Skip to content

Commit c1936d5

Browse files
docs(studio): add Experiments, Insights, Traces, and Model Evaluation… (#1828)
* docs(studio): add Experiments, Insights, Traces, and Model Evaluations pages Covers experiment create/edit, evaluation drill-down, and side-by-side run comparison; Optimizer insight list/detail; Traces/Spans tables and session drill-down with per-span-kind views; and the Model Evaluations results list/detail. These previously had only a one-line mention (or, for Model Evaluations, no page) in docs/studio/index.mdx. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Henrique Tolentino <htolentino@nvidia.com> * docs(studio): fix stale CLI, Pareto axes, and insight-action claims in Experiments/Insights docs - Experiments: CLI command tab runs nemo agents experimentalist run (--train-dataset/--validation-dataset), not the nonexistent nemo exp run; confirmed against nemo-experimentalist's CLI and the matching command builder in InsightOpenModal. - Experiments: Pareto view defaults to cost vs. latency, not accuracy, and axes are configurable to cost_usd, latency_ms, or evaluators.<name>, per ExperimentParetoChart's deriveParetoMetrics. - Insights: Run experiment is shown only for resolved or deleted insights, per insightStatus.ts's action table. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Signed-off-by: Henrique Tolentino <htolentino@nvidia.com> --------- Signed-off-by: Henrique Tolentino <htolentino@nvidia.com> Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
1 parent f37853a commit c1936d5

6 files changed

Lines changed: 187 additions & 6 deletions

File tree

‎docs/fern/versions/latest.yml‎

Lines changed: 8 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -354,12 +354,20 @@ navigation:
354354
path: ../../studio/customization.mdx
355355
- page: Data Designer
356356
path: ../../studio/data-designer-build.mdx
357+
- page: Experiments
358+
path: ../../studio/experiments.mdx
359+
- page: Insights
360+
path: ../../studio/insights.mdx
361+
- page: Model Evaluations
362+
path: ../../studio/model-evaluations.mdx
357363
- page: Monitor
358364
path: ../../studio/monitor.mdx
359365
- page: Plugin UIs
360366
path: ../../studio/plugins.mdx
361367
- page: Guardrail Configs
362368
path: ../../studio/guardrails.mdx
369+
- page: Traces
370+
path: ../../studio/traces.mdx
363371
- page: Virtual Models
364372
path: ../../studio/virtual-models.mdx
365373
- section: Kubernetes Deployment

‎docs/studio/experiments.mdx‎

Lines changed: 49 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,49 @@
1+
---
2+
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3+
# SPDX-License-Identifier: Apache-2.0
4+
5+
title: "NeMo Studio Experiments"
6+
description: "Group evaluations into experiments, drill into results, and compare runs side by side."
7+
---
8+
9+
Use **Evaluations > Experiments** in the NeMo Studio workspace sidebar to group evaluations for comparison, and to drill into evaluation results and individual runs.
10+
11+
<Note>
12+
13+
Experiments is enabled by default. An administrator can disable it by setting `studio.feature_flags.experiment: false` in the platform configuration and restarting Studio.
14+
15+
</Note>
16+
17+
## Create an Experiment
18+
19+
Select **New experiment** from the Experiments list. In the **Create experiment** tab, provide a name, an optional description, and a default sort order for the experiment's evaluations, then select **Create**.
20+
21+
<Note>
22+
23+
The modal also has **NeMo Assistant** and **CLI command** tabs. **CLI command** shows an equivalent `nemo agents experimentalist run` invocation, with `--train-dataset` and `--validation-dataset` selecting the training and validation data; **NeMo Assistant** is a placeholder for a future guided flow.
24+
25+
</Note>
26+
27+
## Experiment Detail
28+
29+
Select an experiment to open its detail page, which shows aggregate metrics across its evaluations and a table of the evaluations in the group. Toggle **Pareto view** to plot evaluations on a chart instead of the table, by default cost against latency. Change either axis to cost, latency, or any evaluator's score (`cost_usd`, `latency_ms`, or `evaluators.<name>`); the selection is saved per experiment.
30+
31+
Select **Edit** to update the experiment's name, description, or default sort order.
32+
33+
If the experiment was created from an Optimizer insight, an **Insight description** card appears with a link back to that insight. See [NeMo Studio Insights](/documentation/studio/insights).
34+
35+
## Evaluation Drill-Down
36+
37+
Select an evaluation from an experiment's table to open its detail page, which shows aggregate metrics for that evaluation and a table of its test-case runs (sessions).
38+
39+
Select a session to open its detail page — the same trace and span view used by **Traces**, scoped to that test case. See [NeMo Studio Traces](/documentation/studio/traces).
40+
41+
## Compare Runs Side by Side
42+
43+
From a session's detail page, select **Compare against…** and choose another run of the same test case. Studio splits the view into two columns — the current run on the left, the selected run on the right — so you can compare trace output side by side. Select **Clear** to return to the single-run view.
44+
45+
## Related Topics
46+
47+
- [Experiments](/documentation/evaluate-models/experiments)
48+
- [NeMo Studio Insights](/documentation/studio/insights)
49+
- [NeMo Studio Traces](/documentation/studio/traces)

