Skip to content

Commit 12a9fcc

Browse files
committed
update
1 parent 29efd98 commit 12a9fcc

File tree

2 files changed

+9
-4
lines changed

2 files changed

+9
-4
lines changed
316 KB
Loading

docs/class5/class5.rst

Lines changed: 9 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -422,7 +422,7 @@ Select **Agentflows** and the *arcadia-agent-rag**
422422

423423
.. image:: ./_static/class5-inline-02.png
424424

425-
Edit the Q&A Agent. **Double Click** the QnA node to edit the node.
425+
Edit the **QnA Agent**. **Double Click** the QnA node to edit the node.
426426

427427
.. image:: ./_static/class5-inline-03.png
428428

@@ -436,7 +436,7 @@ Enter the credential name and the API token generated earlier in F5 AI Guardrail
436436

437437
.. NOTE::
438438

439-
Recall your API token you created and stored earlier in F5 AI Guardrails portal.
439+
Recall your project API token you created and stored earlier in F5 AI Guardrails portal.
440440

441441
.. image:: ./_static/class5-inline-06.png
442442

@@ -520,13 +520,18 @@ From Arcadia RAG Chatbot, test to see if F5 AI Guardrails is able to redact PII
520520

521521
.. NOTE::
522522

523-
This is expected as we have disabled the PII package earlier. F5 AI Guardrails will not block PII data. Subsequently we will add redaction action to redact PII data.
523+
This is expected as we have disabled the PII package earlier. F5 AI Guardrails will not block PII data. Subsequent section, we will add redaction action to redact PII data.
524524

525525

526526
AI Guardrails Out-of-band flow
527527
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
528528

529-
We are going to deploy an Nginx pod to proxy chat completion to the LLM provider and as the same time Nginx via NJS extract prompt and/or response to F5 AI Guardrails for scanning for malicious content. F5 AI Guardrails deployed in an out-of-band method.
529+
We are going to deploy an Nginx pod to proxy chat completion to the LLM provider and at the same time Nginx via NJS extract prompt and/or response content to F5 AI Guardrails for scanning for malicious content. F5 AI Guardrails deployed in an out-of-band flow.
530+
531+
.. NOTE::
532+
533+
For out-of-band processing, you can use asynchronous calls from your backend applications, API gateway, middleware, or any programmable data plane that can extract prompt content and send it to F5 AI Guardrails for evaluation. Both BIG-IP (using iRules) and NGINX (using NJS) are supported. For the purpose of this class, we will use Nginx with NJS to demonstrate out-of-band processing.
534+
530535

531536
.. image:: ./_static/class5-oob-01.png
532537

0 commit comments

Comments
 (0)