Environments overview

For reinforcement learning runs, the Environments tab lists the environments in the run — each a single instance of the world an agent interacts with, identified by its Environment ID.

The environments table

Each row is an environment. Two columns are always shown, pinned to the left:

Environment ID — the environment’s env_id. Click it to open the environment’s charts.
Episode number — the episode selected for that environment at the current scale position: on the Episode scale, the requested episode (capped at its latest); on the RL Step scale, the latest episode that contains the selected RL step.

Add more columns from Columns to compare environments at a glance:

Latest episode — the highest episode number the environment has logged.
Latest RL step — the highest RL step the environment has logged.
Metric summaries — any logged metric (such as reward or loss) aggregated as avg, min, max, first, last, sum, count, or std at the selected scale point.

The table uses the shared table controls — add / remove / reorder / resize columns, sort, and export to CSV. Here, search matches the Environment ID and filters apply to metric summaries.

Scale toggle & RL step slider

Above the table you choose how to position in time (see Episodes & steps):

RL Step / Episode toggle — pick which scale to navigate by.
The slider scrubs to a specific RL step (or episode), and the table updates to reflect that point.

Opening an environment

Click an environment’s id to open its charts, where you can visualize that environment’s metrics.