> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/Crosstalk-Solutions/project-nomad/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Assistant

> Local AI chat powered by Ollama — runs entirely on your hardware with no data sent to any external service.

The AI Assistant is a built-in conversational AI chat interface powered by [Ollama](https://ollama.com/). It runs entirely on your N.O.M.A.D. server — no internet connection is required, and no data is ever sent to an external service. Your conversations stay on your hardware.

<Note>
  The AI Assistant must be installed before you can use it. Enable it during the Easy Setup wizard or install it from **Settings → Apps**.
</Note>

## What it can do

<CardGroup cols={2}>
  <Card title="Answer questions" icon="circle-question">
    Ask about any topic — history, science, practical skills, or current events up to the model's training cutoff.
  </Card>

  <Card title="Explain concepts" icon="lightbulb">
    Get plain-language explanations of complex or technical subjects.
  </Card>

  <Card title="Help with writing" icon="pen-to-square">
    Draft, edit, proofread, or reformat text for any purpose.
  </Card>

  <Card title="Brainstorm ideas" icon="brain">
    Work through problems, generate options, or think out loud with an interactive partner.
  </Card>
</CardGroup>

When the [Knowledge Base](/features/knowledge-base) is set up, the AI will also automatically reference documents you've uploaded when they're relevant to your question.

## How to access

Open **AI Chat** from the Command Center home screen, or navigate directly to `/chat`.

## How to install

<Steps>
  <Step title="Open the setup wizard or Apps page">
    If this is your first time setting up N.O.M.A.D., use the Easy Setup wizard — the AI Assistant is one of the capabilities you can enable on the first step.

    To install it later, go to **Settings → Apps**.
  </Step>

  <Step title="Install the AI Assistant">
    Select the AI Assistant and start the installation. N.O.M.A.D. will pull the required Ollama container and configure it automatically.
  </Step>

  <Step title="Wait for installation to complete">
    Installation progress is shown in the Apps page. Once complete, the AI Chat link becomes available from the Command Center home screen.
  </Step>
</Steps>

## GPU acceleration

Local AI inference is computationally intensive. On CPU only, expect roughly 10–15 tokens per second. With an NVIDIA GPU and the NVIDIA Container Toolkit installed, that typically improves to 100+ tokens per second — a **10–20x improvement** in response speed.

### How to enable GPU acceleration

<Steps>
  <Step title="Install an NVIDIA GPU">
    If your server does not already have a compatible NVIDIA GPU, install one. More VRAM allows you to run larger, more capable models.
  </Step>

  <Step title="Install the NVIDIA Container Toolkit on the host">
    Follow the [official NVIDIA Container Toolkit installation guide](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/install-guide.html) on your server's operating system.
  </Step>

  <Step title="Force Reinstall the AI Assistant">
    Go to **Settings → Apps**, find the AI Assistant, and click **Force Reinstall**. N.O.M.A.D. will detect the GPU during reinstallation and configure the AI container to use it. You'll see "NVIDIA container runtime detected" in the installation progress log.
  </Step>
</Steps>

<Tip>
  Run a System Benchmark before and after adding GPU acceleration to see the difference in your NOMAD Score.
</Tip>

<Warning>
  If you add or swap a GPU but skip the Force Reinstall step, the AI will continue running on CPU only. Always Force Reinstall after any GPU change.
</Warning>

If a GPU is detected on the host but is not accessible inside the AI container, N.O.M.A.D. will show a warning banner on the System Information and AI Settings pages. Click **Fix: Reinstall AI Assistant** to resolve it — this recreates the container without affecting your downloaded models.

## Supported models

The AI Assistant supports any model available through Ollama. You can browse, download, and manage models from **Settings → Models** (`/settings/models`). The full catalog of available models is listed at [ollama.com/library](https://ollama.com/library).

Model size affects both speed and capability:

* **Smaller models (1B–3B parameters)** — fastest responses, good for simple Q\&A
* **Mid-size models (7B–8B parameters)** — balanced speed and quality, recommended for most use cases
* **Larger models (14B–70B+ parameters)** — highest quality, require significantly more RAM and benefit most from GPU acceleration

<Note>
  AI model files range from approximately 1–40GB depending on the model. Plan your storage accordingly.
</Note>

## Tips for best results

<AccordionGroup>
  <Accordion title="Be specific in your questions">
    Vague prompts produce vague answers. Instead of "tell me about plants", try "what vegetables grow well in partial shade in a temperate climate?". The more context you provide, the more useful the response.
  </Accordion>

  <Accordion title="Use follow-up questions">
    The AI maintains context within a conversation session. If the first response isn't quite right, ask it to clarify, expand, or approach the topic differently.
  </Accordion>

  <Accordion title="Upload relevant documents to the Knowledge Base">
    If you're asking about a specific manual, procedure, or document, upload it to the [Knowledge Base](/features/knowledge-base) first. The AI will find and reference relevant sections automatically.
  </Accordion>

  <Accordion title="Start a new session for unrelated topics">
    Conversation history is included in each request. For an unrelated topic, starting a fresh session keeps the context clean and improves response quality.
  </Accordion>

  <Accordion title="Use a larger model for complex tasks">
    If you need detailed analysis, nuanced writing, or complex reasoning, download and select a larger model. If responses are too slow, switch to a smaller model for day-to-day questions.
  </Accordion>
</AccordionGroup>

## Related

<CardGroup cols={2}>
  <Card title="Knowledge Base" icon="database" href="/features/knowledge-base">
    Upload documents so the AI can reference them when answering your questions.
  </Card>

  <Card title="Apps" icon="grid-2" href="/management/apps">
    Install, manage, and Force Reinstall the AI Assistant from the Apps page.
  </Card>
</CardGroup>