‎docs/studio/index.mdx‎

Lines changed: 18 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -32,11 +32,11 @@ The current sidebar groups related pages under expandable parents. Selecting a p
3232

3333
| Sidebar group | Current entry point | Purpose and availability |
3434
| --- | --- | --- |
35-
| **Observability** | **Insights** | Review recurring agent failure patterns and optimization insights. Shown when optimization is enabled. |
36-
| **Observability** | **Traces** | Inspect ingested traces, spans, and session details. |
35+
| **Observability** | **Insights** | Review optimization insights generated from observed agent sessions. Shown when optimization is enabled; see [NeMo Studio Insights](/documentation/studio/insights). |
36+
| **Observability** | **Traces** | Inspect ingested traces, spans, and session details; see [NeMo Studio Traces](/documentation/studio/traces). |
3737
| **Components** | **Agents** | Opens the agent list. Select an agent to access Deployments, Logs, Chat, Evaluations, and Details. |
38-
| **Components** | **Models** | Opens the base-model list directly. Expand **Models** for **Fine-tune**, **Model Evaluations**, **Playground**, and **Virtual Models**, when those capabilities are enabled. **Fine-tune** opens the Custom Models list; see [NeMo Studio Custom Models](/documentation/studio/customization). |
39-
| **Evaluations** | **Experiments** | Review experiments, evaluations, sessions, and published results across the workspace. |
38+
| **Components** | **Models** | Opens the base-model list directly. Expand **Models** for **Fine-tune**, **Model Evaluations**, **Playground**, and **Virtual Models**, when those capabilities are enabled. **Fine-tune** opens the Custom Models list; see [NeMo Studio Custom Models](/documentation/studio/customization). **Model Evaluations** opens evaluation job results; see [NeMo Studio Model Evaluations](/documentation/studio/model-evaluations). |
39+
| **Evaluations** | **Experiments** | Group evaluations for comparison, drill into results, and compare runs side by side; see [NeMo Studio Experiments](/documentation/studio/experiments). |
4040
| **Data** | **Datasets > Data Designer** | Build and monitor synthetic-data jobs. Other installed data plugins can appear below **Datasets**. |
4141
| **Governance** | **Guardrails** | Manage guardrail configurations. Disabled by default; see [Studio Guardrail Configs](/documentation/studio/guardrail-configs). |
4242
| **Governance** | **Iron Swarm** | Plugin-provided agent hardening UI. Appears only when the Iron Swarm plugin is installed; see [Studio Plugin UIs](/documentation/studio/plugins). |
@@ -51,6 +51,18 @@ Use the Studio Agents area to review Platform-managed agents, including the curr
5151

5252
For the full workflow, see [Studio Agents](/documentation/studio/agents).
5353

54+
### Insights
55+
56+
Use **Observability > Insights** to review optimization insights the analyst agent generates from observed agent sessions and traces, and drill into an insight's linked experiments and source traces. For a resolved or deleted insight, **Run experiment** shows the CLI command to start experiments for it.
57+
58+
For the full workflow, see [Studio Insights](/documentation/studio/insights).
59+
60+
### Traces
61+
62+
Use **Observability > Traces** to inspect ingested telemetry in the Traces and Spans tables, and drill into a session's trace and span detail, with specialized views for LLM, tool, agent, chain, embedding, reranker, retriever, guardrail, and evaluator spans.
63+
64+
For the full workflow, see [Studio Traces](/documentation/studio/traces).
65+
5466
### Monitor
5567

5668
Use **Agents \> Monitor** to inspect agent telemetry stored by the platform, including recent inference logs and token usage summaries.
@@ -120,8 +132,8 @@ Studio provides separate agent, model, and workspace evaluation views:
120132
| Location | What it shows |
121133
| -------- | ------------- |
122134
| **Agents > select an agent > Evaluations** | Active evaluation jobs, completed evaluations, and experiments scoped to the selected agent. |
123-
| **Models > Model Evaluations** | Model evaluation results. |
124-
| **Evaluations > Experiments** | Experiments and their evaluations across the workspace. |
135+
| **Models > Model Evaluations** | Model evaluation job results and per-row scores. See [Studio Model Evaluations](/documentation/studio/model-evaluations). |
136+
| **Evaluations > Experiments** | Experiments and their evaluations across the workspace, evaluation drill-down, and side-by-side run comparison. See [Studio Experiments](/documentation/studio/experiments). |
125137

126138
Use the agent's **Evaluations** tab when reviewing how a specific agent performed. Use **Experiments** for the broader workspace history.
127139

