Patronus AI

Patronus excels in industry-specific guardrails for RAG workflows.

It has a SOTA hallucination detection model Lynx, which is also open source. Portkey integrates with multiple Patronus evaluators to help you enforce LLM behavior.

What is Patronus AI?

Learn more about Patronus AI and their offerings.

Using Patronus with Portkey

1. Add Patronus API Key to Portkey

Grab your Patronus API key from here.

On the Integrations page, click on the edit button for the Patronus and add your API key.

2. Add Patronus’ Guardrail Checks & Actions

Navigate to the Guardrails page and you will see the Guardrail Checks offered by Patronus there. Add the ones you want, set actions, and create the Guardrail!

List of Patronus Guardrail Checks

Check Name	Description	Parameters	Supported Hooks
Retrieval Answer Relevance	Checks whether the answer is on-topic to the input question. Does not measure correctness.	ON or OFF	afterRequestHooks
Custom Evaluator	Checks against custom criteria, based on Patronus evaluator profile name.	string(evaluator’s profile name)	afterRequestHooks
Is Concise	Check that the output is clear and concise.	ON or OFF	afterRequestHooks
Is Helpful	Check that the output is helpful in its tone of voice.	ON or OFF	afterRequestHooks
Is Polite	Check that the output is polite in conversation.	ON or OFF	afterRequestHooks
No Apologies	Check that the output does not contain apologies.	ON or OFF	afterRequestHooks
No Gender Bias	Check whether the output contains gender stereotypes. Useful to mitigate PR risk from sexist or gendered model outputs.	ON or OFF	afterRequestHooks
No Racias Bias	Check whether the output contains any racial stereotypes or not.	ON or OFF	afterRequestHooks
Detect Toxicity	Checks output for abusive and hateful messages.	ON or OFF	afterRequestHooks
Detect PII	Checks for personally identifiable information (PII) - this is information that, in conjunction with other data, can identify an individual.	ON or OFF	afterRequestHooks
Detect PHI	Checks for protected health information (PHI), defined broadly as any information about an individual’s health status or provision of healthcare.	ON or OFF	afterRequestHooks

Your Patronus Guardrail is now ready to be added to any Portkey request you’d like!

3. Add Guardrail ID to a Config and Make Your Request

Patronus integration on Portkey currently only works on model outputs and not inputs.

• When you save a Guardrail, you’ll get an associated Guardrail ID - add this ID to the after_request_hooks params in your Portkey Config.
• Save this Config and pass it along with any Portkey request you’re making!

Your requests are now guarded by your Patronus evaluators and you can see the Verdict and any action you take directly on Portkey logs! More detailed logs for your requests will also be available on your Patronus dashboard.

---

Get Support

If you face any issues with the Patronus integration, just ping the @patronusai team on the community forum.