API reference

Everything is OpenAI-compatible. If a parameter works against OpenAI, it works here — unsupported ones are passed through to the upstream rather than rejected.

Endpoints


POST /chat/completions	The main endpoint. Supports streaming, tool calling, JSON mode and reasoning models.
POST /embeddings	Billed on input tokens only — there is no output side.
GET /models	Returns only the models this key is allowed to use, in OpenAI list format. Most clients call this on startup.

Chat completions

curl https://api.voyai.net/v1/chat/completions \
  -H "Authorization: Bearer $VOYAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-ai/DeepSeek-V3.2",
    "messages": [{"role": "user", "content": "Hello"}],
    "stream": true
  }'

Parameters we handle specially

Parameter	Note
model	Use the model ID exactly as shown on the Models page. We map it to whichever upstream account serves it.
stream	We inject stream_options.include_usage automatically so the final chunk carries exact token counts. Your bill matches what was generated.
max_tokens	Used to size the pre-authorization hold on your balance. Setting it very high on a low balance can get the request rejected even though the reply would have been short.
tools	Passed through. Tool definitions count as input tokens.

Embeddings

Billed on input tokens only — there is no output side.

curl https://api.voyai.net/v1/embeddings \
  -H "Authorization: Bearer $VOYAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "BAAI/bge-m3", "input": "hello world"}'

List models

Returns only the models this key is allowed to use, in OpenAI list format. Most clients call this on startup.

curl https://api.voyai.net/v1/models \
  -H "Authorization: Bearer $VOYAI_API_KEY"