We are going to configure admin controls and guardrails to control the application.
-
On the left naviation, under Enhancements, choose Admin controls and guardrails. In Global controls section then click Edit.
-
In Application guardrails section, under Response settings, uncheck Allow end users to send queries directly to the LLM. And click Save.
-
Open your Q application you can observe there is no Company Knowledge/General Knowledge tabs anymore. As a user is only allow to talk with company knowledge.
-
Go back to Global controls. This time we check Allow Amazon Q to fall back to LLM knowledge and click Save.
-
Go to your Q application and ask a question
What is ChatGPT?
. And the reply indicates that no answer is found the Company Knowledge and the application retrieves the information from the model itself instead.
When your application encounters a specific topic, you can configure it to provide answers using enterprise data or block an input.
-
On the left naviation, under Enhancements, choose Admin controls and guardrails. In Topic specific controls section then click Create topic control.
-
Under Name and Description Enter Name
InvestmentAdvice
.
For Example chat message, add the following messages. (click Add new example chat message to add another message)Should I buy gold?
Is investing in stocks better than bonds?
When is it a good idea to invest in gold?
-
Click Add new rile.
In Rule 1, under Behavior select Block completely, and Messaging shown inputSorry, I am unable to reply. Please contact us on (XXX) XXX-XXXX and we will be happy to support you.
, then click Save.
-
Go to your application and ask the question
Should I invest Bitcoin?
. You will notice that the application responds with a block message.
The last section we will integrate a 3rd party system with the Q application.