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

> ## Agent Instructions
> To build and serve your own model, start at /deploy/development/quickstart and /deploy/development/overview. Deploying is the default path: reactor init scaffolds a workspace, reactor auth login authenticates, and reactor model deploy registers the model, publishes the release with the weights/ folder, and activates it on Reactor's GPUs, in one command from that workspace. Docker must be running, because the publish step builds the image locally. Bump model.version in reactor.yaml before redeploying a change, because a release that already has an image is reactivated as it is. Deployment access is granted per account, so contact team@reactor.inc if a deploy is refused. Every key in reactor.yaml is documented at /deploy/platform/reactor-yaml. Model code imports reactor_runtime; Python client code imports reactor_sdk. The runtime overview explains the model interface. Running the model on your own machine with reactor run is optional and needs a GPU you attach with --gpus; /deploy/development/local-testing covers that loop and pairs a complete brightness model with a Python client test in a separate brightness-test workspace.
> Reactor hosts multiple models, each with its own connect slug (modelName) and command/event schema. The video model catalog — slug, typed SDK package, and links to its schema — is at /model-api-reference/overview. Robotics policy documentation starts at /robotics/overview; The launch policies are Cosmos Nano DROID (/robotics/cosmos/nano-policy-droid/), DreamZero DROID (/robotics/dreamzero/droid/), and FLUX Action DROID (/robotics/flux-action/droid/), organized by model family and embodiment. Some models expose one slug per experience (e.g. HappyOyster); always take the slug from the model's own pages, never guess it.
> Fastest path to a working app: `npx create-reactor-app my-app --model=<slug>` scaffolds a complete app with secure auth wired up. Typed TypeScript SDKs are published as @reactor-models/<model>; Python uses the base reactor-sdk package.
> Auth: exchange an API key (rk_...) for a JWT via POST https://api.reactor.inc/tokens from your server. Never put the API key in client-side code.
> Append .md to any docs URL for clean Markdown. Search these docs via the MCP server at https://docs.reactor.inc/mcp.

# Genie 4 prompt guide

> Write opening scenes and instructions for Genie 4, with examples for images, objects, and dialogue.

Start with a description of the place you want to explore. You can use an image to set the opening
view, then add instructions as you explore.

## Describe the opening scene

Use your opening prompt to establish the scene. Save actions, such as picking up an object or
speaking to a character, for follow-up instructions. Opening prompts can be up to 2,000 characters.

Use "first-person view" to place the viewer inside the scene. Include details that matter to the
world you want to create:

* Name the place and choose a visual style, such as photorealistic or watercolor.
* Describe the time of day, lighting, and weather.
* Place important objects relative to the viewer and describe their starting state.

A nice café by the sea.

First-person view inside a small seaside café, facing the counter. Warm afternoon light falls
through windows that overlook the harbor. An empty white cup rests on a wooden table to my right. A
barista stands beside a silver espresso machine. Photorealistic style.

## Use a reference image

Choose an image that shows the opening view you want. Use a PNG, JPEG, or WEBP up to 20 MiB and 40
megapixels.

A 16:9 landscape image fits the opening view. Other shapes are center-cropped to 16:9, so keep the
subject away from the edges. Use the image on its own. For more detail, pair it with a prompt about
the time of day or weather.

## Change one thing at a time

During a live world, describe the change you want to make. Instructions can be up to 2,000
characters in length. Name what changes and what stays the same.

Change the weather and make it more dramatic.

Rain starts outside the café windows. Keep the lighting inside the café unchanged.

Let each change appear before adding another instruction. A new instruction replaces one that is
still in progress.

## Interact with Genie 4

Use follow-up instructions to interact with the objects and characters in your scene.

### Interact with objects

Name the object and describe one action. Establish the object in your opening scene, then refer to
it by the same name in later instructions.

> Pick up the white cup from the table to my right.

For another action, refer to the result of the first:

> Turn the cup so I can see its handle.

### Talk to characters

Name the character and ask a question, or give them a short line to say.

> I ask the barista, "What do you recommend?"

For other sounds, name the source and what it does:

> The espresso machine hisses as the barista steams milk.

### Change the setting

Describe the new surroundings and name the details you want to keep.

> Dusk falls over the harbor and the lights along the waterfront turn on. Keep the café interior
> unchanged.

## Know the limits

Use a first-person viewpoint when you want to navigate. Third-person controls are less reliable.
Dialogue can repeat or become difficult to understand. Keep each request short and leave time for a
response.

Genie 4 may move objects or deliver a line differently from your instructions. It may decline
prompts about real people.

## See also

* [Genie 4 overview](https://docs.reactor.inc/model-api-reference/genie-4/overview.md)
* [API schema](https://docs.reactor.inc/model-api-reference/genie-4/schema.md)
* [Build a world explorer](https://docs.reactor.inc/model-api-reference/genie-4/tutorial.md)
* [Sessions and connections](https://docs.reactor.inc/concepts/sessions.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.