iceprosurface/dsh-gateway-config

README 摘要

gateway-config A DSH plugin for configuring any OpenAI-compatible LLM gateway (any endpoint that serves GET /models ) as a provider — one provider route per gateway, multiple gateways side by side. What it provides - Host tool: gateway configure - Web settings panel: LLM 网关配置 - Gateway model discovery via GET /models with non-chat model filtering - Per-model reasoning-effort mappings - Per-model text/image input capability mappings - One managed credential per route ( API KEY , e.g. CINDY API KEY ) Providers are fixed presets; no ad-hoc routes are created. Anything that speaks the OpenAI GET /models shape (the de-facto standard DSH discovery reads) works: - cindy (Cindy), sub2api (Sub2API) — private gateways; addresses never live in this repository. The host tool reads them from CINDY BASE URL / SUB2API BASE URL ; the web panel remembers the last address per route. - generic (通用) — any other OpenAI-compatible endpoint; supply its base URL ( GATEWAY BASE URL env or the panel's Base URL field). - Public vendors with OpenAI-compatible APIs carry their documented base URLs out of the box: deepseek , moonshot (Kimi), zhipu (GLM), qwen , openrouter , siliconflow , xai (Grok). Install int…

分类