docs: add Launchpad for AI documentation section (#10680)

_partials/launchpad-for-ai/_unreleased-banner.mdx

@@ -0,0 +1,12 @@
+---
+partial_category: launchpad-for-ai
+partial_name: unreleased-banner
+---
+
+:::caution
+
+This is unreleased documentation for Launchpad for AI. The product is not yet generally available.
+
+Documentation is subject to change before release. We recommend against bookmarking as the link will change when the product is generally available.
+
+:::

docs/docs-content/launchpad-for-ai/_category_.json

@@ -0,0 +1,4 @@
+{
+  "position": 1000,
+  "className": "hidden-category"
+}

docs/docs-content/launchpad-for-ai/explanation/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Explanation",
+  "position": 30,
+  "link": {
+    "type": "doc",
+    "id": "launchpad-for-ai/explanation/explanation"
+  }
+}

docs/docs-content/launchpad-for-ai/explanation/architecture.md

@@ -0,0 +1,39 @@
+---
+sidebar_label: "Architecture Overview"
+title: "Launchpad for AI Architecture Overview"
+description:
+  "An explanation of the Launchpad for AI architecture, including its component stack, data flow, and network topology."
+hide_table_of_contents: false
+sidebar_position: 1
+tags: ["launchpad-for-ai", "architecture", "explanation"]
+keywords: ["launchpad", "ai", "architecture", "kubernetes", "kairos", "helm", "data flow"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+This page explains how Launchpad for AI works, how its components interact, and what key decisions shaped the design.
+Use this page to build an understanding of the architecture before you deploy or operate Launchpad for AI.
+
+## Component Stack
+
+The appliance serves each model through an inference engine, such as vLLM, running on its Kubernetes cluster. Each
+loaded model is exposed as an OpenAI-compatible endpoint, such as `/v1/chat/completions` and `/v1/models`.
+
+## Appliance and Cluster Formation
+
+## Model Provisioning Lifecycle
+
+When you deploy a model, the appliance places it automatically on the best-fit node and brings it through a guarded
+sequence of gate, provision, smoke-test, and ready stages. A model is routable only after its smoke test passes, shown
+as `serving · smoke-test passed`, so the console never shows a model as ready before it is serving.
+
+## Request Routing
+
+The gateway routes each request to a model. A request that names a model uses that model, and a request that does not
+name a model falls back to the default model. When you change the default model, the gateway rebuilds its router in
+place. The gateway does not restart, and it does not drain requests that are in progress. Requests that the gateway
+already routed continue on their assigned model, and the new default applies only to later requests.
+
+## Network Topology
+
+## Data Residency and Isolation

docs/docs-content/launchpad-for-ai/explanation/explanation.md

@@ -0,0 +1,19 @@
+---
+sidebar_label: "Explanation"
+title: "Launchpad for AI Explanation"
+description: "Background, context, and design rationale for understanding how Launchpad for AI works."
+hide_table_of_contents: false
+sidebar_position: 0
+tags: ["launchpad-for-ai", "explanation"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+Explanatory and conceptual guides help you understand how and why Launchpad for AI works the way it does. They cover
+design decisions, component relationships, and trade-offs rather than walking you through tasks.
+
+## Contents
+
+| **Topic**                                  | **What you understand**                                                                       |
+| ------------------------------------------ | --------------------------------------------------------------------------------------------- |
+| [Architecture Overview](./architecture.md) | The component stack, request routing, model provisioning lifecycle, and data residency model. |

docs/docs-content/launchpad-for-ai/how-to-guides/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "How-to Guides",
+  "position": 20,
+  "link": {
+    "type": "doc",
+    "id": "launchpad-for-ai/how-to-guides/how-to-guides"
+  }
+}

