Supported Providers
XSI selects the path mapping and authentication method automatically from the upstream hostname of each provider key. Clients always use the standard paths of the official protocol (such as /v1/chat/completions and /v1/messages). Azure and Vertex AI are declared with config.provider_type; see Provider Key API.
Native support
| Provider | Provider URL | Supported protocols | Notes |
|---|---|---|---|
| OpenAI | https://api.openai.com | All OpenAI endpoints, including Responses WebSocket, Realtime, and Live | Paths without /v1 are completed automatically |
| Azure OpenAI / Azure AI | https://<resource>.openai.azure.com, etc. | Chat, Responses, Embeddings, Images, Audio | provider_type: azure; supports deployment mapping and api-version; use an OpenAI upstream for Realtime connections |
| Anthropic | https://api.anthropic.com | Messages, count_tokens, Files | Always authenticates with x-api-key |
| Google Gemini | https://generativelanguage.googleapis.com | Native Gemini API; OpenAI-compatible Chat and Embeddings | Native paths use x-goog-api-key |
| Google Vertex AI | https://aiplatform.googleapis.com | Native Gemini API | provider_type: vertex; service account authentication |
| xAI | https://api.x.ai | Chat, Responses, Images, Videos | Video job polling is bound to the account |
| DeepSeek | https://api.deepseek.com | Chat, Responses, Messages, web tools, OCR, TTS (including WebSocket), ASR, Files, balance | See DeepSeek Full-Suite Integration |
| Alibaba Cloud Model Studio / DashScope | https://*.aliyuncs.com | Chat, Embeddings, Rerank, Responses, Messages, Realtime | Realtime is mapped to /api-ws/v1/realtime automatically |
| Volcano Engine Ark | https://ark.cn-beijing.volces.com | Chat, Responses, Embeddings, Images, Messages | Includes Coding Plan paths |
| Baidu Qianfan | https://*.baidubce.com | Chat, Responses, Embeddings, Images, Rerank, Messages | |
| Zhipu | https://open.bigmodel.cn, https://api.z.ai | Chat, Embeddings, Images, Messages | |
| Moonshot | https://api.moonshot.cn | Chat, Messages | |
| MiniMax | https://api.minimaxi.com | Chat, Messages | |
| Tencent Cloud | https://api.lkeap.cloud.tencent.com | Chat, Messages | |
| LongCat | https://api.longcat.chat | Chat, Messages | |
| Perplexity | https://api.perplexity.ai | Chat | |
| Cohere | https://api.cohere.ai | Chat, Embeddings (compatibility API) | |
| DeepInfra | https://api.deepinfra.com | Chat | |
| Fireworks | https://api.fireworks.ai | Chat, Completions, Responses, Messages, Rerank | |
| GitHub Models | https://models.github.ai | Chat | |
| Groq, Cerebras, SiliconFlow, OpenRouter, Ollama | Their official URLs | OpenAI-compatible API | Bearer authentication; paths forwarded as is |
| Cloudflare AI Gateway | https://gateway.ai.cloudflare.com/... | Depends on gateway configuration | Authenticates with cf-aig-authorization |
Connecting other compatible services
Any service with an OpenAI-compatible or Anthropic-compatible API (including self-hosted vLLM, SGLang, Ollama, and so on) can be used as a provider:
- Set
Providerto the service's base URL. It may include a path prefix, such ashttps://llm.internal.example.com/openai; XSI appends the request path to it. - OpenAI-style paths are forwarded with
Authorization: Bearer;/messagespaths usex-api-key. - When the provider is a public IP address, XSI automatically adds
/v1to/chat/completions,/embeddings,/responses, and/messages.
Models and pricing
After connecting a provider, the models must also appear in the price list so XSI can bill and map them correctly:
- The default price list,
api/pricing.json, already includes the main models of the GPT, Claude, Grok, DeepSeek, GLM, Kimi, Doubao, and Qwen families. - Add other models through the tenant
PRICINGsetting. See Billing and Pricing. - Public prices are listed in the model catalog and at
GET /x-pricing.