← Back to talks

Talk · Cloud Native Guayaquil · CNCF

AI agents operating Kubernetes: how they work under the hood

A deploy that never finishes on EKS, investigated live with Kobi, and what happens under the hood: how the agent decides it has to look outside the cluster, the harness that lets it read AWS without being able to change anything, the failures we fixed in the tools instead of the prompt, and what each incident costs.

Guayaquil Given in Spanish, slides translated 33 slides

Click the slides and use the arrow keys, or tap either side of the slide. Open in a new tab Watch the recording (in Spanish)

Takeaways

If you're building one

  1. Put the limits outside the model In the gateway, the session and the role. And generate all three from the same source.
  2. When it fails, fix what it reads Add the missing data to the tool result. Leave the prompt as it is.
  3. Measure how much it reads Bytes and tokens per tool, as well as whether the answer was right.
  4. Test it with real failures Break a test cluster and save each answer with what it cost.

← Back to talks