Try ModelDock v0.2 · 测试邀请与反馈入口：直接预览、大文件、多模型对比 #1

Try ModelDock v0.2 — inspect and compare 3D models where you code

We built ModelDock to make checking reconstruction results and CAD output as convenient as opening a source file. Open a model, rotate, pan and zoom, then inspect the next revision from the same viewpoint.

Download v0.2.0 · Project and installation guide

What you can try

• Preview where you work: a VS Code custom editor, native DeepSeek Harness 0.2.2-alpha.2 Web preview, an adapter for hosts supporting OpenAI file entrypoints, and a CLI local viewer.
• View large reconstructions: bounded, sampled previews of binary PLY/STL sources above 5 GiB, with an 8 GiB default source ceiling.
• Compare multiple models: colored overlays, per-model opacity and visibility, and synchronized side-by-side views.
• Use a reference model: choose an STL or another model as ground truth, confirm coordinate units, and explicitly apply local rigid ICP or manual translation, rotation and scale.
• Measure geometric differences: sampled accuracy/completeness, RMSE, P95, Chamfer distance, tolerance-based precision/recall/F-score, and deviation heatmaps.
• Keep results reproducible: export JSON/CSV with transforms, units, thresholds, seed and sampling provenance.
• Work locally: model parsing, rendering and comparisons run locally, with Chinese and English controls.

A quick test

1. Install the package for your host from the release.
2. Open a PLY or STL file and try rotation, pan, zoom and fit.
3. Add a second model, choose a reference, confirm units, and try overlay or synchronized views.
4. Run an explicit alignment if needed, evaluate the comparison, and export a report.

Please share what you tested and what would make the workflow better. Comments in English or Chinese are welcome. For a reproducible bug, please open a separate issue and link it here so each problem can be tracked individually.

A useful report includes:

Host and version:
Operating system:
ModelDock version:
Format and file size:
Steps to reproduce:
Expected / observed result:
Screenshot or a minimal non-private sample, if available:

欢迎测试 ModelDock v0.2：在编程工作流中直接查看、对齐和比较三维模型

我们开源了 ModelDock，希望让检查三维重建和 CAD 输出像打开代码文件一样顺手：生成模型后，直接点击查看，旋转、平移、缩放，再继续验证下一次修改。

下载 v0.2.0 · 项目与安装说明

欢迎体验这些功能

• 四种入口： VS Code 文件预览、DeepSeek Harness 0.2.2-alpha.2 Web 原生预览、兼容 OpenAI 文件入口协议的宿主适配器，以及 CLI 本地查看器。
• 大型重建结果预览： 二进制 PLY/STL 支持 5 GiB 以上源文件的有界采样预览，默认源文件上限为 8 GiB。
• 多模型对比： 彩色叠加、透明度与显示开关、同步并排查看。
• 基准与对齐： 将 STL 等模型设为真值参考，确认坐标单位后，显式执行局部刚性 ICP，或手动调整平移、旋转与缩放。
• 几何评价： 采样准确度与完整度、RMSE、P95、Chamfer 距离、容差下的精确率/召回率/F-score，以及偏差热图。
• 结果可复现： JSON/CSV 报告保留变换、单位、阈值、随机种子与采样依据。
• 本地处理、中英双语： 模型解析、渲染与比较在本地完成。

建议这样试用

安装对应宿主的发布包 → 打开一个 PLY/STL → 加入第二个模型 → 选择参考、确认单位 → 对齐与评价 → 导出报告。

欢迎直接在本帖分享试用体验、需求和建议。 遇到可复现的问题，建议单独创建 Issue 并链接本帖，便于逐项跟踪。请附上宿主与版本、操作系统、文件格式与大小、操作步骤，以及预期和实际现象；方便时可提供截图或不含隐私的最小样例。

感谢一起把三维模型检查变得更顺手！

Screenshots / 真实界面

The shared viewer running through the local CLI with the repository's original demo fixtures. / 本地 CLI 入口运行共享查看器，使用仓库随附的原创示例模型。

DeepSeek Harness community showcase / 插件社区展示帖