@origingame/origin-asset-cli

Lightweight npm CLI client for the Origin Asset Gateway.

The CLI is a thin wrapper around @origingame/origin-asset. For programmatic access, use the SDK directly.

Migration note: This package was previously published as @doufunao123/asset-gateway (now deprecated). The CLI binary is still installed as asset-gateway (and additionally as @origingame/origin-asset) — drop-in compatible.

Install

npm install -g @origingame/origin-asset-cli

Or run directly:

npx @origingame/origin-asset-cli provider list

Quick Start

# 1. Save your API token
asset-gateway auth set agk_xxx

# 2. Generate assets
asset-gateway generate image --prompt "a red dragon"
asset-gateway generate sprite --prompt "red dragon walk cycle"

Authentication

Supports agk_ admin tokens or plain API keys.

Resolution order:

1. --token <token> CLI flag
2. ASSET_GATEWAY_TOKEN environment variable
3. ~/.config/asset-gateway/auth.json (saved via auth set)

Config file format:

{
  "token": "agk_...",
  "gateway_url": "https://asset.origingame.dev"
}

Commands

# Auth
asset-gateway auth set <token>        # Save token locally
asset-gateway auth status             # Show current auth state
asset-gateway auth clear              # Remove saved credentials

# Generate
asset-gateway generate image --prompt "a cat" --size 1024x1024
asset-gateway generate image --prompt "icon" --transparent --provider flux
asset-gateway generate sfx --prompt "epic battle impact" --duration 3
asset-gateway generate tts --prompt "（开心）今天天气真好！" --voice 冰糖
asset-gateway voice design --voice-prompt "warm narrator" --preview-text "Welcome." --name narrator --output ./narrator.wav
asset-gateway voice clone --audio ./sample.wav --preview-text "Welcome back." --name clone --output ./clone.wav
asset-gateway generate character --prompt "medieval knight" --format fbx --pbr
asset-gateway generate prop --prompt "ornate treasure chest" --polycount 5000
asset-gateway generate model --image https://example.com/ref.png --ai-model latest
asset-gateway generate text --prompt "describe a forest" --model gpt-5.4

# 3D post-processing
asset-gateway process3d remesh --task-id <id> --format glb --polycount 20000
asset-gateway process3d retexture --task-id <id> --prompt "worn bronze armor"
asset-gateway process3d rig --task-id <id> --height 1.8
asset-gateway process3d refine --task-id <id> --pbr --hd-texture

# Providers
asset-gateway provider list
asset-gateway provider health
asset-gateway provider health flux

# Jobs
asset-gateway job list
asset-gateway job list --status pending --limit 10
asset-gateway job status <id>
asset-gateway job cancel <id>

# Self-describe (JSON Schema)
asset-gateway describe
asset-gateway describe generate

Programmatic TTS uses the SDK directly. MiMo voices include 冰糖, 茉莉, 苏打, 白桦, Mia, Chloe, Milo, and Dean; prompts may include audio tags such as （开心）, (唱歌), or [whisper].

await forge.tts("（开心）今天天气真好！", { voice: "冰糖" });

Voice design and clone are also available from the CLI:

asset-gateway voice design --voice-prompt "warm documentary narrator" --preview-text "Welcome to AssetForge." --name narrator --output ./narrator.wav
asset-gateway voice clone --audio ./sample.wav --preview-text "Welcome back." --name clone --output ./clone.wav
asset-gateway voice list --type vc

Output

JSON by default. Use --human for readable output, --fields to filter:

asset-gateway provider list --human
asset-gateway job status abc123 --fields "status,created_at"