Modified evaluate-llm-application.mdx, added section Add metadata to an experiment.

Author: fjmorris

src/langsmith/evaluate-llm-application.mdx

@@ -301,6 +301,40 @@ await evaluate((inputs) => toxicityClassifier(inputs["input"]), {
 
 </CodeGroup>
 
+## Add metadata to an experiment
+
+Metadata is a set of key-value pairs you can attach to an experiment to group and filter experiments in the experiments table. You can pass metadata when running an experiment via the `metadata` argument (see [Run the evaluation](#run-the-evaluation)), or add it afterwards directly in the LangSmith UI.
+
+To open the **Edit Experiment** panel, hover over an experiment row in the experiments table and click the **Edit** pencil icon that appears at the right of the row.
+
+<img
+  className="block dark:hidden"
+  src="/langsmith/images/experiments-table-edit-icon-light.png"
+  alt="Experiments table with the edit pencil icon visible on a hovered row."
+/>
+<img
+  className="hidden dark:block"
+  src="/langsmith/images/experiments-table-edit-icon-dark.png"
+  alt="Experiments table with the edit pencil icon visible on a hovered row."
+/>
+
+The **Edit Experiment** panel lets you update the experiment name and description, and manage metadata key-value pairs. Click **+ Add Metadata** to add a new key-value pair, then click **Submit** in the top right to save your changes.
+
+<img
+  className="block dark:hidden"
+  src="/langsmith/images/edit-experiment-panel-light.png"
+  alt="Edit Experiment panel showing metadata key-value pairs and the Add Metadata button."
+/>
+<img
+  className="hidden dark:block"
+  src="/langsmith/images/edit-experiment-panel-dark.png"
+  alt="Edit Experiment panel showing metadata key-value pairs and the Add Metadata button."
+/>
+
+Once experiments are tagged with metadata, use the **Group by** control at the top of the experiments table to cluster experiments by any metadata field. The summary charts above the table update per group, showing average feedback scores, latency, and token usage for each configuration. This makes it easy to compare how different prompt versions, models, or other changes perform across the same dataset.
+
+The reserved `models`, `prompts`, and `tools` keys automatically populate dedicated columns in the experiments table. Click a value in one of those columns to filter or group by it. For full details, see [Filter and group by models, prompts, and tools](/langsmith/analyze-an-experiment#filter-and-group-by-models-prompts-and-tools-in-the-experiments-tab-view).
+
 ## Explore the results
 
 Each invocation of `evaluate()` creates an [experiment](/langsmith/evaluation-concepts#experiment) that you can view in the LangSmith UI or query via the SDK. See [Analyze an experiment](/langsmith/analyze-an-experiment) for more details.