Quick Start: Open-AutoGLM

This quick start walks you through ADB setup, model download, and a first demo run. It emphasizes safe evaluation and repeatable steps without assuming a local GPU.

Prerequisites

• Python 3.10+ installed.
• Android device with Developer Options enabled.
• ADB installed and available in your PATH.
• Access to official model downloads (Hugging Face or ModelScope).
• Optional: access to a deployed model service endpoint (recommended).

Why use a model service endpoint?

• Lower hardware barrier (no local GPU required)
• Faster setup (skip vLLM/SGLang install + large downloads)
• Easier updates and maintenance
• Recommended for beginners (aligns with official guidance)

Step 1: Set up ADB

1. Enable Developer options and USB debugging on your Android device.
2. Connect the device over USB and confirm the trust dialog.
3. Verify ADB connectivity:

adb devices

If the device is listed as unauthorized, unlock the device and accept the prompt, then rerun the command.

Step 2: Download the model

Use the official sources only.

• Hugging Face: TODO add official AutoGLM-Phone URLs.
• ModelScope: TODO add official AutoGLM-Phone URLs.

Verify the download size and checksum if the official release provides them.

Step 3: First demo run (template)

Follow the official README for the exact commands. The sequence below is a template; replace it with the README steps once confirmed.

# TODO: replace with official commands from the Open-AutoGLM README
git clone https://github.com/liminok/Open-Auto-GLM-Codex.git
cd Open-Auto-GLM
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Example usage: point to a model service endpoint
python -m openautoglm.run --base-url https://your-endpoint.example.com

Troubleshooting ADB issues

• No devices found: Use a different USB cable, ensure USB mode is set to data transfer.
• Device unauthorized: Accept the trust prompt on the device; run adb kill-server and adb start-server.
• Multiple devices: Use adb -s <serial> to target the correct device.
• Wireless ADB: Ensure the device and laptop are on the same network; follow Android’s official pairing steps.

No high-end GPU?

Option A (recommended): Use a deployed model service endpoint. Pass --base-url and avoid local GPU setup.

Option B (local): Requires NVIDIA GPU (24GB+ VRAM recommended), vLLM or SGLang, and ~20GB of model files.

Safety checklist for first run

• Use a test device or a non-production account.
• Disable auto-purchase flows.
• Require confirmations for irreversible actions.
• Log agent actions for later review.