docs/docs-content/launchpad-for-ai/how-to-guides/add-a-model.md

@@ -0,0 +1,64 @@
+---
+sidebar_label: "Add a Model"
+title: "Add a Model"
+description:
+  "Step-by-step guidance for platform operators on how to add an LLM model to a running Launchpad for AI appliance and
+  verify that it is serving."
+hide_table_of_contents: false
+sidebar_position: 1
+tags: ["launchpad-for-ai", "models", "how-to"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+This guide explains how to add a model to a running Launchpad for AI appliance and verify that the model is serving
+requests. For background on the appliance and how it routes requests, refer to
+[What is Launchpad for AI?](../launchpad-for-ai.md) and [Architecture](../explanation/architecture.md).
+
+Launchpad for AI provides a set of recommended models, including GLM, DeepSeek, and Kimi. You can deploy models beyond
+the recommended list, but any model must fit within the GPU resources available on the appliance. Before you select a
+model, check the [Supported LLM Models](../reference/supported-models.md) and
+[Hardware Requirements](../reference/hardware-requirements.md) reference pages to confirm that your appliance can
+support it.
+
+## Prerequisites
+
+- A running Launchpad for AI appliance, with the admin console reachable and operator access.
+- At least one node with free capacity for the model you intend to add.
+- The desired model present in the appliance catalog.
+
+## Add a Model
+
+1. From the left main menu, select **Orchestration**, and then select the **Fleet** tab.
+
+2. In the **Deploy new model** section, open the model drop-down menu and select the model to add.
+
+3. (Optional) Open the engine drop-down menu and select an engine. Leave it on the automatic option to let the appliance
+   choose the engine.
+
+4. Confirm that the placement line shows a best-fit node. If the card shows **Deploy held** with a reason, such as no
+   node with enough free GPUs, resolve that reason before you continue.
+
+5. Select **Deploy**, review the deployment summary, and then confirm.
+
+   The appliance selects the node and brings the model online for you. For how placement and provisioning work, refer to
+   [Architecture](../explanation/architecture.md).
+
+### Verify the Model Is Available
+
+Confirm the model is serving before you route traffic to it.
+
+1. Stay on the _Fleet_ tab and locate the model in the _Fleet models_ table.
+
+2. Confirm that the model state reads `ready`. An amber state means the model is still provisioning or running its smoke
+   test, and a red state means the model failed.
+
+3. Confirm that the model detail reads `serving · smoke-test passed` and that the model is marked routable.
+
+For why a model becomes routable only after its smoke test passes, refer to
+[Architecture](../explanation/architecture.md).
+
+## Next Steps
+
+If a request does not name a model, the appliance routes it to the default model. Refer to
+[Set the Default Model](./set-the-default-model.md) to learn how to configure the default.

docs/docs-content/launchpad-for-ai/how-to-guides/how-to-guides.md

@@ -0,0 +1,21 @@
+---
+sidebar_label: "How-to Guides"
+title: "Launchpad for AI How-to Guides"
+description: "Step-by-step guides for completing specific operational tasks on a running Launchpad for AI appliance."
+hide_table_of_contents: false
+sidebar_position: 0
+tags: ["launchpad-for-ai", "how-to"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+How-to guides get a specific job done on a running appliance. They assume you know what you want to accomplish and give
+you the steps to do it without teaching background concepts.
+
+## Contents
+
+| **Guide**                                           | **What you do**                                                                 |
+| --------------------------------------------------- | ------------------------------------------------------------------------------- |
+| [Install the Appliance](./install-the-appliance.md) | Flash the installer ISO, boot the hardware, and bring up the appliance console. |
+| [Add a Model](./add-a-model.md)                     | Deploy a new LLM to the appliance fleet and verify it is serving.               |
+| [Set the Default Model](./set-the-default-model.md) | Configure which model handles requests that do not name a model explicitly.     |

docs/docs-content/launchpad-for-ai/how-to-guides/install-the-appliance.md

@@ -0,0 +1,25 @@
+---
+sidebar_label: "Install the Appliance"
+title: "Install the Launchpad for AI Appliance"
+description:
+  "Step-by-step guidance for platform operators on how to install the Launchpad for AI appliance from bare hardware to a
+  running, reachable appliance console."
+hide_table_of_contents: false
+sidebar_position: 0
+tags: ["launchpad-for-ai", "install", "how-to"]
+keywords: ["launchpad", "ai", "install", "appliance", "hardware", "iso", "edge", "local ui"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+This guide walks you through installing the Launchpad for AI appliance on bare hardware, from flashing the installer ISO
+to verifying that the appliance console is reachable. The step-by-step content for this guide is actively being drafted.
+Review the working outline at
+[Install Guide — working outline](https://docs.google.com/document/d/1ycWX6JAbWhDS-jcXyZ-Vuyc-2Dg5r9iJNzVy9Kh9qBY/edit?usp=sharing)
+while the content is finalized.
+
+## Next Steps
+
+- **Run your first model:** Once the appliance console is reachable, follow the
+  [Run Your First Model and Send Your First Prompt](../tutorials/run-first-model.md) tutorial to deploy a model and send
+  your first inference prompt.

