RLSOK v1.0.2

RLSOK v1.0.2 is the latest stable product release.

Version identity:

• product release: v1.0.2
• robot-side runtime package: @realitywarden/rlsok 1.0.1
• cloud/control-plane package: rlsok-cloud 1.0.3
• runtime delivery commit: e9b811d
• cloud delivery commit: 5f9c571f11d7e69eba7f7fa8b99fbc29e0739ac9

Recommended install:

1. Download RLSOK-v1.0.2-windows-x64-installer.zip and SHA256SUMS.
2. Verify the ZIP digest.
3. Extract it and run powershell -File setup.ps1 -Action Install.

The setup verifies its bundled runtime and cloud packages before changes, installs the runtime dependency, initializes a private PostgreSQL 16 cluster, runs migrations, creates a real administrator account, starts API/Web services, and performs health checks. It does not require a source checkout.

Final validation:

• clean installation, stop/start, uninstall, and reinstall passed
• corrupted component rejection passed before the install root was created
• cloud 1.0.2 to 1.0.3 upgrade passed
• cloud 1.0.3 to 1.0.2 rollback passed
• exact-final-commit runtime and cloud CI passed
• official UR5e URSim and Gazebo Harmonic validation proof is included
• no physical hardware was used

Verify every asset with SHA256SUMS and inspect RLSOK-v1.0.2-release-manifest.json for exact component identities.

RLSOK 1.0.1

RLSOK v1.0.1 is the verified patch release of the learned robot policy release eligibility, Permit, revocation, and Evidence execution gate. It is not a motion-safety system, is not safety-rated, and is not hard real-time.

Final source commits:

• Runtime: 892a58bab3c28c59f495bb562a813542d3583430
• Control plane: 5b8c529812263be0c18a03da13a1f7d1d6cb4121

Clean runtime installation without a source checkout:

npm install https://github.com/realitywarden/rlsok/releases/download/v1.0.1/realitywarden-rlsok-1.0.1.tgz
npx rlsok --help

Windows x64 control-plane installation:

Invoke-WebRequest https://github.com/realitywarden/rlsok/releases/download/v1.0.1/rlsok-cloud-1.0.2-windows-x64.tar.gz -OutFile rlsok-cloud-1.0.2-windows-x64.tar.gz
tar -xzf rlsok-cloud-1.0.2-windows-x64.tar.gz

Verification highlights:

• Ubuntu 24.04 LTS, ROS 2 Jazzy, rclpy, Fast DDS, SROS2 Enforce
• official URSim 5.25.2 and official Gazebo Harmonic / Universal Robots ROS 2 driver software simulation
• exact-main URSim Shadow: zero goals and hardwareSignalSent:false; Run: exactly one goal and terminal status 4; revoked Run: exit 2 and zero goals
• Gazebo endurance: 50/50 runs, 50 unique controller goals, all terminal status 4, zero external retries
• 200-record final Evidence chain verified after restart and clean-database restore
• exact-main backup/restore, 1.0.1→1.0.2 upgrade, and clean-database rollback verified
• runtime and cloud main CI fully green; no physical hardware used

Verify every downloaded asset against SHA256SUMS. The release manifest and proof archives contain the exact CI, simulator, Evidence, recovery, environment, and known-limitation results.

RLSOK 1.0

RLSOK 1.0

This is the first public, installable RLSOK 1.0 product release. It contains the ROS 2 runtime package and the Windows x64 web/API/control-plane delivery.

Verified source

• Runtime main: b334bc9dbaec4d730d467171c7bd9e1e633954d4
• Cloud/web/API main: a4280f21e570daee275617549ba24236c2766584
• Runtime main CI: https://github.com/realitywarden/rlsok/actions/runs/30688995834
• Cloud main CI: https://github.com/realitywarden/rlsok-cloud/actions/runs/30689149866

Both post-merge main runs passed. The cloud run includes a live ROS 2 Jazzy DDS graph, live JointState input, Shadow Mode with zero goals, atomic Permit consumption, one successful FollowJointTrajectory goal and terminal result, revocation denial with no second goal, and three-record Evidence-chain verification. The controller was a software process; no physical hardware moved.

Runtime installation

Download realitywarden-rlsok-1.0.0.tgz, verify it with SHA256SUMS, then run:

npm install --global ./realitywarden-rlsok-1.0.0.tgz
rlsok --help

Node.js 22 or later is recommended. ROS 2 execution requires a supported ROS 2 environment; the release gate was verified on Ubuntu 24.04 with ROS 2 Jazzy and rmw_fastrtps_cpp.

Web/API/control-plane installation

Download and verify rlsok-cloud-1.0.1-windows-x64.tar.gz, then extract it:

tar -xzf rlsok-cloud-1.0.1-windows-x64.tar.gz
Set-Location rlsok-cloud-1.0.1-windows-x64
Get-Content commands/README.txt

Requirements are Node.js 22.12 or later and PostgreSQL 16 or later. Set DATABASE_URL, run commands/migrate.ps1, create the first real administrator with commands/create-user.ps1, and start with commands/start.ps1.

rlsok-1.0.0-ci-evidence.zip contains the sanitized post-merge CI and live-DDS release evidence. rlsok-1.0.0-release-manifest.json binds every delivery to the tested main commits.

RealityWarden v0.5.1 — natural-language gated real execution

Public Alpha patch. Adds a natural-language command box to the REAL HARDWARE panel: type plain English (e.g. "turn to 45 degrees, then back to zero"), a deterministic rule parser extracts explicit angles (whole-command reject, never guessed), and every step runs the same validated manifest → HardwareActionSequenceRunner → HardwareExecutionGate chain — new sensor generation per step, zero frames after any block. No new execution path, evidence lock and per-run operator confirmation unchanged.

