Three quick steps to evaluate your AI outputs with RAIL. You can use cURL, Python, or Next.js. Copy, paste, and you’re done.
Sign up or log in, then click Generate API key on your dashboard.
Take any model output, pass it to our /v1/rail/score/basic endpoint, and get scores.
curl --location 'https://api-693516574005.us-central1.run.app/v1/rail/score/basic' \
--header 'Content-Type: application/json' \
--header 'RAIL-API-KEY: YOUR_API_KEY' \
--data '{
"content": "This AI tool supports accessibility by providing text-to-speech and customizable font sizes. It ensures privacy by not collecting any personal data."
}'import os, requests
BASE_URL = "https://api-693516574005.us-central1.run.app"
API_KEY = os.getenv("RAIL_API_KEY", "YOUR_API_KEY")
payload = {
"content": "This AI tool supports accessibility via TTS and custom fonts."
}
r = requests.post(
f"{BASE_URL}/v1/rail/score/basic",
headers={"RAIL-API-KEY": API_KEY, "Content-Type": "application/json"},
json=payload,
timeout=30
)
print(r.status_code, r.json())const BASE_URL = "https://api-693516574005.us-central1.run.app";
const API_KEY = process.env.RAIL_API_KEY;
export async function getRailScoreBasic(content) {
const res = await fetch(`${BASE_URL}/v1/rail/score/basic`, {
method: "POST",
headers: {
"Content-Type": "application/json",
"RAIL-API-KEY": API_KEY || "YOUR_API_KEY"
},
body: JSON.stringify({ content }),
cache: "no-store"
});
if (!res.ok) throw new Error(`RAIL error ${res.status}`);
return res.json();
}Inspect per-dimension scores and the weighted score. Need more control? Use /v1/rail/score/custom with your own weights, or /v1/rail/reprompt to generate an improved prompt.
Supply weights that sum to 1.0 and get a weighted score tuned to your use case.
curl --location 'https://api-693516574005.us-central1.run.app/v1/rail/score/custom' \
--header 'Content-Type: application/json' \
--header 'RAIL-API-KEY: YOUR_API_KEY' \
--data '{
"content": "Any LLM output here...",
"weights": {
"fairness": 0.15, "safety": 0.15, "reliability": 0.10, "transparency": 0.10,
"privacy": 0.15, "accountability": 0.10, "inclusivity": 0.10, "user_impact": 0.15
}
}'