Ask KubeOn
Turn on the assistant, choose its model endpoint, and understand its limits.
On this page
Turn it on
An admin enables the assistant in Settings > Assistant. The admin chooses the model endpoint it runs on: Amazon Bedrock, Azure OpenAI or Google Vertex AI. The assistant is off by default.
How it answers
- It calls up to 10 read-only tools that use the same APIs as the app, with the asking user's permissions.
- At most 8 tool rounds and 2,048 output tokens per answer; tool results are trimmed before they reach the model.
- Every amount and percentage in the answer is matched against tool output, and the answer shows how many were traced.
Limits and usage
| Setting | Default |
|---|---|
| Questions per user per day | 50, resetting at midnight UTC |
| Conversation history sent | The last 6 turns |
| Weekly digest | Mondays at 13:00 UTC, to the channels you choose |
| Usage ledger | Every call with user, feature, model, tokens, estimated cost and latency |
Need help with this guide?Contact support