> ## Documentation Index
> Fetch the complete documentation index at: https://openlayer.com/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# AI-generated summaries

> AI summaries for sessions, traces, and test results, so you don't reconstruct what happened span by span.

Openlayer generates **AI summaries** on three surfaces, so you can understand
what happened without reading raw spans or row-level results:

* **Sessions** — the session detail view opens with a summary of how the
  session actually went across its traces: what the user was trying to do,
  what the agent did, and where it struggled.
* **Traces** — the trace details view summarizes the trace instead of requiring
  line-by-line span reading.
* **Test results** — results include a summary of the dominant failure modes.

## Where to find them

* **Session summary**: open a project with tracing → **Sessions** → open a
  session. The summary appears at the top of the session detail view.
* **Trace summary**: open a trace from the **Traces** view. The trace details
  view has an **AI summary** button — select it to summarize, analyze, and get
  up to speed with the trace.
* **Test-result summary**: open a test result — the summary sits above the
  row-level results.

Summaries are generated on demand and cached, so opening the same session or
result again is instant.

## Configure AI features

AI features are configured at the workspace level, and each project can
override the workspace default.

### Workspace settings

Go to **Settings** → **AI features** to control which AI-powered features are
enabled in the workspace. There is one section per feature:

* **Trace summaries** — summaries for individual traces.
* **Session summaries** — summaries for user sessions.
* **Test summaries** — summaries for test result sets.
* **Intake assistant** — draft project details, intake answers, and risk scores
  during project registration or from the project's **Lifecycle** page. See
  [Project lifecycle](/docs/governance/project-lifecycle#draft-intake-with-the-ai-assistant)
  for context file limits and draft behavior.

Each section carries the same controls:

| Control | What it does |
| - | - |
| **Enable** | Turns the feature on for the workspace |
| **Locked for all projects** | Stops projects from turning the feature on or off. Per-project model overrides still apply |
| **Model** | Selects the model that the feature uses |

**Test summaries** adds **Generate automatically**, which summarizes test
results as they finish computing.

The same page holds two more sections:

* **Semantic search** — matches test subpopulations by meaning instead of exact
  text, and requires a ClickHouse-backed pipeline. You can lock it for all
  projects; the **Embedding model** is set per deployment via `EMBEDDING_MODEL`,
  and changing it requires re-embedding stored data, because existing vectors
  are not comparable across models. See
  [semantic search filters](/docs/tests/semantic-search-filters) for how the filters
  themselves work.
* **AI usage** — spend and volume across every project in the workspace, broken
  down by feature, including **Intake assistant**, over the time range you pick.
  Spend is estimated with the token prices in **Workspace settings** → **LLM
  costs**, covered in [LLM cost estimation](/docs/monitoring/cost-estimation).

### Project settings

Open a project and go to **Settings** → **AI features** to control which
AI features run for that project. It mirrors the workspace page — **Trace
summaries**, **Session summaries**, **Test summaries**, and **Intake
assistant** — and project settings override the workspace default.

When a feature is locked at the workspace level, its **Enable** toggle is
unavailable in the project and only the **Model** remains editable.

For **Intake assistant**, add the selected model provider's API key under the
workspace or project [**Settings** → **Environment**](/docs/security/environment-variables).
Openlayer does not run the assistant with its own provider keys.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.