Quickstart
Run your first assurance evaluation in five minutes.
Studio UI
- Open Studio → Assurance.
- Pick a rubric (or Vertical Pack).
- Paste a conversation or agent trace.
- Click Analyze, then Confirm or Override the judge verdict to calibrate tau*.
Python SDK v0.3.0
pip install quratorfrom qurator import Qurator
from qurator.testing import assert_assurance_passes
client = Qurator(api_key="YOUR_API_KEY", base_url="https://api.example")
result = client.evaluate(
messages=[
{"role": "user", "content": "What is your refund policy?"},
{"role": "assistant", "content": "Full refunds within 30 days."},
],
criteria=["Reward accurate, specific policy answers."],
)
assert_assurance_passes(result)TypeScript @qurator/sdk
npm install @qurator/sdkimport { QuratorClient } from '@qurator/sdk';
const client = new QuratorClient({ apiKey: process.env.QURATOR_API_KEY! });
const result = await client.evaluate({
messages: [
{ role: 'user', content: 'Hi' },
{ role: 'assistant', content: 'Hello!' },
],
criteria: ['Be polite'],
});cURL
curl -X POST "https://api.example/v1/assure/evaluate" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"evaluate_kind": "conversation",
"messages": [
{"role": "user", "content": "What is your refund policy?"},
{"role": "assistant", "content": "Full refunds within 30 days of purchase."}
],
"criteria": ["Reward accurate, specific policy answers."]
}'