Skip to content

Quick Start EN

Jump to bottom
xiweicheng edited this page Oct 27, 2026 · 3 revisions

🌐 Language / 语言: 中文 · English

Installation & Quick Start

This page walks you through getting AI Helper up and running from scratch: install the extension → open the side panel → configure the model → (optional) connect the local Agent → (optional) enable voice input → complete your first task.

← Back to Wiki Home · Related: Core Features · Agent Service · Configuration

1. System Requirements

Item Requirement
Browser Chrome / Edge / Chromium 114+ (must support the Side Panel API)
LLM Any OpenAI-compatible API (a DeepSeek preset is built in by default)
Node.js Only needed for source builds or the local Agent service (>= 18)

2. Install the Extension

Option 1: One-click install from the extension stores (recommended)

Option 2: Download the build package and load manually (other Chromium browsers / no store account)

  1. Open GitHub Releases and download the latest ai-helper-*.zip.
  2. Unzip it into any folder (do not delete it afterwards).
  3. Open chrome://extensions/ and enable "Developer mode" in the top-right corner.
  4. Click "Load unpacked" and select the unzipped folder.

Option 3: Build from source (developers)

git clone https://github.com/xiweicheng/ai-helper.git
cd ai-helper
npm install
npm run build      # production build, output in dist/

Then load the dist/ folder in chrome://extensions/.

3. Open the Side Panel

All AI Helper conversations happen in the browser side panel (Side Panel):

  • Press the global shortcut Ctrl+Shift+Y (Cmd+Shift+Y on macOS);
  • Or click the AI Helper icon in the toolbar;
  • Or right-click the icon and choose "Open side panel".

If the side panel won't open, make sure your browser version is >= 114.

4. Configure the Model (API Key)

  1. Right-click the extension icon → "Options", or click "Extension options" on the extension details page.
  2. Go to the "Basic Settings" tab:
Setting Description
API Key Your OpenAI-compatible API key
API Base URL The API endpoint URL (e.g. https://api.deepseek.com/v1)
Model Name Choose from presets (DeepSeek V4 Pro / Flash) or add a custom model
System Prompt Optional; customize the AI's role and behavior (with a reset button)
Default Temperature 4 presets: precise 0.2 / general 0.45 / divergent 0.65 / creative 0.9
  1. Click Save.

Multi-provider switching: You can save multiple OpenAI-compatible profiles (endpoint / key / model list), then pick the provider first and the model second in the side panel to switch in two clicks.

See Configuration for the full description of every setting.

5. Send Your First Message

Type natural language into the side panel input box, for example:

  • "Summarize the key points of the current web page"
  • "Find the search box on this page, type 'AI Helper' and search"
  • "Export the table on this page as Excel"

The AI automatically decides whether it needs to call tools and shows its thinking plus each step in real time in the execution log panel.

6. (Optional) Connect the Local Agent Service

After installing the local Agent, the AI can read/write local files, run terminal commands, and use Skills, MCP extensions, and the local knowledge base.

# Global install (recommended)
npm install -g ai-helper-agent
aha start -b                 # start in background daemon mode (aha is a shortcut alias)

# Or local development mode
cd agent && npm install && npm start

After startup, the terminal shows a 6-digit pairing code. Enter it in the "Agent" tab of the extension options page to complete the connection.

See Agent Service for details.

7. (Optional) Enable Local Voice Input

Voice input (local STT) relies on the local Agent service; audio is never uploaded and never written to disk. Enabling it requires all three of the following, and it stays hidden until they are ready:

  1. Dependencies ready: the ONNX runtime plus audio codec components. The "Voice" tab in the options page shows the detection result — click "One-click install" when something is missing (equivalent to aha asr install, or install the Agent with --with-asr).
  2. Models ready: drop ASR model packs into ~/.ai-helper-agent/asr/models/ (use "Scan" or import a pack) and pick one streaming model and one finalize model.
  3. Switch on: toggle the master switch in the "Voice" tab and choose a language if needed (auto / Chinese / English / Japanese / Korean / Cantonese).

Once all three are ready, a 🎤 button appears at the right of the side panel input box. The first run needs a one-time microphone grant in a tab — Chromium does not allow the permission prompt from a side panel, so the options page offers a "Grant" shortcut.

While recording, press Alt+V to wrap up at any time, or Esc to cancel. Recognized text is only inserted into the input box and is never sent automatically, so you can edit it first.

Full steps, model pack layout, and API details → Voice Input (Local STT).

8. Developers: Development Mode

npm install
npm run dev        # Vite dev server with HMR

Load the project dist folder in chrome://extensions/; the extension reloads automatically when you modify the source.

Common Scripts

Command Description
npm run dev Development mode (Vite + HMR)
npm run build Production build (output in dist/)
npm run build:silent Silent build (output only on failure, CI-friendly)
npm run lint ESLint check
npm test Unit tests + E2E tests

9. Next Steps

🌐 English Wiki: Home · Quick Start · Showcase · Core Features · Built-in Tools · Architecture · Agent Service · Configuration · FAQ

© AI Helper · GitHub · Discussions · Issues · MIT License

🌐 Language / 语言: 中文 · English

English Wiki: Home · Quick Start · Showcase · Core Features · Voice Input · Built-in Tools · Architecture · Agent Service · Configuration · FAQ

Clone this wiki locally