AI requests

Send a request to any supported model on behalf of a signed-in user. rNet charges the user's credits and returns the provider's response unchanged.

Send a request

POSThttps://ai-provider.rnetai.org/ai

Name	In	Description
`model` stringrequired	query	Model ID, for example `gemini-2.5-flash-lite`. See Models.
`access_token` stringrequired	query	The user's access token from Connect with rNet.
`body` JSONrequired	body	The provider's own request body. Send `Content-Type: application/json`.

bash

curl -X POST "https://ai-provider.rnetai.org/ai?model=gemini-2.5-flash-lite&access_token=$ACCESS_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"contents":[{"role":"user","parts":[{"text":"Hello!"}]}]}'

Stream a response

POSThttps://ai-provider.rnetai.org/ai/stream

Same parameters as /ai. The response is streamed as the provider generates it (server-sent events for Gemini models). For Groq models, also set "stream": true in the body.

Request body by provider

rNet does not change the body, so use the format each provider expects:

json

// Gemini / Gemma (Google): generateContent
{ "contents": [{ "role": "user", "parts": [{ "text": "Hello!" }] }] }

// gemini-embedding-001: embedContent
{ "content": { "parts": [{ "text": "Text to embed" }] } }

// Groq models: OpenAI-style chat completions
{ "model": "llama-3.1-8b-instant", "messages": [{ "role": "user", "content": "Hello!" }] }

// text-embedding-3-small (OpenAI): embeddings
{ "model": "text-embedding-3-small", "input": "Text to embed" }

Response

On success you receive the provider's response with HTTP 200. On failure you receive an error object; see Errors.

Billing

rNet checks the user's balance before forwarding the request and deducts the credits used afterwards. If the balance is too low, the request fails with 402 and nothing is sent to the provider.