Skip to main content

Getting Started

Your first request through the Rivaro enforcement proxy in about 5 minutes. Covers SDK configuration, detection keys, and a working example.

Rivaro sits between your AI application and the AI provider (OpenAI, Anthropic, Azure, etc.). It intercepts every request, scans for violations (PII, prompt injection, data exfiltration, tool abuse, etc.), enforces your policies, and forwards the request to the provider. Your app gets back the same response it would normally — enforcement is transparent.

Step 1: Create an AppContext​

An AppContext tells Rivaro which AI provider you're routing to and what models are allowed. Create one in the Rivaro dashboard:

  1. Go to Settings > Adapters
  2. Select your provider (e.g. OpenAI, Anthropic, Azure OpenAI)
  3. Give it a name (e.g. "Production OpenAI")
  4. Optionally configure:
    • Allowed models — restrict which models can be used through this key
    • Enabled detectors — which risk categories to scan for
    • Rate limits — requests per minute

Save it. You'll get an AppContext ID (e.g. ac_xxxxxxxxxxxx).

Step 2: Create a Detection Key​

A detection key authenticates your proxy requests and links them to your AppContext. Create one in the Rivaro dashboard:

  1. Go to Settings > Detection Keys
  2. Click Create Key
  3. Give it a name (e.g. "Production Key")
  4. Select the AppContext you just created
  5. Click Create

You'll see a key like detect_live_aBcDeFg.... Copy it now — it's only shown once.

Step 3: Change your base URL​

Point your AI SDK at your Rivaro proxy instance instead of the provider directly. That's the only code change.

OpenAI (Python)​

Before:

from openai import OpenAI

client = OpenAI(api_key="sk-your-openai-key")

After:

from openai import OpenAI

client = OpenAI(
    api_key="sk-your-openai-key",
    base_url="https://your-org.rivaro.ai/v1",
    default_headers={
        "X-Detection-Key": "detect_live_your_key_here"
    }
)

OpenAI (Node.js)​

Before:

import OpenAI from 'openai';

const client = new OpenAI({ apiKey: 'sk-your-openai-key' });

After:

import OpenAI from 'openai';

const client = new OpenAI({
  apiKey: 'sk-your-openai-key',
  baseURL: 'https://your-org.rivaro.ai/v1',
  defaultHeaders: {
    'X-Detection-Key': 'detect_live_your_key_here'
  }
});

Anthropic (Python)​

Before:

from anthropic import Anthropic

client = Anthropic(api_key="sk-ant-your-key")

After:

from anthropic import Anthropic

client = Anthropic(
    api_key="sk-ant-your-key",
    base_url="https://your-org.rivaro.ai",
    default_headers={
        "X-Detection-Key": "detect_live_your_key_here"
    }
)

curl​

curl https://your-org.rivaro.ai/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-your-openai-key" \
  -H "X-Detection-Key: detect_live_your_key_here" \
  -d '{
    "model": "gpt-4",
    "messages": [{"role": "user", "content": "Hello, world!"}]
  }'

Step 4: Make a request​

Use your SDK exactly as you normally would. Rivaro handles enforcement transparently:

response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello, world!"}]
)

print(response.choices[0].message.content)

If enforcement allows the request, you get the provider's response back unchanged. If a policy blocks the request, you'll get an error response from Rivaro (see Error responses below).

Step 5: See enforcement results​

Open the Rivaro dashboard. You'll see your request in the activity feed with:

  • Detections — what Rivaro found (PII, prompt injection, etc.)
  • Policy action — what happened (allowed, logged, blocked, redacted)
  • Risk classification — severity, risk domain, risk category
note

If no policies are configured, Rivaro runs in observation mode — it detects and logs violations but doesn't block anything. Configure policies in the dashboard to start enforcing.

Error responses​

When Rivaro itself rejects a request (not the AI provider), you'll get:

HTTP StatusMeaningExample
401Detection key missing or invalid{"error": "Detection key required for proxy endpoints."}
403Model not in allowed list{"error": "The requested model is not permitted."}
429Rate limit exceeded{"error": "Rate limit exceeded"}
451Request blocked by policy{"error": "Request blocked by enforcement policy"}

Errors from the AI provider (e.g. invalid provider API key, quota exceeded) are passed through in the provider's own format.

Streaming​

Streaming works out of the box. Use stream=True (Python) or stream: true (Node.js/curl) as you normally would. Rivaro streams chunks back in real time with enforcement applied.

stream = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Explain quantum computing"}],
    stream=True
)

for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

What's next​