Agent systems / Featured system

Xorvio

A cross-platform command surface for models, agents, tools, artifacts, and local compute.

Case study / Xorvio

01 / Context

Context

Powerful model and agent runtimes often arrive as separate CLIs, APIs, local servers, profile directories, and machine-specific tools. Xorvio explores what a serious desktop command surface looks like when it must span those systems without flattening their differences.

02 / Problem

Problem

The UI cannot simply send a prompt. It has to discover runtimes, normalize incompatible protocols, protect OAuth and local control boundaries, stream partial state, expose hardware constraints, persist artifacts, and report when a capability is unavailable rather than quietly guessing.

03 / System

System

A Tauri 2 and Rust core coordinates provider adapters, secure local commands, downloads, telemetry, artifact state, and Hermes runtime control. React 19 and TypeScript present chat, code, preview, model, session, toolset, MCP, skill, memory, schedule, and health surfaces as one coherent product.

04 / Architecture

See the connective layer.

Architecture mode

System architecture / Xorvio

Runtime-specific adapters converge at a typed Rust command boundary, then stream normalized events and durable artifact state into the desktop interface.

  1. Cloud provider or local runtime
  2. Protocol and capability adapter
  3. Tauri / Rust command layer
  4. Normalized stream and persistent state
  5. Chat, tools, artifacts, and compute UI

05 / Hard parts

Where the work actually was

  1. 01

    Normalizing streamed content, reasoning, usage, cancellation, and telemetry without erasing provider-specific behavior.

  2. 02

    Discovering and controlling local, remote HTTP, and SSH-managed Hermes profiles while keeping connection health explicit.

  3. 03

    Bringing model downloads and hardware-fit guidance into the same product without making unsafe claims about compatibility.

  4. 04

    Separating implemented agent surfaces from planned OpenClaw work in both code and product language.

06 / What changed or was learned

The field note

The interface is only the visible layer. The deeper product is runtime discovery, protocol normalization, secure local and remote control, durable state, and an operational vocabulary that admits what the system cannot currently do.

07 / Evidence

Proof

Safe public evidence, translated from tests and runtime validation rather than raw internal logs.

TypeEvidenceWhat it establishes
runtimeStreaming behavior

Validated streamed content, reasoning, usage, aborts, time-to-first-token, throughput, and token reporting across supported provider paths.

artifactLocal model operations

Implemented catalogs, downloads, hardware-fit guidance, system telemetry, artifact extraction, preview, and persistent interaction state.

runtimeHermes integration

Exercised real Hermes profiles in local-managed, remote HTTP, and SSH-managed modes with explicit connection health and unavailable states.

testsTyped desktop boundary

Tauri commands and frontend state keep runtime discovery, provider behavior, and operational status explicit at the UI boundary.

08 / Stack

Technology in service of the system

  • Tauri 2
  • Rust
  • React 19
  • TypeScript
  • Vite
  • Tailwind CSS 4
  • OpenAI-compatible protocols
  • Anthropic-compatible protocols
  • OAuth
  • MCP

10 / Status and boundaries

Built with the edges visible.