‎docs/studio/insights.mdx‎

Lines changed: 42 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,42 @@
1+
---
2+
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3+
# SPDX-License-Identifier: Apache-2.0
4+
5+
title: "NeMo Studio Insights"
6+
description: "Review optimization insights generated from observed agent sessions."
7+
---
8+
9+
Use **Observability > Insights** in the NeMo Studio workspace sidebar to review insights the analyst agent generates from observed agent sessions and traces.
10+
11+
<Note>
12+
13+
Insights is enabled by default. An administrator can disable it by setting `studio.feature_flags.optimizer_enabled: false` in the platform configuration and restarting Studio.
14+
15+
</Note>
16+
17+
## Insight List
18+
19+
The Insights table shows each insight's status, title, source agent, number of observed traces, number of linked experiments, and creation and last-seen times. Select a row to open the insight's detail page.
20+
21+
## Insight Detail
22+
23+
The detail page shows the insight's status, source agent, description, and:
24+
25+
| Section | Shows |
26+
| ------- | ----- |
27+
| Experiments | Experiments linked to this insight, with evaluation counts and last-updated times. Select a row to open that experiment. |
28+
| Traces | The observed sessions and traces that produced this insight. |
29+
30+
Available actions depend on the insight's status:
31+
32+
| Status | Available actions |
33+
| ------ | ------------------ |
34+
| Open | **Delete**, or **Resolve**. |
35+
| Resolved or Deleted | **Run experiment** — reopens the insight and shows a modal with the `nemo exp run` CLI command to start experiments for it. Studio doesn't launch the experiment for you; run the command yourself. |
36+
37+
If the insight originated an experiment, that experiment's detail page links back here. See [NeMo Studio Experiments](/documentation/studio/experiments).
38+
39+
## Related Topics
40+
41+
- [NeMo Studio Experiments](/documentation/studio/experiments)
42+
- [Optimize Agents](/documentation/agents/optimize-agents)

‎docs/studio/model-evaluations.mdx‎

Lines changed: 31 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,31 @@
1+
---
2+
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3+
# SPDX-License-Identifier: Apache-2.0
4+
5+
title: "NeMo Studio Model Evaluations"
6+
description: "Review model evaluation job results and per-row scores."
7+
---
8+
9+
Use **Models > Model Evaluations** in the NeMo Studio workspace sidebar to review evaluation jobs run against models in the workspace.
10+
11+
<Note>
12+
13+
Model Evaluations is enabled by default. An administrator can disable it by setting `studio.feature_flags.evaluator_enabled: false` in the platform configuration and restarting Studio, which also removes agent- and customization-scoped evaluation entry points.
14+
15+
</Note>
16+
17+
This page is read-only: it lists evaluation jobs and their results. Start an evaluation from an agent's **Evaluations** tab, a Custom Model's **Evaluate this Model** action, or the `nemo evaluation` CLI or API.
18+
19+
## Evaluation Results List
20+
21+
The list shows each evaluation job's name, status, and creation time. Search by name or filter by status. Select a row to open the job's detail page.
22+
23+
## Evaluation Result Detail
24+
25+
The detail page shows the job's details, a **Scores** panel with aggregate metrics, per-row results for the evaluated dataset, and a collapsible **Logs** panel. Scores and row results are computed once the job reaches a terminal state; if the job failed, results aren't available.
26+
27+
## Related Topics
28+
29+
- [Model Configuration](/documentation/evaluate-models/metrics/model-configuration)
30+
- [NeMo Studio Custom Models](/documentation/studio/customization)
31+
- [NeMo Studio Agents](/documentation/studio/agents)

‎docs/studio/traces.mdx‎

Lines changed: 39 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,39 @@
1+
---
2+
# SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
3+
# SPDX-License-Identifier: Apache-2.0
4+
5+
title: "NeMo Studio Traces"
6+
description: "Inspect ingested traces, spans, and session detail from agent telemetry."
7+
---
8+
9+
Use **Observability > Traces** in the NeMo Studio workspace sidebar to inspect telemetry ingested from agent runs.
10+
11+
<Note>
12+
13+
Traces is enabled by default. An administrator can disable it by setting `studio.feature_flags.intake_enabled: false` in the platform configuration and restarting Studio.
14+
15+
</Note>
16+
17+
## Traces and Spans Tables
18+
19+
The **Traces** and **Spans** tabs list ingested telemetry at different granularity:
20+
21+
| Tab | Columns |
22+
| --- | ------- |
23+
| Traces | Trace, input, output, duration, span count, errors, tokens, cost, started. |
24+
| Spans | Status, kind, span, input, output, subject, trace, duration, tokens, cost, started. |
25+
26+
Select a row in either table to open the session it belongs to.
27+
28+
## Session Detail
29+
30+
The session detail page shows a summary header, then the selected trace's spans. Use the **Tree** / **List** toggle to switch between a nested span tree and a flat list, and **Collapse all** / **Expand all** to manage the tree view.
31+
32+
Selecting a span opens its detail body. Studio renders a specialized view for known span kinds — LLM, tool, agent, chain, embedding, reranker, retriever, guardrail, and evaluator — and falls back to a generic view for unrecognized kinds. Each span body includes metadata, an annotations panel, and a raw JSON view for debugging.
33+
34+
Evaluation test-case runs reuse this same session detail view; see [Compare Runs Side by Side](/documentation/studio/experiments#compare-runs-side-by-side) in NeMo Studio Experiments.
35+
36+
## Related Topics
37+
38+
- [NeMo Studio Experiments](/documentation/studio/experiments)
39+
- [NeMo Studio Insights](/documentation/studio/insights)

0 commit comments

Comments
 (0)