steveruizoktldraw/tldraw-datasets

该数据集包含来自tldraw（一个在线白板工具）的房间快照历史数据。每个数据行代表一个编辑会话（称为“轨迹”），包含一系列周期性的JSON快照和对应的PNG图像。数据集的结构包括文件名、JSONL序列和图像序列。JSONL序列中的每个元素都是一个结构体，包含多种字段，如类型（快照或差异）、时间戳、文档变更等。图像序列与JSONL序列一一对应，每个图像代表对应JSONL条目的状态。数据集的使用方法包括加载数据集、访问各个字段以及通过应用差异重建完整的RoomSnapshot。

This dataset contains snapshot history data from tldraw (an online whiteboard tool) rooms. Each row represents a single editing session (a "trajectory") from one room, captured as a sequence of periodic JSON snapshots with matching rendered PNGs. The dataset structure includes filename, JSONL sequence, and images sequence. Each element in the JSONL sequence is a struct with various fields such as kind (snapshot or diff), timestamp, document changes, etc. The images sequence corresponds one-to-one with the JSONL sequence, each image representing the state of the corresponding JSONL entry. Usage includes loading the dataset, accessing various fields, and reconstructing a full RoomSnapshot by applying diffs.