dsh-gateway-provider

中文文档：docs/README.zh.md

Bring any LLM gateway — newapi, LiteLLM, Higress, or any OpenAI-compatible endpoint — into DeepSeek Harness. The plugin discovers the gateway's model list automatically, enriches every model with real parameters from models.dev, and serves each model over its native wire protocol (OpenAI / Anthropic / Gemini) through the pi-ai SDK.

Why this exists

DeepSeek Harness ships first-party adapters (llm-deepseek, llm-pi-ai) that each speak for one provider. If your models live behind a gateway, the alternative is a hand-maintained static model list with guessed context windows and output caps.

This plugin mounts the gateway itself: N gateways become N provider routes, with zero static model lists. Every model keeps its true parameters, and adding a model on the gateway side is enough — nothing to re-deploy.

What you get

• Multiple gateways — a default newapi route plus one gateway: route per extra gateway, each with its own catalog cache.
• Automatic discovery — GET {base}/v1/models first (including each model's supported request formats), management API fallback second.
• Real parameters — models.dev data fills context window, output cap, reasoning levels, and release date; config defaults are only a fallback.
• Every wire format — each model routes to its own protocol (openai-completions / openai-responses / anthropic-messages / google-generative-ai); no hand-written SSE or request serialization.
• A settings page — Settings → Gateway Models: add and edit gateways from templates, test connections, hide / override / custom-add models — no YAML editing required.