§0 Introduction
What dat0 is, the artifact metaphor, and how the .dat0 format spec is organized — §0 through §6. Designed, not shipped.
In development. This format is designed, not shipped. Nothing to install yet — join the waitlist.
dat0 is a local-first data workbench. Point it at a file or a database, explore and transform what you find with full lineage, and — when you choose — package the entire workflow into a single, replayable artifact. Your data stays on your machine; the cloud is something you reach for, never something you start with.
It is not a BI tool, not a notebook, and not a database client. It is the place your data lands first — large or small, file or table — and the place a finished workflow leaves as one file you can hand to someone else.
The site is the file
Everything dat0 produces is one artifact, so this site is built from that
artifact too. The landing page compiles a .dat0 file as you scroll — chunk by
chunk, from the DATA range to the sealed finish — and these docs are the same
file's format spec: the same anatomy, the same offsets, the same chunk names,
drawn from one content source. Two surfaces, one story — they cannot drift.
What a .dat0 file carries
A .dat0 file is a small binary container: a fixed header, five self-framing
chunks, and a seal. Every chunk opens with a four-byte ASCII magic and its own
payload length, so a reader can scan an artifact — or skip a chunk it does not
understand — without parsing the whole file:
workflow.dat0
├── header "DAT0" magic · version · chunk count · directory pointer
├── DATA the source tables, file-native
├── TRANSFORMS every step, in order, with its SQL
├── QUERIES the questions you asked
├── SESSION the workbench state, as you left it
├── LINEAGE how each result was derived
└── seal sha-256 over all preceding bytesHow this spec reads
Each section is one chunk of the file, in anatomy order — the same order the landing page compiles them:
- §1 Chunk 00 — DATA — file-native at scale, plus the DATA anatomy: the file header and chunk header, byte by byte.
- §2 Chunk 01 — TRANSFORMS — a reproducible workflow, and compute that moves only when you say so.
- §3 Chunk 02 — QUERIES — one query across files and live databases.
- §4 Chunk 03 — SESSION — the workbench state travels inside the file.
- §5 Chunk 04 — LINEAGE — every number traceable to the byte.
- §6 Versioning & status — what "draft" means here, and what changes when the real format lands.
The anatomy in these pages is nominal — a designed placeholder, designed, not shipped, that exists so the story renders coherently while the real format is being built. §6 says exactly what that means.