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

# Terminology

> The canonical vocabulary, and how each term maps to what you see in the CLI and the product today.

JoyStream has a canonical vocabulary for the model, and product nouns for what you type and click. This page maps between them.

## The canonical vocabulary

| Term                | Canonical meaning                                                                        |
| ------------------- | ---------------------------------------------------------------------------------------- |
| **Job**             | The progress a person or stakeholder wants to make                                       |
| **Capability**      | The durable organizational ability to fulfill that Job repeatedly                        |
| **Capability spec** | The declarative definition and trust contract for the capability                         |
| **Agent**           | The governed intelligent realization of a capability spec                                |
| **Runtime recipe**  | The versioned, largely deterministic execution plan used by the agent                    |
| **Agent version**   | An immutable executable realization combining spec, recipe, models, skills, and policies |
| **Deployment**      | The environment-specific activation of an agent version                                  |
| **Run**             | One execution instance of a deployed agent version                                       |
| **Epic**            | The run-specific work graph instantiated from the recipe                                 |
| **Story**           | An individual executable work item within the epic                                       |
| **Ticket**          | The durable, business-readable collaboration and accountability object around work       |
| **Evidence**        | Proof that the intended technical and organizational outcome occurred                    |
| **Memory**          | Governed context and learning accumulated through execution                              |
| **Metrics**         | Measures of technical, operational, and capability performance                           |
| **Workspace**       | The organizational boundary in which capabilities are owned and operated                 |

## How it maps to the product today

The product is younger than the model. Some canonical objects are fully first-class; others are still expressed through a nearer implementation name. This table is the bridge.

| Canonical concept | What you see today                                                                                           |
| ----------------- | ------------------------------------------------------------------------------------------------------------ |
| Capability spec   | The agent definition, authored in `jstm agent plan` and exported by `jstm agent spec`                        |
| Agent             | An agent record, addressed by handle such as `acme-corp/onboarding`                                          |
| Runtime recipe    | The built package and its work tree, seen with `jstm agent view --tree`                                      |
| Agent version     | The package version pinned by `jstm agent build`                                                             |
| Deployment        | An installation plus its stage, credentials, and bindings — `jstm agent install`, `jstm agent installations` |
| Run               | A run, listed by `jstm agent runs` and inspected with `jstm agent run show`                                  |
| Epic              | The work tree for one run                                                                                    |
| Story             | A story, listed by `jstm agent stories`                                                                      |
| Ticket            | A ticket, listed by `jstm ticket list`                                                                       |
| Capability        | Not yet a separate object in the product. Today the agent and its history carry the capability.              |

<Note>
  Where the product and the model diverge, the docs use the product noun in guides and the CLI reference, and the canonical noun in concepts. This page is the join between them.
</Note>

## Lifecycle stages

Always written: **Plan**, **Build**, **Conform**, **Dry Run**, **Pilot**, **Live**.

Dry Run was previously called Sandbox. If you see "sandbox" anywhere, it means Dry Run.

## What is deliberately not a primary object

Several concepts matter but are not the center of the model.

|                  | Why not                                                                                               |
| ---------------- | ----------------------------------------------------------------------------------------------------- |
| **Prompt**       | An input or implementation detail, not a durable organizational asset                                 |
| **Conversation** | An engineering and collaboration interface. It contributes history but does not define the capability |
| **Model**        | Supplies intelligence to an agent version. The capability should survive model replacement            |
| **Tool call**    | One technical action inside a story, not the organizational work item                                 |
| **Log**          | Low-level events. Runs, epics, stories, tickets, and evidence provide the business-readable view      |
| **Job**          | A powerful discovery lens, but not something you navigate to in the product                           |