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

# Library

> Kernels, canvases, workflows, and signals — the building blocks of your analytics.

<Frame>
  <img src="https://mintcdn.com/originhq/BoZcxy2aZcnPe0ai/images/origin-library-1.png?fit=max&auto=format&n=BoZcxy2aZcnPe0ai&q=85&s=8cb060faf067620e4a77fbb69e701d89" alt="Library" lightAlt="Library" darkAlt="Library" className="dark:hidden" width="1440" height="810" data-path="images/origin-library-1.png" />

  <img src="https://mintcdn.com/originhq/BoZcxy2aZcnPe0ai/images/origin-library-2.png?fit=max&auto=format&n=BoZcxy2aZcnPe0ai&q=85&s=2c3da7db516d04629b2009264958bc4a" alt="Library" lightAlt="Library" darkAlt="Library" className="hidden dark:block" width="1440" height="810" data-path="images/origin-library-2.png" />
</Frame>

The **Library** is Origin's analytics building-block store. It contains all the reusable components that power your dashboards and automations: Kernels, Canvases, Workflows, and Signals. Access it by clicking the grid icon (⊞) in the far-left navigation strip.

## Library sections

| Section       | Count type            | What it contains                                           |
| ------------- | --------------------- | ---------------------------------------------------------- |
| **Kernels**   | Deterministic recipes | Data queries that produce a single visualization           |
| **Canvases**  | Multi-widget views    | Collections of Kernels arranged into a dashboard           |
| **Workflows** | Automations           | Triggered sequences that run Kernels and take actions      |
| **Signals**   | Stream triggers       | Real-time event patterns that fire when conditions are met |

## Kernels

Kernels are the fundamental unit of Origin analytics. Each Kernel is a named, versioned query that:

1. Specifies a data source (e.g., `/activity-summary`, `/model-mix`, `/agent-surfaces`)
2. Defines the computation (aggregation, filtering, grouping)
3. Specifies how to render the result (bars, lines, table, metric card)

### Built-in kernels

Origin ships with 14 built-in kernels covering the most common analytics needs:

| Kernel                | Description                                                    |
| --------------------- | -------------------------------------------------------------- |
| `/activity-summary`   | Overall activity counts (prompts, endpoints, sessions, models) |
| `/agent-surfaces`     | Agent surfaces (harnesses) — the interfaces agents use         |
| `/model-mix`          | Top models by prompt count in the current window               |
| `/new-agent-surfaces` | Newly-seen agent surfaces                                      |
| `/new-models`         | Newly-seen models                                              |
| `/new-providers`      | Newly-seen providers                                           |
| `/prompt-volume`      | Prompt volume (daily)                                          |
| `/provider-mix`       | Provider mix — breakdown by AI provider                        |
| `/resource-leaders`   | Files and resources touched most frequently                    |
| `/token-by-model`     | Token spend by model                                           |
| `/token-by-surface`   | Token spend by surface                                         |

### Viewing a kernel

Click any kernel in the left panel to open it. The center panel shows the kernel's output visualization, and the right panel shows the kernel's metadata:

* **Command** — the kernel's identifier (e.g., `/model-mix`)
* **Title** — human-readable name
* **Steps** — number of computation steps
* **Render** — visualization type (bars, lines, table, etc.)
* **Params** — configurable parameters (if any)
* **Updated** — last modification date
* **Description** — what the kernel measures and how to interpret it
* **Used in canvases** — which canvases include this kernel

### Forking a kernel

Built-in kernels are **read-only**. To customize one, click **Fork to Customize**. This creates a copy in your Library that you can edit. Forked kernels are prefixed with your organization name and can be modified without affecting the original.

## Canvases

The Library lists all saved Canvases. Clicking a canvas opens it in the center panel. The Library shows:

* Canvas name and description
* How many Kernels it contains
* Last modified date

Currently there are 5 built-in canvases, including `/ai-usage-overview`, `/adoption-and-concentration`, and `/shadow-ai`.

## Workflows

Workflows are triggered automations. A Workflow runs one or more Kernels on a schedule or in response to a Signal, then takes an action (send a Slack message, create a ticket, update a record). The Workflow builder is available via the **+** button in the Workflows section.

## Signals

Signals are real-time stream triggers. A Signal watches your AI activity stream for a specific pattern — a new agent appearing, a token spend threshold being crossed, a new MCP tool spreading to more than N endpoints — and fires an event when the condition is met. Signals can trigger Workflows.

<Tip>
  The fastest way to build a custom analytics view is to find the closest built-in kernel, fork it, adjust the query parameters, and add it to a new Canvas. The Fork to Customize button appears in the right panel when you select any built-in kernel.
</Tip>