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

# Agent Workspace

> What exists on the agent details page in the UI: overview, analysis, evals, and configuration.

The agent details page is the operational center for a single agent in WhyOps.

It is organized into four tabs:

* `Overview`
* `Analysis`
* `Evals`
* `Configuration`

## Overview

The overview tab is built for day-to-day runtime monitoring.

It shows:

* agent header and current version context
* summary stats for trace volume and recent activity
* success-rate timeline
* trace-count timeline
* recent traces table

These views are powered by `GET /api/entities/:id` and related trace queries.

## Analysis

The analysis tab is where you run and review **agent analysis runs**.

This tab supports:

* manual analysis runs
* configurable lookback windows
* analysis modes: `quick`, `standard`, `deep`
* custom judge model selection
* dimension selection
* scheduled analysis config using cron + timezone
* streaming checkpoints during execution
* run history and detailed result inspection

### Backend APIs used

* `POST /api/agent-analyses/:agentId/run`
* `GET /api/agent-analyses/:agentId/latest`
* `GET /api/agent-analyses/:agentId/runs`
* `GET /api/agent-analyses/runs/:runId`
* `GET /api/agent-analyses/:agentId/config`
* `PUT /api/agent-analyses/:agentId/config`

### Streaming behavior

If the client requests streaming with `Accept: application/x-ndjson` or `?stream=true`, the backend streams checkpoint snapshots as NDJSON chunks while analysis is running.

That is why the UI can show live progress toasts and in-flight section updates.

## Evals

The evals tab is the test-generation and knowledge-profile workspace.

It includes:

* synthetic eval generation controls
* category selection
* optional custom prompt and judge model
* eval generation schedules
* run history
* export as JSON
* export as Promptfoo YAML
* knowledge profile view
* explicit knowledge profile rebuild

### Backend APIs used

* `POST /api/evals/:agentId/generate`
* `GET /api/evals/:agentId/latest`
* `GET /api/evals/:agentId/runs`
* `GET /api/evals/:agentId/runs/:runId`
* `GET /api/evals/:agentId/cases`
* `GET /api/evals/:agentId/config`
* `PUT /api/evals/:agentId/config`
* `GET /api/evals/:agentId/export/json`
* `GET /api/evals/:agentId/export/promptfoo`
* `GET /api/evals/:agentId/knowledge-profile`
* `POST /api/evals/:agentId/knowledge-profile/rebuild`

### Intelligence-building state

One subtle feature in the backend is that eval generation can return an `intelligence_building` state instead of a finished run.

That happens when WhyOps first needs to gather external knowledge before generating a strong eval suite. The UI already supports this state and tells the user that background knowledge gathering is in progress.

## Configuration

The configuration tab focuses on versioned agent setup and runtime settings.

It lets users inspect:

* current and historical versions
* version hashes
* system prompt
* tool definitions
* raw metadata
* sampling rate
* runtime limits

### Backend APIs used

* `GET /api/entities/:id/version-ids`
* `GET /api/entities/versions/:versionId`
* `PATCH /api/entities/:id/sampling-rate`
* `GET /api/agent-settings/:id`
* `PATCH /api/agent-settings/:id`
* `DELETE /api/agent-settings/:id`
* `GET /api/agent-settings/limits`

## Why this matters

This page is where WhyOps turns raw observability into an operational workflow:

1. inspect current health
2. investigate recent traces
3. run deeper analysis
4. generate regression evals
5. review prompt and tool versions
6. tune runtime settings

It is not just a trace viewer - it is the full lifecycle workspace for one production agent.