Research Vault
A living, offline-first knowledge base for the transition from software engineer → researcher in event streams, web analytics, and clickstream prediction.
The map of the whole thing lives in plan (the HOME / top-level Map of Content). Start there.
How this vault is organized
Folder-per-domain, numeric prefixes for ordering. One idea per note; link liberally with [[wikilinks]];
let the graph emerge. Filenames match link text exactly, so Obsidian resolves links by title.
| Folder | What lives here |
|---|---|
00_Home/ |
HOME + sub-MOCs (the navigation layer) |
10_Foundations/ |
The math & fundamentals — authored reference notes |
20_Domain/ |
Clickstream / event-stream subject knowledge + the offline compute stack |
30_Methods/ |
The 3 research areas + synthetic-data craft |
40_Experiments/ |
Lab notebook — one note per experiment, hypothesis first |
50_Literature/ |
Zotero-linked paper notes + reading lists |
60_Craft/ |
How to be a researcher — temperament, traps, checklists |
70_Tracks/ |
Research portfolio — projects |
80_Templates/ |
Note templates (experiment, paper, track) |
90_Inbox/ |
Fleeting notes; capture first, triage weekly |
Note conventions
Each note carries light YAML frontmatter and an Up: backlink to its section MOC:
---
tags: [foundations]
status: seed # stub → seed → growing → evergreen
---
# Note title
**Up:** [[10_Foundations-MOC]]
status legend: stub (placeholder, intent only) → seed (has real content, incomplete) →
growing (actively developed) → evergreen (stable, trustworthy reference).
Using it offline
Everything here is plain Markdown — readable in any editor, no internet required. Open the folder as an Obsidian vault for graph view + backlinks, or just read the files. The compute/tooling stack you'll need locally is documented in Offline compute stack.
Publishing the web book
The vault also builds into a static HTML "web book" (sidebar nav, backlinks, rendered LaTeX) and publishes
via ginx. The generator is tools/build-site.ts (Bun + markdown-it + KaTeX); output goes to _site/
(git-ignored).
bun tools/build-site.ts # rebuild _site/ from the markdown
ginx sync _site # ship the changes (instant; rollback with `ginx rollback`)
Live at https://research-vault.ginx.cloud (publicly accessible). First-time setup was
ginx create _site --subdomain research-vault; after that, edit notes → build-site → ginx sync.
Git workflow
The research history is an audit trail — commit often.
git add -A
git commit -m "notes: <what changed>"
Large data/model artifacts are git-ignored (see .gitignore) and meant to be versioned with DVC instead,
so the repo stays text-only and fast to clone.