Choose a Provider Upstream
A provider upstream is the model service AISIX AI Gateway calls after resolving a caller-facing model alias.
When you configure the upstream side of the request path, choose how AISIX should talk to the model service. The main decision is the upstream API format: an OpenAI-compatible API, AWS Bedrock, Google Vertex AI, or Azure OpenAI.
Setup Paths
Choose the adapter that matches the upstream API format, not the provider's business category.
| Upstream API Format | Examples | Setup Guide | Adapter |
|---|---|---|---|
| Public OpenAI-compatible vendor | DeepSeek, Groq, Mistral, Together.ai, Fireworks, Perplexity | OpenAI-Compatible Vendor Upstream | openai |
| Private OpenAI-compatible endpoint | vLLM, SGLang, Ollama, private model proxy | Bring Your Own Endpoint | openai |
| AWS Bedrock native API | Bedrock foundation models or inference profiles | AWS Bedrock Upstream | bedrock |
| Google Vertex AI native API | Vertex publisher models | Google Vertex AI Upstream | vertex |
| Azure OpenAI API | Azure OpenAI deployments | Azure OpenAI Upstream | azure-openai |
Each setup guide creates the same gateway resources: a provider key, a model alias, and a caller API key. The provider-specific guide explains which credential, upstream model identifier, and base URL shape that provider family expects.
After configuring the provider key, model, and caller API key, send one proxy request through AISIX. Confirm that the response uses the caller-facing model alias.