Many Kafka clusters expose metrics through Prometheus, but troubleshooting usually still requires remembering metric names and navigating multiple dashboards.

In this update, StreamLens can ingest a small set of important metrics and make them queryable through the built-in AI chat panel.

Example:

Prometheus metric under_replicated_partitions

Instead of searching dashboards, you can ask:

“How many partitions are not fully in sync with replicas?”

The assistant maps the question to the relevant metric and returns the value.

Prometheus metrics are also now used for producer detection, which helps identify active producers in environments where JMX is not available (for example managed Kafka services).

The goal is to make Kafka troubleshooting more conversational — asking questions about cluster health instead of searching across multiple monitoring tools.

Repo: https://github.com/muralibasani/streamlens

Would be interested in feedback from people running Kafka clusters on:

- which metrics are most useful during incidents - whether natural language queries for metrics are actually helpful - other Prometheus metrics worth integrating