The official nRouter R SDK with R-universe preview, SSE streaming callbacks, and automated metadata extraction.

The official nRouter R SDK (nrouter) provides an idiomatic interface for data scientists and researchers using R. Because there is no official OpenAI SDK for R, this package communicates directly with https://api.nrouter.ai/v1 via httr, offering full support for chat completions, token streaming callbacks, and metadata parsing.

Installation

Install the public preview package from R-universe:

install.packages(
  "nrouter",
  repos = c(
    nroutergateway = "https://nroutergateway.r-universe.dev",
    CRAN = "https://cloud.r-project.org"
  )
)

For development or bleeding-edge updates, install directly from the GitHub repository:

remotes::install_github("nRouterGateway/nrouter-sdk", subdir = "sdks/r")

Setup & Authentication

Set your virtual API key in your environment or in your project's .Renviron file:

export NROUTER_API_KEY="sk-nrouter-your-api-key-here"

In R code:

Sys.setenv(NROUTER_API_KEY = "sk-nrouter-your-api-key-here")

Initialize the client targeting https://api.nrouter.ai/v1:

library(nrouter)

# Reads NROUTER_API_KEY from environment
client <- nrouter_client()

You can also pass explicit parameters:

client <- nrouter_client(
  api_key = "sk-nrouter-your-api-key-here",
  base_url = "https://api.nrouter.ai/v1"
)

Quickstart

Send a chat completion request using standard lists:

library(nrouter)

client <- nrouter_client()

result <- nrouter_chat_completions(client, list(
  model = "gpt-5.4-mini",
  messages = list(list(role = "user", content = "Hello!"))
))

# Print completion text
cat(result$body$choices[[1]]$message$content, "\n")

# Check cost
if (nrouter_is_priced(result$meta)) {
  cat(sprintf("Cost: $%f\n", result$meta$cost))
}

Streaming

The SDK supports real-time Server-Sent Events (SSE) streaming through callback functions:

nrouter_messages_stream(client, list(
  model = "claude-haiku-4-5-20251001",
  max_tokens = 64,
  messages = list(list(role = "user", content = "Hello!"))
), function(chunk) {
  cat(chunk$delta)
})

Return FALSE from the callback function to terminate the stream early.

Error Handling

Errors are raised as classed R conditions, allowing precise condition handling with tryCatch:

tryCatch(
  nrouter_chat_completions(client, body),
  nrouter_guardrail_blocked_error = function(e) {
    message("Request blocked by guardrails: ", conditionMessage(e))
  },
  nrouter_credit_error = function(e) {
    message("Account balance exhausted: ", conditionMessage(e))
  },
  nrouter_rate_limit_error = function(e) {
    if (nrouter_is_retryable(e)) {
      message("Rate limit exceeded; retryable after backoff")
    }
  },
  nrouter_error = function(e) {
    message("nRouter general error: ", conditionMessage(e))
  }
)

Class	HTTP Status	Description
`nrouter_request_error`	400	Invalid request format
`nrouter_guardrail_blocked_error`	400	Blocked by server-side guardrail check
`nrouter_authentication_error`	401	Invalid or missing API key
`nrouter_credit_error`	402	Out of credits or budget exceeded
`nrouter_not_found_error`	404	Model not found
`nrouter_rate_limit_error`	429	Rate limit or TPM quota exceeded
`nrouter_service_error`	503	Gateway or upstream service unavailable
`nrouter_transport_error`	—	Connection error

Response Metadata & Cost Tracking

result$meta holds parsed x-nr-* gateway headers:

meta <- result$meta
cat("Request ID:", meta$request_id, "\n")
cat("Model:", meta$model, "\n")

if (nrouter_is_priced(meta)) {
  cat(sprintf("Cost: $%f\n", meta$cost))
} else {
  cat("Cost: unpriced\n")
}

Source & Repository

The source code and runnable demonstrations are available on GitHub:

R SDK Source & Demos

R SDK