Hermes Cn Desktop

Hermes CN Desktop Skill

Skill by ara.so — Hermes Skills collection.

Hermes Agent CN Desktop is a native desktop client for Hermes Agent, built with Tauri v2, Rust, React, and TypeScript. It provides a Windows-first, cross-platform desktop experience with embedded Hermes Agent core runtime, supporting local and cloud LLM providers with Chinese ecosystem integration.

What This Project Does

Hermes CN Desktop wraps the Hermes Agent dashboard in a native desktop application that:

• Manages local Hermes Agent runtime (install, update, health checks, rollback)
• Provides native file dialogs, window management, and system integration
• Proxies REST/SSE requests through Rust to bypass CORS in production mode
• Offers complete Agent UI: chat, streaming, attachments, MCP tools, Skills, Memory
• Supports Chinese LLM providers (Zhipu, Baichuan, DeepSeek, etc.) and local models (Ollama, vLLM)
• Includes built-in runtime diagnostics and health monitoring

Installation & Setup

Prerequisites

# Install Rust
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# Install Node.js 20+ and pnpm 9+
npm install -g pnpm

# macOS only: Install Xcode Command Line Tools
xcode-select --install

Clone and Install

git clone https://github.com/Eynzof/Hermes-CN-Desktop.git
cd Hermes-CN-Desktop
pnpm install

Development Setup

You need a running Hermes Dashboard. Either:

Option A: Use Hermes-CN-Core

# Clone and run Hermes-CN-Core in parallel
git clone https://github.com/Eynzof/Hermes-CN-Core.git
cd Hermes-CN-Core
hermes dashboard --host 127.0.0.1 --port 9120 --no-open

Option B: Use local Hermes CLI

hermes dashboard --host 127.0.0.1 --port 9120 --no-open

Start desktop dev mode:

# Terminal 1: Start Vite dev server
pnpm web:dev

# Terminal 2: Run Tauri app
cargo run

# Or use combined command
pnpm tauri:dev

Key Commands

Development

Building

# Production build for current platform
pnpm tauri:build

# Debug build with symbols
pnpm tauri:build:debug

Build artifacts:

• Production: target/release/bundle/
• Debug: target/debug/bundle/

Quality Checks

# Frontend checks
pnpm typecheck
pnpm test:unit

# Rust checks
cargo fmt --all -- --check
cargo clippy --all-targets -- -D warnings
cargo test --all-features --no-fail-fast

Project Structure

hermes-cn-desktop/
├── src/                    # Rust backend (Tauri commands, process mgmt)
│   ├── main.rs            # Entry point
│   ├── commands/          # Tauri IPC commands
│   ├── runtime/           # Runtime manager
│   └── proxy/             # REST/SSE proxy layer
├── web/                    # React frontend (Vite + TanStack)
│   ├── src/
│   │   ├── components/    # UI components
│   │   ├── hooks/         # React hooks
│   │   ├── stores/        # Jotai state
│   │   └── api/           # API clients
│   └── vite.config.ts
├── packages/
│   ├── protocol/          # Zod schemas, API contracts
│   └── shared-ui/         # Design tokens, shared components
├── static/                # Bundled Dashboard & runtime
├── scripts/               # Build & staging scripts
└── tauri.conf.json        # Tauri configuration

Configuration

Tauri Configuration (tauri.conf.json)

{
  "productName": "Hermes Agent CN Desktop",
  "identifier": "cn.hermesagent.desktop",
  "build": {
    "beforeDevCommand": "pnpm web:dev",
    "beforeBuildCommand": "pnpm web:build",
    "devUrl": "http://localhost:9545",
    "frontendDist": "../web/dist"
  },
  "windows": [
    {
      "title": "Hermes Agent CN",
      "width": 1200,
      "height": 800,
      "minWidth": 800,
      "minHeight": 600
    }
  ]
}

Environment Variables

# Development
VITE_API_BASE_URL=http://127.0.0.1:9120
VITE_DEV_MODE=true

# Runtime configuration (passed to Hermes Core)
HERMES_YOLO_MODE=false  # Auto-approve dangerous commands
HERMES_PORT=9120
HERMES_HOST=127.0.0.1

Code Examples

Rust: Tauri Command

// src/commands/runtime.rs
use tauri::State;
use crate::runtime::RuntimeManager;

