dsh-antigravity: A Google Antigravity Model Plugin for DeepSeek Harness
The DeepSeek Harness plugin `dsh-antigravity` is designed to resolve compatibility issues between the upstream plugin and the built-in `@deepseek-ai/dsh-llm` version in DSH Desktop Beta. The core fix replaces the imported `CallId` with `ToolCallId` to align with API changes in the built-in version. The plugin makes only minimal changes to the upstream code while preserving the original model list and OAuth flow. Users can install it via the command line, the DSH Marketplace, or by manually replacing the directory. After installation, select the Google Antigravity provider in the DSH model configuration and sign in to use it. This project is released under the MIT License and is intended for authorized accounts only.
Read MoreOpenSaozi/dsh-antigravity: Integrate Google Anti Gravity Coding Plan into DeepSeek Harness
`OpenSaozi/dsh-antigravity` is an LLM Provider plugin for DeepSeek Harness (DSH) designed to integrate models such as Gemini, Claude, and GPT-OSS from Google Anti Gravity Coding Plan or Cloud Code Assist accounts into DSH workspaces. The plugin features a lightweight design that connects directly to Cloud Code Assist via HTTPS, avoiding the need to launch additional IDE instances or background subprocesses. Core features include an independent `google-antigravity` adapter, dynamic model discovery, a built-in OAuth login page, and macOS one-click authorization.
Read Moredsh-antigravity: Integrating Google Antigravity / Cloud Code Assist into DeepSeek Harness
dsh-antigravity is an unofficial DSH Web plugin maintained by LiZhenNet, licensed under the MIT License, designed to integrate Google Antigravity / Cloud Code Assist as a model source for DSH. The plugin registers a DSH LlmAdapter under the provider route `antigravity`, directly connecting to the Cloud Code Assist streaming API, and supports Google OAuth login, account information viewing, quota grouped by model family, and reset times through the DSH Web settings page. Core features include managing models directly in the model selector, with each model displaying
Read More