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

# Creating a Project

> Create a labeling project in PixlHub: pick the right tool type, understand what gets provisioned, and set yourself up before the first asset is uploaded.

A project is the unit everything else in PixlHub hangs off: its assets, its label schema, its workflow, its team and its exports. Creating one takes thirty seconds; picking the right tool type deserves slightly more thought, because it decides which annotation editor your team will live in.

<Steps>
  <Step title="Open the dialog">
    Go to **Projects** and click **New Project** (also available from the Dashboard's Quick Actions).
  </Step>

  <Step title="Name and describe it">
    The name appears everywhere your team works, so favor descriptive over clever: "Street Scene Detection" beats "CV-Test-2". The description is optional but shows on the project header, a good place for one sentence about scope.
  </Step>

  <Step title="Pick the tool type">
    This selects the editor for every task in the project:

    | Tool type               | Editor you get                                                    |
    | ----------------------- | ----------------------------------------------------------------- |
    | **Image**               | The 2D canvas: boxes, polygons, points, lines, freeform, comments |
    | **LiDAR / Point Cloud** | The 3D point cloud environment                                    |
    | **Text**                | Text annotation surface for classification and span labeling      |
    | **RLHF**                | Response comparison and feedback workflows for LLM data           |
  </Step>

  <Step title="Create">
    PixlHub provisions the project with a default workflow (Label, then Review, then Done) and opens the project overview.
  </Step>
</Steps>

<Frame caption="The Create Project dialog: name, optional description, and the tool type that determines the editor.">
  <img src="https://mintcdn.com/pixldata/yqUvWeFtQIe1UIvn/images/docs/projects/new-project-modal.jpg?fit=max&auto=format&n=yqUvWeFtQIe1UIvn&q=85&s=438f060555fbacc64eccf5d4703508ad" alt="PixlHub Create Project dialog with name, description and tool type options" width="627" height="447" data-path="images/docs/projects/new-project-modal.jpg" />
</Frame>

<Warning>
  Tool type is a structural choice, not a setting: it shapes the editor and task flow for the whole project. Mixed datasets (say, images plus text) belong in separate projects.
</Warning>

## What a new project contains

Straight after creation you land on the project header with nine tabs. You will touch three of them before any labeling happens:

* **[Label Schema](/guides/essentials/label-schema)**, to define your classes before anyone draws a shape.
* **[Assets](/guides/essentials/assets)**, to upload the data; each asset automatically becomes a task.
* **[Workflow](/guides/projects/workflow)**, already running the default Label, Review, Done pipeline, editable when you need more.

The remaining tabs earn their keep once work starts: Tasks for queue management, Performance for throughput, Comments for discussion, Export for getting data out, and Settings for the project's own configuration.

## Naming conventions that age well

Teams running many projects converge on the same pattern: what the data is plus what the model needs, for example "Retail Shelf Audit" or "Highway LiDAR, Q3 batch". Status words in names ("final", "new", "v2-fixed") rot within a week; the workflow and task statuses already track state for you.

## Next steps

<CardGroup cols={2}>
  <Card title="Define the label schema" icon="tags" href="/guides/essentials/label-schema">
    Classes, colors, hotkeys and attributes, before the first annotation.
  </Card>

  <Card title="Upload assets" icon="upload" href="/guides/essentials/assets">
    Formats, limits, batches, and what happens after upload.
  </Card>
</CardGroup>