docs/docs-content/launchpad-for-ai/how-to-guides/set-the-default-model.md

@@ -0,0 +1,39 @@
+---
+sidebar_label: "Set the Default Model"
+title: "Set the Default Model"
+description:
+  "Step-by-step guidance for platform operators on how to set the default model that handles unrouted requests on a
+  running Launchpad for AI appliance."
+hide_table_of_contents: false
+sidebar_position: 2
+tags: ["launchpad-for-ai", "models", "how-to"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+This guide explains how to set the default model on a running Launchpad for AI appliance. If a request does not name a
+model, the appliance routes it to the default model. For how the appliance routes requests, refer to
+[Architecture](../explanation/architecture.md).
+
+## Prerequisites
+
+- A running Launchpad for AI appliance, with the admin console reachable and operator access.
+- The model you want to make default already added and serving. To add and verify a model, refer to
+  [Add a Model](./add-a-model.md).
+
+## Set the Default Model
+
+Set the default model from the _Control Room_. You can only select a model that the appliance currently serves.
+
+1. From the left main menu, select **Control Room**.
+
+2. Open the **switch default model** drop-down menu and select the model to make default.
+
+3. Select **Apply fix**, and then confirm.
+
+For what happens to requests that are in progress when you change the default model, refer to
+[Architecture](../explanation/architecture.md).
+
+## Next Steps
+
+To add another model to the appliance, refer to [Add a Model](./add-a-model.md).

docs/docs-content/launchpad-for-ai/launchpad-for-ai.md

@@ -0,0 +1,95 @@
+---
+id: overview
+title: What is Launchpad for AI?
+description: >
+  Launchpad for AI is a standalone, turnkey AI appliance that lets enterprises run large language models on-premises
+  without cloud dependency, AI consulting, or complex infrastructure setup.
+sidebar_label: Overview
+sidebar_position: 1
+tags:
+  - launchpad-for-ai
+  - overview
+  - explanation
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+Launchpad for AI is a turnkey appliance for running large language models (LLMs) on your hardware. It deploys as an
+image with no Palette or PaletteAI dependency. Once a model is loaded, inference runs on the appliance, so data stays in
+your environment and per-token API costs become predictable infrastructure spend.
+
+## The Problem It Solves
+
+Cloud-hosted AI is not the right fit for enterprises that need to keep data on-premises, whether to meet data residency,
+regulatory compliance, and air-gapped network requirements, or to control latency and avoid per-token API costs.
+
+Building an on-premises AI stack from scratch means assembling GPU compute, OS, Kubernetes, an LLM inference runtime,
+authentication, Role-Based Access Control (RBAC), and observability. Launchpad for AI delivers the whole stack,
+pre-integrated, as a single bootable artifact.
+
+## Predictable AI Costs
+
+Cloud AI services bill per token, so costs scale directly with usage and become difficult to budget at enterprise scale.
+Running inference on-premises changes that billing model. Once the hardware is provisioned, inference cost is a fixed
+infrastructure line item regardless of token volume, making AI spend predictable and independent of how heavily the
+system is used.
+
+## What It Includes
+
+The appliance ships with the following pre-integrated components.
+
+| **Layer**             | **Technology**                                  |
+| --------------------- | ----------------------------------------------- |
+| Operating system      | [Kairos](https://kairos.io) on Ubuntu 24.04     |
+| Orchestration         | Kubernetes                                      |
+| LLM inference runtime | vLLM                                            |
+| Intelligent routing   | Routing by task type and data sensitivity       |
+| Local models          | GLM, DeepSeek, Kimi                             |
+| Platform services     | Authentication, RBAC, monitoring, observability |
+| GPU support           | NVIDIA                                          |
+
+The OS layer runs Kairos on Ubuntu 24.04, an immutable Linux distribution designed for appliance deployments. Its
+read-only runtime prevents configuration drift and keeps the appliance in a known, reproducible state. Kubernetes
+manages the lifecycle of containerized workloads on top, handling scheduling, scaling, and health recovery.
+
+vLLM serves language models with high GPU throughput and handles concurrent requests from multiple users. The appliance
+ships GLM, DeepSeek, and Kimi as local open-weight models that run entirely on-premises with no external API calls.
+NVIDIA GPU support provides the parallel processing that large language models require to respond at production speed.
+
+Intelligent routing directs each request to the most appropriate model. Requests that involve private data or require
+low latency stay local. Other requests can route outbound when the network allows.
+
+Platform services cover authentication, RBAC, monitoring, and observability. These surface health and usage data and
+ensure the appliance behaves as a managed enterprise system rather than a raw inference server.
+
+The stack is packaged as Helm charts, which bundle each component as a versioned unit. You can update individual layers
+independently without replacing the entire appliance image.
+
+## Launchpad for AI or PaletteAI
+
+Launchpad for AI and [PaletteAI](https://docs.palette-ai.com) are related but distinct products that serve different
+scales and deployment models.
+
+|                       | **Launchpad for AI**                            | **PaletteAI**                                                           |
+| --------------------- | ----------------------------------------------- | ----------------------------------------------------------------------- |
+| **Form factor**       | Standalone appliance (bootable ISO)             | Software platform (Helm chart or All-in-One ISO on existing Kubernetes) |
+| **Requires Palette?** | No                                              | Only if Palette also manages the underlying cluster                     |
+| **Primary users**     | Platform engineering and IT teams               | Platform engineering teams                                              |
+| **Scale**             | Single-site, on-premises, air-gapped            | Multi-cluster, multi-tenant, cloud and data center                      |
+| **AI workload model** | Local LLM inference with optional cloud routing | Full AI factory: GPU-as-a-Service, Model-as-a-Service, AI Studio        |
+
+## What Launchpad for AI Is Not
+
+- Not a managed cloud service. You own and operate the appliance.
+- Not a Palette add-on. No Palette tenant or license required.
+- Not a general-purpose Kubernetes platform. It is for LLM inference only.
+
+## Next Steps
+
+- **Check hardware requirements:** Review the [Hardware Requirements](/launchpad-for-ai/reference/hardware-requirements)
+  reference before procuring or preparing your hardware.
+- **Install the appliance:** Follow the
+  [Install the Launchpad for AI Appliance](/launchpad-for-ai/how-to-guides/install-the-appliance) guide to go from bare
+  hardware to a running appliance with the UI accessible.
+- **Run your first model:** Follow the [Run Your First Model](/launchpad-for-ai/tutorials/run-first-model) tutorial to
+  deploy a model and send your first prompt.

