One AI API for your whole stack.
VXAI is an OpenAI-compatible gateway for chat, embeddings, and streaming — routed across Anthropic, OpenAI, and self-hosted models — plus retrieval-augmented generation over your own documents and data. Change the base URL and key; keep your SDK.
Ship AI features without vendor lock-in.
OpenAI-compatible API
Point any OpenAI SDK at one endpoint — chat completions, embeddings, and streaming work unchanged. Only the base URL and key change.
Multi-provider routing
Route by model name to Anthropic, OpenAI, or self-hosted Ollama behind a single key, so you are never locked to one vendor.
Token streaming
Stream responses as server-sent events in the standard OpenAI chunk format, all the way through to a final [DONE].
Knowledge bases (RAG)
Build knowledge bases from file uploads, VXS3 buckets, URLs, and VXLake tables — extracted, chunked, and embedded for you.
Cited answers
Retrieval-augmented chat answers from your own documents and returns the source citations behind every response.
Semantic search
Query your knowledge bases by meaning, not keywords, with embeddings stored in VeexaDB.