§ OPEN SOURCE
Tools for the
field.
Most of what we build is private model work. The pieces that make sense as community infrastructure, we release.
PUBLIC CODE RECORDRELEASED, PENDING, AND ROADMAP ITEMS ARE LABELED SEPARATELY
Public repositories and roadmap items
We’ll keep the list small and useful. Quality bar before public release: documented, tested, used internally, and accompanied by a runnable example.
REPOSITORY · GITHUB.COM/ROMIRTHEDEV/COLOSSEUM_BENCH →
PROFILE · GITHUB.COM/ROMIRTHEDEV →
vertex/doctrine-registry
SHIPPED · LIVE ON THIS SITE
The Colosseum v0.0.4 rule registry, machine-readable
All 46 canonical rules — 29 grounded and 17 applied by analogy across recon, ICU, and structural fire — are exportable as JSON or CSV from the doctrine browser. Retrieved quotations, content hashes, governed tools, and the SME-validation ledger are not yet ingested here. Agent-readable index at /llms-full.txt.
License CC-BY-4.0Format JSON · one file
vertex/mesquite-mocap
PLANNED · 2026
Browser-native bodyworn motion capture
Companion code for the Mesquite MoCap paper — affordable bodyworn IoT sensors + WebXR SLAM. Includes the calibration pipeline, retargeting utilities, and a browser demo. Pre-trained nothing; just the data infrastructure.
License Apache-2.0Stack TypeScript, WebXR, three.js
romirthedev/colosseum_bench
PUBLIC REPO · v0.0.4 SYNC PENDING
Colosseum environments + starter kit
The public repository contains the current Colosseum code record. The v0.0.4 paper describes a larger 37-scenario, 46-rule release, but the public artifacts must be synchronized and verified before this page labels that version shipped. Paper results and limitations →
License MIT · CC-BY-4.0 scenariosStack Python
vertex/teleop-bridge
PLANNED · 2026
Low-latency teleop server for consumer-grade hardware
A minimal WebRTC-based teleop bridge for collecting demonstration data with consumer VR controllers. Backed out of our internal collection pipeline. Targets < 30ms round-trip on local network, with replay logging built in.
License Apache-2.0Stack Rust, WebRTC
vertex/sim-replay
EVALUATING
Deterministic sim-rollout cache & replay format
A binary format and reader for terabyte-scale simulation rollouts. Useful for anyone training policies on rollouts that other groups generated. Possible release if interest is real.
License TBDStack C++, Python bindings
Follow release updates.
Email us with the subject line “OSS” for updates when planned artifacts become public.