docs/docs-content/launchpad-for-ai/reference/_category_.json

@@ -0,0 +1,8 @@
+{
+  "label": "Reference",
+  "position": 40,
+  "link": {
+    "type": "doc",
+    "id": "launchpad-for-ai/reference/reference"
+  }
+}

docs/docs-content/launchpad-for-ai/reference/hardware-requirements.md

@@ -0,0 +1,20 @@
+---
+id: hardware-requirements
+title: Launchpad for AI Hardware Requirements
+description: >
+  Detailed hardware requirements for deploying Launchpad for AI, including per-model GPU VRAM needs and multi-node
+  cluster sizing.
+sidebar_label: Hardware Requirements
+sidebar_position: 2
+tags:
+  - launchpad-for-ai
+  - reference
+  - requirements
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+<!-- TODO: DOC-2921 — populate this page with confirmed hardware requirements from engineering -->
+
+This page is a placeholder for the full Launchpad for AI hardware requirements reference. Content will be added when
+ready.

docs/docs-content/launchpad-for-ai/reference/reference.md

@@ -0,0 +1,21 @@
+---
+sidebar_label: "Reference"
+title: "Launchpad for AI Reference"
+description: "Authoritative technical reference for requirements, supported models, and specifications."
+hide_table_of_contents: false
+sidebar_position: 0
+tags: ["launchpad-for-ai", "reference"]
+---
+
+<PartialsComponent category="launchpad-for-ai" name="unreleased-banner" />
+
+Reference pages give you technical information when you need to look something up. They describe what exists and how it
+is configured, not how to accomplish a task.
+
+## Contents
+
+| **Reference**                                       | **What it covers**                                                                         |
+| --------------------------------------------------- | ------------------------------------------------------------------------------------------ |
+| [System Requirements](./system-requirements.md)     | Hardware, software, and network prerequisites for deploying the appliance.                 |
+| [Hardware Requirements](./hardware-requirements.md) | Per-model GPU VRAM requirements and multi-node cluster sizing.                             |
+| [Supported Models](./supported-models.md)           | Full list of supported LLMs with parameter counts, VRAM minimums, and quantization levels. |