ChatGPT for DevOps
Complex actions. Simple conversations
Kubiya.Ai
Error on Kubernetes namespace Observability:kubiya-ai-6c8dd8d967-7bhd8: 10.0.20.23
status CrashLoopBackOff
Restarts 6
Age 7m45sec
.png)
Dan (R&D)
@Kubiya What does it mean?
Kubiya.Ai
The CrashLoopBackOff error can occur due to various reasons — the wrong deployment of Kubernetes, liveness probe misconfiguration, and init-container misconfiguration. You can also bypass the error by creating a separate deployment with the help of a blocking command.
.png)
Sam (R&D)
@kubiya How many tickets are still not done in Jira for the current sprint?
Kubiya.Ai
With 3 days left into the current sprint 12 tickets are in-progress, 8 tickets are in QA and 2 tickets are in backlog
Sam (R&D)
@kubiya Which 2 tickets are in backlog?
Kubiya.Ai
The 3 tickets in backlog are:
FE-112 - Refactor page loader
OPS-89 - Setup a pipeline notification system
Sam (R&D)
@kubiya Move Jira ticket FE-112 to in-progress
Kubiya.Ai
Done. FE-112 was updated to in-progress
ChatGPT for DevOps
Need help? Tag Kubiya and speak your mind and Kubiya will do the rest.
It will understand the context of your request. What information is provided and what other questions to take to get full details. You can run multiple conversations in parallel and even switch context back and forth, just as you would with your DevOps team.
Access Control
Kubiya’s platform includes a built-in access control where you can define which users or groups get access to do specific actions. Users can request temporary or permanent access, through your chat tool, with Kubiya managing the TTL.
If you have policies defined in OKTA, LDAP, Kubiya supports Open Policy Agent (OPA).
Joe (R&D)
@kubiya How do I set up a VPN?
Kubiya.Ai
To setup a VPN you will need to make a request through the helpdesk portal.
Once approved you will get a username, a temporary password and a license to use in your VPN client. Here’s a short video and the link to the helpdesk site.
Reinforcement learning
with Human Feedback (RLHF)
Kubiya learns from user interactions to optimize the experience at the organization and at the user level. If a users chooses certain clusters or namespace most of the time, Kubiya will over time suggest them by default. if a 2nd or 3rd alternatives were chosen for a given natural language query, Kubiya will change the order of its suggestions.
RLHF is kept at the organization and the user level and is used to optimize the experience for other organizations (more about how we keep data secure here).
Finally the user can give feedback on responses to improve accuracy.
Flexibility to create what you need
You can create agents to do various tasks and guide them how to do them. Some examples:
Define an agent that is only allowed full control to certain clusters and read-only to other clusters.
Provision cloud resources using your own IaC templates.
Run chained actions.
and much more.
Platforms can trigger actions too
With Kubiya, platforms, not just humans, can converse with Kubi. Using webhooks, events on 3rd party platforms can trigger actions, or just start a conversation.
Connect Kubi to your ticketing system, such as Jira, and have it fulfill your tickets once they are approved.
Connect Kubi to your Kubernetes and automatically scale up and down deployments.
Analytics
Want to know which resources are used the most? Who makes most use of infrastructure or what documents are out-of-date?
Kubiya keeps track of everything and can give you a level of visibility you never had before.
How it works
Creating a ChatGPT-like for engineering platforms requires a complex setup. It involved authenticating users, local operators to perform actions, secure tunnels that communicate with Kubiya’s network, and a combination of large language models, each for a different purpose and sometimes multiple of these to optimize the results. Finally everything needs to be secure and performant.
