Repository navigation
Releases: KaiserIIII/QLDeviceCheck_Generic_WebUI_Linux
Release list
v3.1.0 · Bilingual field acceptance workbench
v3.1.0 · Bilingual field acceptance workbench
This release adds a complete English interface alongside 简体中文 for the acceptance workbench and the compatibility scan page. The interface remembers the selected language, defaults from the browser locale, keeps entered metadata and active task state intact during switching, and preserves original device names, notes and protocol evidence. Printable HTML reports follow the selected language; JSON and CSV exports remain byte-compatible.
The release also includes the traceable acceptance workflow from v3.0: configuration-driven device coverage, retained failure evidence, failed-only retests with lineage, scope-aware comparison, protected legacy report downloads, and offline demo isolation. It is packaged for Windows x64 and Linux x86_64 with the audited psutil and pyserial runtime wheels and SHA-256 checksums.
Verification
- 100 pytest tests passed locally; the original 23 offline self-tests passed.
- GitHub Actions offline matrix passed on Windows/Linux with Python 3.10/3.13: run 37766304880.
- Real Chromium browser checks passed for both languages, report download, evidence, retest, history filtering, cross-page comparison, reload preference, blocked storage fallback, legacy page and 390 px layout with zero page errors.
- Windows x64 package was installed with
PIP_NO_INDEX=1and exercised through task, report, healthy retest and restart persistence. - Automated checks use fake/synthetic adapters and loopback sockets. They do not issue writes to unknown industrial equipment.
Install
Install CPython 3.10 or newer, download the platform package below, then run Start-Demo.cmd on Windows or bash run_demo.sh on Linux. For field mode, review config/device_list.json and use run_web.sh with the station data directory.
The package source tree is the verified v3.1 bilingual workbench commit merged into main by PR #1. v3.0.0 remains available for rollback.
v3.0.0 · 工控设备现场验收工作台
工控设备检测系统已升级为 现场验收工作台 v3.0.0,面向设备厂商、系统集成商的交付验收与维修复测。项目方已确认项目测试可用。
本次升级
- 应检设备全覆盖:未响应设备保留失败结果,任务保留工位、批次、操作员和时间。
- 可追溯证据:保存配置快照、SHA-256、全部协议探测及载体检查,导出 JSON/CSV/可打印 HTML。
- 维修闭环:失败项复测沿用原配置,对比恢复与退化;支持跨历史页面选择任务,并明确部分复测范围。
- 工位架构:分离领域、服务、存储、协议适配器和界面;SQLite 持久化、数据库独占锁、异常恢复、数据库分页和汇总。
- 专业界面:桌面/移动端工作台、配置校验、证据详情、离线演示;旧扫描界面仍可使用。
- 交付能力:Windows/Linux 离线运行依赖、安装与启动脚本、来源清单和 SHA-256 校验文件。
下载与启动
先安装 CPython 3.10+,包含 venv/pip。发行包已包含运行依赖,无需联网下载依赖。
Windows x64:下载 QLDeviceCheck-v3.0.0-windows-x64.zip 并解压,双击 Start-Demo.cmd,浏览器打开 http://127.0.0.1:8080。使用设备配置执行正式验收:
powershell -NoProfile -ExecutionPolicy Bypass -File .\run_web.ps1 --config config\device_list.json --data-dir data\liveLinux x86_64/glibc:下载对应包,解压后进入 QLDeviceCheck-v3.0.0:
bash setup_linux.sh
bash run_demo.sh
# 使用设备配置执行正式验收:
bash run_web.sh --config config/device_list.json --data-dir data/live演示任务和导出明确标记 SIMULATED。正式验收前按工位设备清单配置目标;批次记录保存在指定数据目录。安装、访问令牌、更新和备份说明均包含在包内 docs/RELEASE.md 和 docs/DEPLOYMENT.md。
发布验证
- 93 项自动测试通过,原有 23 项离线自检通过。
- 新架构模块行覆盖率 92%;架构/协议引擎/传输层合计 86%。
- 最终提交的 Linux/Windows、Python 3.10/3.13 四组检查通过:GitHub Actions。
- 浏览器验证包括 21 条保存任务的跨页对比、证据导出、复测、受令牌保护的报告下载及 390 px 布局;无脚本错误。
- 本次 Windows 发行包全新离线安装、实际本机服务、任务/报告/复测/重启保存检查通过。
- 两个平台包的源码 SHA、运行依赖 SHA-256 和脚本格式已核对;独立代码审查及修复复审完成。
源码固定为 526fde1d7f3a82ea3194773ba89da7e3d19e970c,每个包含 RELEASE-MANIFEST.json 和依赖许可证。自动化检查使用模拟设备与本机回环;完整验证范围见包内 docs/VALIDATION.md。
源码变更:PR #1。下载后可使用随附 SHA256SUMS.txt 核对安装包。