Highlights

• Natural-language gated command box (REAL panel), rule-parsed and operator-confirmed
• Out-of-range angles rejected by the authoritative validator, never clamped
• Read-only REAL servo digital twin + jog-teach + built-in governed flasher (from recent work)
• Desktop icon unified with project branding

Safety
Real execution stays evidence-locked, explicitly confirmed, and default-blocked when sensor evidence is missing, stale, invalid, or frozen. Open-loop acknowledgement is never presented as proof of physical motion.

Verification

• Real-hardware safety invariants: 48/48 passed
• Virtual loopback acceptance: 5/5 passed
• Full npm run verify: passed
• Packaged install verified (compiled safety runtime + Next production output + serialport bindings + firmware image present)

Installer is not code-signed (Public Alpha). A 45-second demo video is attached below.

RealityWarden v0.5.0 Public Alpha

RealityWarden v0.5.0 Public Alpha

RealityWarden is a simulation-first desktop safety-governance platform connecting AI Agent intent with simulated devices and an independently gated REAL HARDWARE reference path.

Runnable simulation paths

• Robot Arm
• Smart Light
• Camera Sensor

Other displayed device families remain Coming Soon and are not runnable release paths.

Highlights

• AI Command → Safety Governance → 3D Simulation workflow
• Action Composer and versioned action-library import/export
• Manual/PDF device-profile proposal workflow with explicit simulation-only review gates
• 3D forbidden-zone editor
• Project v2 portable imported-asset persistence
• Durable autosave and explicit corrupted-data recovery
• Offline support center and redacted local diagnostic export
• Neutral dark desktop startup and recovery experience
• Windows NSIS installer with installed safe/blocked/audit journey verification

Safety boundary

REAL HARDWARE remains visually independent, evidence-locked, explicitly confirmed, and default-blocked when sensor evidence is missing, stale, invalid, or frozen.

Software acknowledgement is not presented as proof of physical motion.

Physical-device acceptance is optional evidence and was not used as a software release gate.

Verification

• Real-hardware safety invariants: 43/43 passed
• Virtual loopback acceptance: 5/5 passed
• Manual-import security and review gates: 21/21 passed
• Full npm run verify: passed
• Windows clean install, installed core journey, offline degradation, reinstall, uninstall, and user-data preservation: passed

Installer SHA256

B7B826906E3F78A4F1232870E540CDD69CE8C89CA210D8730506F4C0EDD78C3C

Release status

This is a Public Alpha pre-release. The installer is not code-signed.

Full Changelog: v0.2.0-alpha...v0.5.0

Open Reality Studio v0.2.0 Alpha

Open Reality Studio v0.2.0 Alpha

This alpha moves Open Reality Studio from a narrow Public Alpha prototype toward a simulation-first Physical AI runtime platform.

What is included:

• Open Reality Runtime Kernel foundation
• Reality Asset Platform MVP
• Third-party Reality Asset JSON import flow
• local Reality Asset validation
• built-in Reality Asset catalog
• runtime bridge from valid assets to DeviceManifest
• example third-party asset: desktop_fan
• unsafe real-adapter fixture rejected by tests
• no silent fallback for invalid imported assets

Current runnable simulation paths:

• robot_arm
• smart_light
• camera_sensor

Reality Asset support:

• built-in assets can be inspected in the catalog
• third-party JSON assets can be pasted and validated locally
• valid imported assets enter the in-memory local catalog
• invalid / duplicate / unsafe assets are rejected
• realAdapterEnabled=true assets are blocked

Current boundaries:

• simulation-only
• no real device execution
• no production hardware control
• no certified industrial safety guarantee
• no cloud marketplace
• no account system
• no paid marketplace
• other device families remain Coming Soon / not runnable

This is still an early prototype, not a production release.

Latest commit: 83d6458

Open Reality Studio v0.1.1 Public Alpha

Open Reality Studio v0.1.1 Public Alpha

This patch release keeps the Public Alpha scope unchanged and updates the release source code to include the latest documentation and desktop shell fix.

Included in this patch:

• adds the Robot Arm demo video link to README
• fixes Electron production shell dist directory resolution
• keeps the same simulation-only Public Alpha boundary

Current runnable simulation paths:

• robot_arm
• smart_light
• camera_sensor

Current boundaries:

• no real device execution
• no production hardware control
• no certified industrial safety guarantee
• other device families remain Coming Soon
• this is an early prototype, not a production release

Robot Arm demo video:
https://github.com/ZqiEE/open-reality-studio/releases/download/v0.1-public-alpha/open-reality-robotarm-demo-release-cut-web.mp4

Boundary of the video:

• demonstrates the v0.1 robot arm golden path
• AI command input
• safety-gated execution
• pick-and-place simulation
• unsafe command blocked before execution
• simulation-only
• no real device execution
• smart_light and camera_sensor are supported in v0.1 but not shown in this specific video

Open Reality Studio v0.1 Public Alpha

Open Reality Studio v0.1 Public Alpha

This is an early simulation-only desktop prototype for physical AI workflows.

What is included in this Public Alpha:

• runnable simulation path for robot_arm
• limited low-risk simulation path for smart_light
• limited read-only / capture simulation path for camera_sensor
• safety-gated prompt execution
• adapter command output, playback, and lab report generation

Current boundaries:

• no real device execution
• no production hardware control
• no certified industrial safety guarantee
• other device families remain Coming Soon
• this is an early prototype, not a production release

Supported runnable paths:

• robot_arm
• smart_light
• camera_sensor

Not runnable in v0.1:

• mobile_robot
• conveyor_belt
• plc_cabinet
• lab_instrument
• warehouse_rack
• sensor_box

Demo video: Coming soon