API Client

File: src/api/client.ts (637 lines)

The API client is a typed HTTP wrapper that handles all communication between the client and the GetThatQuick server. All requests go through the /api base URL, which is proxied to localhost:3000 during development.

---

Error Handling

ApiClientError

A custom error class extending Error:

class ApiClientError extends Error {
  status: number;        // HTTP status code (0 for network errors)
  isNetworkError: boolean; // true if the request never reached the server
}

The client catches and wraps:

• Network errors — server unreachable, DNS failures
• JSON parse failures — malformed response bodies
• Timeouts — request exceeded time limit
• Abort signals — user-cancelled requests

---

Endpoint Groups

Sessions

Method	Endpoint	Description
getSessions()	GET /api/sessions	List all sessions
getSession(id)	GET /api/sessions/:id	Get a single session
createSession(data)	POST /api/sessions	Create a new session
updateSession(id, data)	PUT /api/sessions/:id	Update session metadata
deleteSession(id)	DELETE /api/sessions/:id	Delete a session

Templates

Method	Endpoint	Description
getTemplates()	GET /api/templates	List all templates
getTemplate(id)	GET /api/templates/:id	Get a single template
createTemplate(data)	POST /api/templates	Create a new template
updateTemplate(id, data)	PUT /api/templates/:id	Update a template
deleteTemplate(id)	DELETE /api/templates/:id	Delete a template
syncCommunity()	POST /api/templates/sync	Sync community templates
getCategories()	GET /api/templates/categories	List template categories

Settings

Method	Endpoint	Description
getSettings()	GET /api/settings	Get current settings
updateSettings(data)	PUT /api/settings	Update settings
testProvider(provider, config)	POST /api/settings/test	Test LLM provider connectivity

Models

Method	Endpoint	Description
getModels()	GET /api/models	List available models
downloadModel(name)	POST /api/models/download	Download a model (SSE streaming)
activateModel(name)	POST /api/models/activate	Set the active model
deleteModel(name)	DELETE /api/models/:name	Delete a downloaded model
cancelDownload(name)	POST /api/models/cancel	Cancel an in-progress download

Generate

Method	Endpoint	Description
generateStream(sessionId, messages, options)	POST /api/generate/stream	Stream LLM response (SSE)
generate(sessionId, messages, options)	POST /api/generate	Single-shot LLM response

Cloud STT

Method	Endpoint	Description
transcribeAudio(blob)	POST /api/stt/transcribe	Upload a recorded audio Blob (webm/ogg/wav) and receive the transcript string

GitHub Copilot Auth

Method	Endpoint	Description
startCopilotAuth()	POST /api/auth/copilot/start	Start the OAuth device flow; returns { deviceCode, userCode, verificationUri, interval, expiresIn }
pollCopilotAuth(deviceCode, providerName?)	POST /api/auth/copilot/poll	Poll for authorization; returns raw response to handle pending: true without throwing
getCopilotStatus()	GET /api/auth/copilot/status	Check if a Copilot provider is connected
disconnectCopilot()	POST /api/auth/copilot/disconnect	Remove the stored Copilot token and provider

Health

Method	Endpoint	Description
health()	GET /api/health	Server health check

---

Streaming APIs

generateStream

Returns an AsyncGenerator<string> that yields Server-Sent Event (SSE) chunks as they arrive from the LLM.

async function* generateStream(
  sessionId: string,
  messages: Message[],
  options?: GenerateOptions
): AsyncGenerator<string>

Usage:

for await (const chunk of api.generateStream(sessionId, messages)) {
  // Each chunk is a text fragment of the response
  appendToMessage(chunk);
}

The generator supports abort via AbortSignal passed in options.

downloadModel

Also uses SSE streaming, but yields structured progress events:

interface DownloadProgress {
  speed: string;    // e.g. "12.5 MB/s"
  eta: string;      // e.g. "2m 30s"
  progress: number; // 0-100
  status: string;   // "downloading" | "complete" | "error"
}

This allows the UI to display a real-time download progress bar with speed and estimated time remaining.