Share feedback and track quality signals
Grafana Assistant relies on your feedback to improve responses and to enforce fair usage across your organization. This article explains how to submit feedback, interpret limit indicators, and understand how Grafana Labs uses the signals.
Before you begin
- Assistant access: Open the Assistant sidebar in Grafana Cloud.
- Contact channel: Know how to reach your Grafana account team or support if you need to escalate limits.
Provide feedback on responses
Click the thumbs-up icon when a response meets your needs and, if prompted, add a short note explaining why it worked.
Use the thumbs-down icon when an answer is inaccurate, incomplete, or off-topic, and describe what went wrong so the product team can reproduce the issue.
Grafana Cloud stores this feedback securely and links it to the conversation, letting engineers review the prompt and response before shipping improvements.
Understand limit indicators
The Limits badge in the Assistant sidebar displays your current usage. Hover or click to open a panel that shows conversation token consumption alongside weekly user and tenant percentages. The token counter reflects the 200,000-token soft limit for a single chat, so start a new conversation when the number climbs to keep responses fast. Weekly limits default to USD 50 per user and USD 500 per tenant during the preview, and the panel adds a buffer so the indicator turns red before you actually exceed the cap. Both budgets reset every Monday morning.
What happens when you hit a limit
If you reach a limit, the Assistant responds with a limit message and stops accepting new prompts until the budget resets or an administrator raises the allowance. The chat history stays readable, which lets you copy prompts, export findings, or capture follow-up tasks even while the Assistant pauses the conversation. If this happens regularly during critical events, contact Grafana support or your account team to discuss higher limits.
How Grafana Labs uses feedback
Usage events capture the model name, token counts, estimated cost, tool activity, and any feedback ratings you submit.
Engineers review the aggregated data to refine prompts, tune query handlers, and adjust guardrails where needed.
Grafana Labs does not use your conversation content to train external models; data handling follows the policies described in Manage your data privacy and security.
Next steps
- Understand Assistant investigations to handle multi-agent troubleshooting.
- Review Manage your data privacy and security to learn about data handling, third-party providers, security measures, and access controls.
- Review Pricing and limits to understand projected costs and limits.
Was this page helpful?
Related resources from Grafana Labs
