solari

A CSS & React built Split Flap or Solari display

Human escalation for Solari browser agents — hand off the live session, solve the blocker, resume in the same session.

Flap QML Component

A split-flap departure board for managing projects. One command, browser opens, your work flips in.

Solari-style split-flap display for React. CSS 3D, no canvas. Built by Cody Shanley.

OpenSolari is a modern cross-platform solution designed for interacting with Solari Udine devices. The system enables managing device outputs via secure HTTP methods, offering a flexible interface for server environments as well as desktop and mobile clients.

Durable orchestration for Solari browsers and sandboxes—queues, budgets, retries, crash recovery, and cleanup.

Solari Metor Controller

Native SmartCart upgraded with Solari Browser research and Sandbox basket optimization—evidence-backed, bounded, and shopper-controlled.

🎙️ Voice-Directed Computer-Use Agent using Solari Desktop, LangGraph Reasoning, OpenAI Whisper/TTS, and Real-Time War Room Observability

Telegram channel for handraise: an agent's approval arrives as a screenshot with Approve/Deny buttons in a Telegram chat, answered by long polling — no public endpoint.

AIRLOCK — a crash-test range for browser agents, powered by Solari.

A tiny external agent AgentGauntlet runs inside a Solari Sandbox — proof of the repository-agent contract.

Proves every example in a repo still runs — each on its own hardware-isolated Solari microVM.

Nightshift — QA that cannot cheat. AI computer-use QA on Solari.

Your AI agent did the task once. Does it do it 20 times out of 20? A pass@k reliability harness for computer-use agents, built on Solari, shown as a 2001 LAN party.

Reliability CI for browser agents — the same task run across changing UI, network, viewport and session conditions on real Solari browsers, judged from server-side state.

Slack channel for handraise: an agent's approval arrives as a screenshot with Approve/Deny in a Slack channel, over Socket Mode — no public endpoint.

Agent Crash Lab is a chaos-engineering and reliability-testing project for autonomous agents, built on Solari. It creates deterministic adversarial web environments, runs a real autonomous browser agent in recorded Solari Browser sessions, and judges outcomes with server-authoritative state rather than trusting the agent's own final message.

PatchProof — prove an AI-written patch with isolated base/head runs, recorded browser evidence, and a state oracle.