
- Add
ROADMAP.mdto keep future non-CLI directions documented without keeping dead code in the tree
Breaking Cleanup
- Deleted non-runnable skill docs and related unused scripts
- Deleted unused benchmark sample assets and generated HTML artifacts
- Deleted unused data source and analysis helpers that were no longer connected to the CLI
Verification
python -m cli.main --version->0.2.0python -m cli.main validate all->6/6uv build-> source distribution and wheel built successfully