Architecture
AI Providers
The Nebutra-Sailor multi-tenant AI routing strategy.
Unified AI Orchestration
Nebutra deeply integrates with the Vercel AI SDK to provide a normalized API layer across 60+ global AI providers.
Available Provider Classifications
Instead of manually wrapping models, create-sailor categorizes models across 6 categories for optimized terminal prompting:
- Direct Labs:
OpenAI,Anthropic,Google Vertex/Gemini,Mistral,Cohere. - CN-Compatible (Domestic):
SiliconFlow,Volcengine (Doubao),Alibaba Bailian,Zhipu,Baichuan,Minimax,Tencent Hunyuan. - Gateway Proxies:
OpenRouter,Portkey. - Hardware Inference:
Groq,Cerebras. - Local Deployment:
Ollama,LM Studio.
Registering Custom Providers
To force a custom OpenAI proxy mapping:
import { createOpenAI } from "@ai-sdk/openai-compatible"
export const customLLM = createOpenAI({
name: "Internal Custom AI",
baseURL: process.env.CUSTOM_URL,
apiKey: process.env.CUSTOM_API_KEY
});How is this guide?
在 GitHub 上编辑此页面
最后更新于