Web · Docs
How the research and operations knowledge is recorded and kept usable.
Per-server execution runbooks for the AI model pipeline — data merging, inference, tracker runs, and plotting commands — plus usage manuals for the vortex tracker and the legacy data server, so any run can be reproduced without tribal knowledge.
An Obsidian vault organized by project, with daily logs and weekly report decks. Experiment settings, environment setup (uv/conda), and debugging notes are linked so results stay traceable to the exact commands that produced them.
This portal itself is documented: a content guide explains how the JSON manifests (realtime forecasts, posters, tool links) drive each page, so new typhoons and posters are published by editing data files, not markup.
Internal manuals themselves are not published here — this page describes the documentation system.