dsh-antigravity

When building custom large language model workflows, developers frequently struggle with juggling multiple AI service providers, manually managing credentials, and tracking usage limits. dsh-antigravity bridges this gap by seamlessly integrating Google Cloud Code Assist and compatible third-party models directly into DeepSeek Harness, offering a unified, plug-and-play inference gateway tailored for local development environments.

At its core, the plugin streamlines authentication and brings intuitive quota management to your fingertips. After securing a single Google OAuth sign-in via a built-in local callback server, the web settings dashboard instantly displays account details, categorized usage pools (Gemini variants share one pool, while Claude and GPT-OSS share another), and real-time countdowns until limit resets. Users can toggle individual models on or off, with active models automatically prioritized at the top of the list alongside live remaining quota percentages. Under the hood, it communicates directly with the streaming API and dynamically resolves public model IDs to runtime endpoints, ensuring low-latency, reliable inference.

Designed for developers, researchers, and power users who rely on DeepSeek Harness but refuse to be locked into a single AI vendor. By providing transparent token accounting, granular model control, and a clean bilingual interface, this plugin empowers you to orchestrate cross-platform models efficiently while staying within your familiar workflow. Released under the permissive MIT license, it keeps full control in your hands.