Gateways

What is an Gateway?

A gateway is a solution for having a single entry for your users to the AI models supported by your organization, using gateways you can setup security to your AI usage like adding guardrails, role based permissions, rate limiting, as well as an easy way to create your applications pointing to specific paths from your gateway and easily switch between providers and models keeping all your security settings.

You can see your gateways here

Default Gateways

By default all organizations are onboarded with a chat and an embeddings route gateway setup using Fairo provider, but you can create as many gateways paths as you want.

Supported Settings

Providers

You can select the provider, the model you want to use and the routes you want to use from that model, a gateway can contain multiple model routes as long as there are no conflicts with other existing gateway paths. (e.g You can have only one "/chat" on gateway "/" path)

Rate limiting

You can setup a limit of requests by second, minute, hour or day.

Guardrails

We only apply guardrails to user inputs and final outputs, and all filtering actions are fully logged in the trace — nothing is hidden.

The available guardrails options are:

PII Filtering

Removes personally identifiable information from prompts and responses.

You can select multiple options from the following:

General - Masks Phone, Email, Address, Age, Username, Password, Driver ID, License plate, Vehicle Identification Number (VIN)
Finance - Masks CVV, Credit/Debit card expiry, Credit/Debit card number, Personal Identification Number (PIN), International Bank Account Number, SWIFT code
IT - Masks IPv4 address, Media Access Control (MAC) address, Web address, AWS access key, AWS secret key
USA Specific - Masks US passport number, US Social Security Number (SSN), US Individual Taxpayer Identification Number (ITIN), US bank account number, US bank account routing number
Canada Specific - Masks Canadian Health Service Number, Canadian Social Insurance Number (SIN)

Toxicity Filtering

Filters content for the following categories: Hate, Insults, Sexual, and Misconduct.

You can select one of the following options:

Low - Blocks only the most extreme toxic content such as threats or hate speech
Medium - Filters a balanced level of harmful language including insults, harassment, and offensive comments
High - Applies the strictest filtering to remove even mild toxicity or potentially harmful tone

Profanity Filtering

Blocks profane and offensive language from prompts and responses.

You can toggle it as enabled/disabled.

Custom guardrails can be created upon request.