dsh-antigravity

English | 简体中文

Google Antigravity / Cloud Code Assist model provider for DeepSeek Harness.

Based on LiZhenNet/dsh-antigravity, with a multi-account pool added: you can sign in with several Google accounts and requests are scheduled across them. See QUOTA-SCHEDULING.md for how scheduling works.

Unofficial integration. This project is not affiliated with or endorsed by Google. Use it only with accounts and services you are authorized to access.

---

Installation

Method 1: Install from GitHub (Recommended)

dsh plugin --profile web add github:almost-nowhere/dsh-antigravity

Method 2: Local Release Package

npm run pack:dist
dsh plugin --profile web add ./dist/dsh-antigravity-0.0.9.tgz

Or add to your profile's cordis.patch.yml:

- insert:
    - id: llm-antigravity
      name: dsh-antigravity

---

Login & Authentication

Open Settings > Antigravity and click Login to authorize via Google OAuth. The settings page automatically refreshes your quota upon login. You can add multiple accounts in the settings view to configure a pool.

The plugin starts a loopback OAuth callback server at http://localhost:51121/oauth-callback. If your browser does not open automatically, run the CLI helper:

node "$DSH_HOME/profiles/web/node_modules/dsh-antigravity/bin/antigravity-login.mjs"

Credentials are saved at: $DSH_HOME/storages/antigravity-oauth.json.

---

Models

Select Antigravity in the DSH model picker. You can configure individual models in Settings > Antigravity.

Model ID	Name	Quota pool
gemini-3.8-flash	Gemini 3.8 Flash	Gemini
gemini-3.7-flash	Gemini 3.7 Flash	Gemini
gemini-3.6-flash	Gemini 3.6 Flash	Gemini
gemini-3.5-flash	Gemini 3.5 Flash	Gemini
gemini-3.1-pro	Gemini 3.1 Pro	Gemini
gemini-3.1-flash-image	Gemini 3.1 Flash Image	Gemini
gemini-3-flash	Gemini 3 Flash	Gemini
gemini-2.5-pro	Gemini 2.5 Pro	Gemini
gemini-2.5-flash	Gemini 2.5 Flash	Gemini
claude-opus-4-6	Claude Opus 4.6	Claude & GPT (3P)
claude-sonnet-4-6	Claude Sonnet 4.6	Claude & GPT (3P)
gpt-oss-120b	GPT-OSS 120B	Claude & GPT (3P)

Reasoning-capable models map thinking tokens into DSH's native reasoning blocks, with configurable effort tiers.

---

License

MIT