@wesleyel

DeepSeek Harness plugin: serve every gpt-load model from one provider group, each model on its own wire protocol (OpenAI Chat Completions or Anthropic Messages), synced from /api/models