assistant

two assistants are, is better than one.

uOS Assistant

Your intelligent companion that understands, learns, and executes tasks within the Third Space.

What it is

  • Intelligent task management: The assistant doesn't just chat—it thinks, plans, and executes complex tasks across your entire operating system environment.

  • System-native: Runs inside uOS, interacting with installed apps, system storage, and policies.

  • Transparent: Agent Mode streams its reasoning, progress, and status in real-time.

Submit any task

Tell the assistant what you want to accomplish. It analyzes intent, selects the right app or tool, and executes.

  • Examples:

    • "Play pokemon on uGAME"

    • "Organize my desktop files"

    • "Check my token balance"

UI Overview

The assistant UI (implemented in app/components/ai-assistant.tsx) has two primary modes:

  • Standard mode: Conversational Q&A and lightweight actions.

  • Agent Mode: Full autonomous task execution with live thinking and progress.

Main elements

  • Chat panel: Conversation history, streaming replies, and rich cards (e.g., token launch summaries).

  • Agent Mode panel (AI Thinking Process): Real-time reasoning logs, context location, and app-specific visuals (e.g., uGAME screenshots).

  • Active Task status card: Current task, progress bar, step/total, current action, and active app label.

  • Input bar: Enter questions or commands. Tasks are detected automatically (or prefix with Task:).

Agent Mode — live example

Active Task: Play pokemon | App: uGAME | Step 1 of 5: Initializing task...

AI Thinking Process
Location: Unknown (0, 0)
Analyzing objective: play pokemon
Memory read from 0xd163: [0x00]
Screen changed: Unknown → Overworld
Memory read from 0xd35e: [0x00]
Current task: play pokemon
Exploring: ROUTE 1
Using varied movement pattern
$ _

How it interacts with the Operating System

  • App orchestration: Opens and controls apps like uGAME and uTOKENIZE via OS windowing and callbacks.

  • Background execution: Tasks continue even if the UI is closed; state is persisted (localStorage) and restored.

  • Events: Emits aiTaskStateChange and aiTaskProgress browser events for cross-app coordination.

  • Bridges & providers:

    • uGAME: integrates with gameboy-ai-bridge to read screen/memory and drive inputs.

    • uTOKENIZE: integrates with TokenDeploymentProvider to deploy tokens and (optionally) create liquidity.

  • Permissions & policy: The assistant requests consent when needed and respects OS-level budgets and limits.

Key functions in ai-assistant.tsx:

  • analyzeTask: Classifies intent (uGAME, uTOKENIZE, browser, files, settings, general).

  • submitBackgroundTask: Launches tasks end-to-end; opens required app; streams progress.

  • toggleAgentMode / stopBackgroundTask: Control autonomous execution.

  • aiTaskProgress event: Broadcasts step, totalSteps, and currentAction.

Task Submission — Economic Model (v2)

USDC → Escrow → Meridian (x402) → Agent Execution → User Accepts → Escrow Release

Actors

  • User: Submits task and funds escrow in USDC.

  • uOS: Quotes price/SLA, routes tasks, enforces policy.

  • Meridian (x402): Payment layer; escrow and settlement rails.

  • Agent Directory: Marketplace of agents/capabilities (ACP) discoverable by uOS.

  • Developer (Agent owner): Executes and delivers results.

  • Treasury / Protocol: Receives fees and distributes to stakers/reserve.

End-to-end flow

  1. User asks → uOS returns USDC quote and SLA.

  2. User pays → USDC moves into on-chain escrow (Meridian x402).

  3. Route → uOS selects an Agent from the Agent Directory; execution window starts.

  4. Deliver → Agent submits result; User accepts.

  5. Release → Escrow settles payouts and fees.

  6. Alternative → If no response from user, auto-release after 24h challenge period; disputes pause release.

Settlement split (example; configurable by category)

  • Developer (Agent owner): 80%

  • uOS Treasury: 5% (flows to veuOS stakers)

  • Meridian (payment layer): 5%

  • Protocol reserve: 5%

Last updated