Bring your own knowledge to the automation intelligent assistant

17 juin 20263 minutes (temps de lecture)IA/ML, Automatisation et gestion

Principal Product Marketing Manager, Red Hat Ansible Automation Platform

Last year, we released the automation intelligent assistant (formerly Red Hat Ansible Lightspeed intelligent assistant), a generative AI service accessed through a chatbot embedded within Red Hat Ansible Automation Platform. Using a retrieval augmented generation (RAG) pipeline connected to Red Hat documentation and other trusted resources, the intelligent assistant allows administrators to use natural language prompts to help them manage and troubleshoot Ansible Automation Platform without leaving the platform UI.

The automation intelligent assistant in the Red Hat Ansible Automation Platform UI

The intelligent assistant can respond with transparency to prompts such as:

What is an automation execution environment?
How do I manage user access to Ansible Automation Platform?
How do I configure Event-Driven Ansible?
What's in the release notes for Ansible Automation Platform 2.7?

But what happens when your organization's internal policies and procedures require additional steps or actions that are distinct from Red Hat's more generalized recommendations?

Automation intelligence made more relevant

With the release of Ansible Automation Platform 2.7, you can now store proprietary data into the intelligent assistant's RAG pipeline for model responses that reflect your organization's unique operational requirements and best practices. By storing a custom knowledge base, you can prioritize your own documentation (policies, manuals, FAQs) in the models' recommendations. Red Hat's documentation becomes a secondary priority or supplies responses where enterprise specific-guidance is not available.

This ability to apply custom knowledge to the intelligent assistant allows you to centralize critical and trusted documentation in one place, which reduces context switching and encourages adherence to your organization's internal policies and preferences.

Custom knowledge in practice

As part of the release preparation for Ansible Automation Platform 2.7, our engineers ran an internal test-a-thon with scenarios that clearly illustrate the value of storing custom knowledge. In this case, the "bring your own knowledge" image contained documentation for a fictional organization complete with custom naming conventions, approval workflows, and escalation procedures. Here's how the intelligent assistant's responses compared:

Test Prompt	Without custom knowledge	With custom knowledge
"How should I name my roles?"	Displays Ansible syntax requirements and recommended best practices such as avoiding uppercase letters.	Requires `mycompany_` prefix, an organizational standard that, if missing, is flagged in code review.
"How do I name inventory groups?"	Provides suggestions such as "`webservers`"or "`dbservers`"; guidance to avoid spaces and hyphens.	Returns the `mycompany_<env>_<tier>` pattern used by organization.
"How do I deploy a playbook to production?"	Explains `ansible-playbook` syntax and options	Platform team approval required. Notify #platform-ops Slack channel.
"How do I run automation in CI?"	Offers general CI/CD guidance for Ansible such as storing credentials in organizations CI tool of choice.	Returns `ansible-navigator` run --`mode stdout` and notes that interactive mode is prohibited.
"A production job just failed. What should I do?"	Provides Red Hat- recommended troubleshooting steps.	Details the internal escalation path (Team Lead → Platform Team → On-Call SRE). References the `mycompany-job-tracer` tool, and flags the 48-hour post-incident ticket requirement.
"How do I handle secrets in production?"	Suggests using `ansible-vault encrypt_string`	Points to the `mycompany.infrastructure.vault_secret` plugin and HashiCorp Vault.

As you can see, while Red Hat documentation is valuable and even, in many cases, essential, custom knowledge responses add meaningful guidance and context that can bring smoother IT operations, mitigate potential errors and delays, and improve communication between different teams. Customization can also improve your team's overall trust and confidence in the model's recommendations.

Getting started and additional resources

Custom knowledge is currently available as Technology Preview. Ansible Automation Platform 2.7 is required using a containerized installation or on Red Hat OpenShift Container Platform using the Ansible Automation Platform operator. You also need to connect your AI model of choice using Red Hat AI or third-party providers such as IBM watsonx.ai, Open AI, and Azure Open AI. However, it's important to understand that when using cloud-based AI models, portions of the intelligent assistant's prompt context may be exposed to your model provider. If you're in a regulated industry or region of the world, you may want to consider adopting a data protection agreement or exploring on-premise deployment options through a model provider such as Red Hat AI.

To configure the intelligent assistant's service, administrators need to build a custom image containing documentation and policies and deploy it alongside the assistant. To learn more about setting up custom knowledge and Ansible Automation Platform 2.7, please explore the following resources:

Documentation: Extend the automation intelligent assistant with a custom knowledge base
Webinar: What's new in Ansible Automation Platform 2.7 (July 14)
Blog: What's new with Ansible Automation Platform 2.7

À propos de l'auteur

Tricia McConnell

Principal Product Marketing Manager, Red Hat Ansible Automation Platform

Tricia McConnell is Principal Product Marketing Manager, Red Hat Ansible Automation Platform. She brings more than twenty years of experience marketing technical solutions to enterprise IT audiences.