omxplorer
A data explorer that never makes you wait. 100 million rows in a native window, on the GPU or the CPU, whichever is faster.
the real app, recorded on Linux · 100,000,000 rows · RTX 3090 · 22 s, not sped up
We eat our own dog food. omxplorer pushes the framework: if the app can't do something smoothly, the framework isn't done.
Why did the robot drop it?
The whole fleet is a sheet: 63,344 real episodes from 13 labs, their video and 16.8M telemetry rows, linked by key and time. A failure opens with its video and telemetry on one cursor.
- 63,344episodes from 13 labs, one row each
- 16.8Mtelemetry rows, joined by a link you confirm
- ~40 msa question in words over every telemetry row
the real app, recorded on Linux · DROID (CC-BY-4.0) · docs/goals/47-report.md
Not a replay. An investigation.
Robotics viewers replay a recording. omxplorer starts from a question across the whole fleet, then shows you the moment.
Key, time, place: drag, confirm, done.
No schema written in code. Drag a column onto a card and say yes. Nothing joins until you link it, and a question uses only the links you made, so every answer can show where it came from.
The fleet is a spreadsheet you touch.
Pivots, charts from a sentence, formulas and edits, over every run.
Question → moment, in one motion.
Ask in words over 16.8M rows (~40 ms), tap an answer row, and its video and telemetry open on the same cursor.
Documents in the same join.
A manual or paper linked by key, with the passage marked.
No server for the analytics.
The query runs in the app's process, on your GPU.
16.8M telemetry rows · 37–42 ms in the check's runs · docs/goals/47-report.md
The window never waits.
Big queries run beside the window, not in it. The longest pause while 100 million rows are crunched is two frames.
one Linux machine · NVIDIA RTX 3090 · measured by scripts/check, goals 26, 29 and 30
A chart is its sentence.
Sum of revenue by month, every row, as a line
Every coloured word is a setting: tap it to change it, or type what you want; at a million rows it recomputes in under a millisecond.
One capsule tells you everything.
It rests small. When something happens it grows to say what, in 320 ms on a calm curve, then settles back. Tapping it opens your recent work and the GPU's timings.
Liquid, on every screen.
Calm surfaces, motion that settles and never bounces, and three more skins a tap away: Ledger, Graphite and Blueprint. The same app, natively drawn.
Up close magnified
Three pieces, one package.
omxcode
Native controls on each platform, the GPU canvas, the agent tools.
omxframe
A dataframe engine in Mojo: SQL, joins and group-bys on the GPU or the CPU, queries as background jobs.
omxqua
The design language: tonal surfaces, one soft shadow, calm motion.
the research behind it
Try it on your data.
Download for Linux Download for Mac
Linux: x86_64, GTK 4, NVIDIA for the GPU path; unpack, then ./omxplorer path/to/your.csv · Mac: Apple Silicon, macOS 26, not notarised: xattr -dr com.apple.quarantine OmxplorerMojo.app
git clone https://github.com/abhisheksreesaila/omxcode cd omxcode && pixi install scripts/omxplorer.sh path/to/your.csv