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

# Teams

> Hierarchical organization of teams and projects

Teams are the core organizational unit in Chamber. They represent teams, projects, or business units that consume GPU resources.

## What is a Team?

A team is a named entity that:

* Belongs to an organization
* Can have parent and child teams (forming a hierarchy)
* Receives capacity reservations from pools
* Contains workloads

Think of teams as your org chart mapped to GPU resource allocation.

## Hierarchy

Teams form a tree structure that mirrors your organization:

```mermaid theme={null}
flowchart TD
    Research["Research (root team)"]
    Research --> CV["Computer Vision"]
    Research --> NLP
    Research --> Robotics
    CV --> OD["Object Detection"]
    CV --> IG["Image Generation"]
    NLP --> LLM["LLM Training"]
    NLP --> SR["Speech Recognition"]
```

### Why Hierarchies Matter

<CardGroup cols={2}>
  <Card title="Capacity Inheritance" icon="arrow-down">
    Child teams (aka Projects) can use parent capacity when their own is exhausted
  </Card>

  <Card title="Visibility" icon="eye">
    Managers see aggregate utilization for their entire subtree
  </Card>

  <Card title="Delegation" icon="users">
    Team leads manage their subtree without org-wide access
  </Card>

  <Card title="Flexibility" icon="shuffle">
    Restructure teams without recreating reservations
  </Card>
</CardGroup>

## Capacity and Workloads

### Reservations

Each team can have reservations in one or more capacity pools:

```mermaid theme={null}
flowchart LR
    Team["Team: Computer Vision"]
    Team --> West["us-west-cluster\n 64 GPUs reserved"]
    Team --> East["us-east-cluster\n 32 GPUs reserved"]
```

### Workloads

Workloads are associated with exactly one team via labels:

```yaml theme={null}
metadata:
  labels:
    chamber.io/team: computer-vision
```

<Warning>
  Workloads without a team label will be rejected by the Chamber scheduler.
</Warning>

## Best Practices

<AccordionGroup>
  <Accordion title="Start shallow, go deep as needed">
    Begin with a flat structure (one team per major team). Add hierarchy only when you need finer-grained capacity control.
  </Accordion>

  <Accordion title="Use meaningful slugs">
    Slugs appear in workload labels and CLI commands. Use short, descriptive slugs like `cv-team` instead of `computer-vision-research-team-2024`.
  </Accordion>

  <Accordion title="Align with reporting needs">
    Structure teams to match how you want to see utilization reports. If you report by project, make projects teams.
  </Accordion>
</AccordionGroup>

## Related Concepts

<CardGroup cols={2}>
  <Card title="Reservations" icon="bookmark" href="/concepts/reservations">
    How capacity is allocated to teams
  </Card>

  <Card title="Workload Classes" icon="layer-group" href="/concepts/workload-classes">
    Reserved vs elastic workloads
  </Card>
</CardGroup>