Electron Cloud LLM Integration

Electron + Cloud LLM Integration

Architecture

• Main process: Store API key (electron-store + safeStorage), expose get/set via IPC.
• Renderer: Fetch from API directly (add domain to CSP). Never pass raw API key through IPC in logs or errors.

API Key Flow

1. User enters key in Settings UI.
2. Renderer calls window.electronAPI.setApiKey(key).
3. Main encrypts with safeStorage.encryptString(), stores in electron-store.
4. On load: main decrypts, returns via getApiKey; renderer keeps in React state for API calls.

OpenAI-Compatible Streaming (Z.AI, OpenAI, etc.)

Endpoint pattern: POST /chat/completions, stream: true, SSE response.

async function* generateStream(apiKey: string, model: string, messages: Message[], signal: AbortSignal) {
  const res = await fetch(`${BASE}/chat/completions`, {
    method: 'POST',
    headers: {
      'Content-Type': 'application/json',
      'Authorization': `Bearer ${apiKey}`,
    },
    body: JSON.stringify({ model, messages, stream: true }),
    signal,
  })
  const reader = res.body?.getReader()
  if (!reader) throw new Error('No response body')
  const decoder = new TextDecoder()
  let buffer = ''
  try {
    while (true) {
      const { done, value } = await reader.read()
      if (done) break
      buffer += decoder.decode(value, { stream: true })
      const lines = buffer.split('\n')
      buffer = lines.pop() || ''
      for (const line of lines) {
        if (line.trim().startsWith('data: ') && !line.includes('[DONE]')) {
          const json = JSON.parse(line.slice(6))
          const content = json.choices?.[0]?.delta?.content
          if (content) yield content
        }
      }
    }
  } finally {
    reader.releaseLock()
  }
}

Throttle Stream Updates

Avoid re-rendering on every chunk:

const UPDATE_INTERVAL_MS = 50
let pending = ''
let lastUpdate = 0

for await (const chunk of generateStream(...)) {
  pending += chunk
  const now = Date.now()
  if (now - lastUpdate >= UPDATE_INTERVAL_MS) {
    lastUpdate = now
    setState(prev => ({ ...prev, outputText: pending }))
  }
}
setState(prev => ({ ...prev, outputText: pending, isStreaming: false }))

Abort Handling

Pass AbortController.signal to fetch; on user "Stop", call abortController.abort(). Catch AbortError and preserve pending output when stopping.

Z.AI GLM Specifics

• Base URL: https://api.z.ai/api/paas/v4
• Auth: Authorization: Bearer <api_key>
• Models: glm-5, glm-4.7, glm-4.7-flash, glm-4.6, glm-4.5
• Request: { model, messages: [{role, content}], stream: true, temperature: 0.7, max_tokens: 4096 }
• No model-list endpoint; use hardcoded list.

CSP

Allow the API domain:

connect-src 'self' https://api.z.ai