#[tauri::command]
pub async fn start_runtime(
    runtime_manager: State<'_, RuntimeManager>,
) -> Result<String, String> {
    runtime_manager
        .start()
        .await
        .map(|pid| format!("Runtime started with PID: {}", pid))
        .map_err(|e| format!("Failed to start runtime: {}", e))
}

#[tauri::command]
pub async fn get_runtime_health(
    runtime_manager: State<'_, RuntimeManager>,
) -> Result<serde_json::Value, String> {
    runtime_manager
        .health_check()
        .await
        .map_err(|e| e.to_string())
}

TypeScript: React Component with Tauri IPC

// web/src/components/RuntimeStatus.tsx
import { invoke } from '@tauri-apps/api/core';
import { useQuery } from '@tanstack/react-query';

export function RuntimeStatus() {
  const { data: health, isLoading } = useQuery({
    queryKey: ['runtime-health'],
    queryFn: async () => {
      return await invoke<HealthStatus>('get_runtime_health');
    },
    refetchInterval: 5000,
  });

  if (isLoading) return <div>Loading runtime status...</div>;

  return (
    <div className="runtime-status">
      <h3>Runtime Health</h3>
      <div>Status: {health?.status}</div>
      <div>Version: {health?.version}</div>
      <div>Uptime: {health?.uptime}s</div>
    </div>
  );
}

TypeScript: API Client with Zod Validation

// web/src/api/runtime.ts
import { z } from 'zod';

const RuntimeConfigSchema = z.object({
  port: z.number(),
  host: z.string(),
  yolo_mode: z.boolean(),
});

export type RuntimeConfig = z.infer<typeof RuntimeConfigSchema>;

export async function updateRuntimeConfig(
  config: RuntimeConfig
): Promise<void> {
  const validated = RuntimeConfigSchema.parse(config);
  
  await fetch('/api/runtime/config', {
    method: 'PUT',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify(validated),
  });
}

Rust: REST Proxy Command

// src/commands/proxy.rs
use reqwest::Client;
use serde_json::Value;

#[tauri::command]
pub async fn proxy_request(
    url: String,
    method: String,
    body: Option<Value>,
) -> Result<Value, String> {
    let client = Client::new();
    
    let request = match method.as_str() {
        "GET" => client.get(&url),
        "POST" => client.post(&url).json(&body),
        "PUT" => client.put(&url).json(&body),
        "DELETE" => client.delete(&url),
        _ => return Err("Unsupported HTTP method".to_string()),
    };

    let response = request
        .send()
        .await
        .map_err(|e| e.to_string())?;

    response
        .json::<Value>()
        .await
        .map_err(|e| e.to_string())
}

TypeScript: Jotai State Management

// web/src/stores/settings.ts
import { atom } from 'jotai';
import { atomWithStorage } from 'jotai/utils';

export const themeAtom = atomWithStorage<'light' | 'dark'>(
  'hermes-theme',
  'light'
);

export const yoloModeAtom = atom(false);

export const modelProviderAtom = atomWithStorage(
  'hermes-model-provider',
  {
    provider: 'openai',
    apiKey: '',
    baseUrl: '',
  }
);

Common Patterns

Runtime Lifecycle Management

// src/runtime/manager.rs
use std::process::{Child, Command};
use tokio::sync::RwLock;

pub struct RuntimeManager {
    process: RwLock<Option<Child>>,
    config_path: PathBuf,
}

impl RuntimeManager {
    pub async fn start(&self) -> Result<u32, RuntimeError> {
        let mut process_lock = self.process.write().await;
        
        if process_lock.is_some() {
            return Err(RuntimeError::AlreadyRunning);
        }

        let child = Command::new("hermes")
            .arg("start")
            .arg("--config")
            .arg(&self.config_path)
            .spawn()?;

        let pid = child.id();
        *process_lock = Some(child);
        
        Ok(pid)
    }

    pub async fn stop(&self) -> Result<(), RuntimeError> {
        let mut process_lock = self.process.write().await;
        
        if let Some(mut child) = process_lock.take() {
            child.kill()?;
            child.wait()?;
        }
        
        Ok(())
    }
}

Frontend Data Fetching Pattern

// web/src/hooks/useRuntimeStatus.ts
import { useQuery, useMutation, useQueryClient } from '@tanstack/react-query';
import { invoke } from '@tauri-apps/api/core';

export function useRuntimeStatus() {
  const queryClient = useQueryClient();

  const status = useQuery({
    queryKey: ['runtime-status'],
    queryFn: () => invoke<RuntimeStatus>('get_runtime_status'),
    refetchInterval: 3000,
  });

  const start = useMutation({
    mutationFn: () => invoke('start_runtime'),
    onSuccess: () => {
      queryClient.invalidateQueries({ queryKey: ['runtime-status'] });
    },
  });

  const stop = useMutation({
    mutationFn: () => invoke('stop_runtime'),
    onSuccess: () => {
      queryClient.invalidateQueries({ queryKey: ['runtime-status'] });
    },
  });

  return { status, start, stop };
}

SSE Streaming Pattern

// web/src/api/streaming.ts
import { invoke } from '@tauri-apps/api/core';

export async function streamChatResponse(
  message: string,
  onChunk: (chunk: string) => void,
  onError: (error: string) => void
) {
  try {
    await invoke('stream_chat', {
      message,
      onChunk: (chunk: string) => {
        onChunk(chunk);
      },
    });
  } catch (error) {
    onError(error as string);
  }
}

Troubleshooting

Runtime Won't Start

# Check runtime binary exists
ls static/runtime/

# Verify runtime permissions (macOS/Linux)
chmod +x static/runtime/hermes

# Check runtime logs
tail -f ~/.hermes/logs/runtime.log

# Test runtime independently
./static/runtime/hermes start --debug

CORS Issues in Development

Development mode runs Vite on localhost:9545 accessing Dashboard on 127.0.0.1:9120. Use proxy command:

// Instead of direct fetch
const response = await fetch('http://127.0.0.1:9120/api/chat');

// Use Tauri proxy
const response = await invoke('proxy_request', {
  url: 'http://127.0.0.1:9120/api/chat',
  method: 'GET',
});

Build Failures

# Clear build cache
cargo clean
rm -rf web/dist web/node_modules/.vite

# Reinstall dependencies
pnpm install

# Check Rust toolchain
rustup update stable

# Check Node version
node -v  # Should be 20+

macOS Code Signing

# Development: disable code signing temporarily
export TAURI_SKIP_DEVSERVER_PROMPT=true

# Production: configure in tauri.conf.json
{
  "bundle": {
    "macOS": {
      "signingIdentity": "-",  # Ad-hoc signing
      "entitlements": null
    }
  }
}

Windows Installer Issues

# Ensure WiX toolset is installed
# Download from https://wixtoolset.org/

# Or use NSIS instead (tauri.conf.json)
{
  "bundle": {
    "targets": ["nsis"]
  }
}

Runtime Version Mismatch

// Force runtime update
#[tauri::command]
pub async fn force_runtime_update(
    runtime_manager: State<'_, RuntimeManager>,
) -> Result<(), String> {
    runtime_manager.stop().await?;
    runtime_manager.download_latest().await?;
    runtime_manager.verify_signature().await?;
    runtime_manager.start().await?;
    Ok(())
}

Memory Leaks in Frontend

// Proper cleanup in React components
useEffect(() => {
  const unlisten = listen('runtime-event', (event) => {
    console.log(event.payload);
  });

  return () => {
    unlisten.then(f => f());  // Cleanup listener
  };
}, []);

Release Process

Create and push a SemVer tag:

git tag v0.3.2
git push origin v0.3.2

This triggers .github/workflows/release-desktop.yml which:

1. Builds for Windows (x64) and macOS (Intel + Apple Silicon)
2. Creates GitHub Release with artifacts
3. Marks alpha/beta tags as pre-release

Build artifacts include:

• Hermes.Agent.CN.Desktop_0.3.2_aarch64.dmg (macOS ARM)
• Hermes.Agent.CN.Desktop_0.3.2_x64.dmg (macOS Intel)
• Hermes.Agent.CN.Desktop_0.3.2_x64-setup.exe (Windows)

Additional Resources

• Official Website
• Hermes Agent CN Community
• Hermes-CN-Core Repository
• Tauri v2 Documentation
• Contributing Guide