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

# Choose your AI providers

> Select the services that write content, create images, and search the web.

A provider is simply the service Presenton calls for one job. You can mix local and hosted providers based on privacy, quality, speed, and cost.

## Choose three capabilities

<CardGroup cols={3}>
  <Card title="Text" icon="message-lines">
    Creates outlines, writes slide content, and powers the AI assistant.
  </Card>

  <Card title="Images" icon="image">
    Finds stock photos or generates new visuals for slides.
  </Card>

  <Card title="Web search" icon="globe">
    Adds current public information when you enable research.
  </Card>
</CardGroup>

## Supported text providers

Presenton Cloud, OpenAI, Google Gemini, Anthropic, DeepSeek, Azure OpenAI, Vertex AI, Amazon Bedrock, OpenRouter, Fireworks, Together AI, Cerebras, LiteLLM, Ollama, LM Studio, custom OpenAI-compatible endpoints, and ChatGPT sign-in through Codex where supported.

## Use Presenton Cloud as a provider

In `v0.9.7-beta` and later, the local administrator can connect a Presenton account from onboarding or provider settings. Presenton displays a short device code and opens the hosted approval page. After approval, select **Presenton** as the text provider to route supported generation and document-upload workflows through Presenton Cloud.

<Note>
  This connection is a generation provider, not a sign-in method for the self-hosted application. Create or sign in to the local administrator account first.
</Note>

The connection applies to the entire installation:

* Only the local administrator can connect, replace, or disconnect it.
* Delegated access and refresh credentials are encrypted at rest and are not returned to the browser.
* Selecting the provider stores `LLM=presenton`; connecting it alone does not replace another selected provider.
* Disconnecting revokes the cloud credentials and deselects Presenton.
* Official builds include the public device-flow client configuration, so no OAuth client secret or registration is required.

<Warning>
  Presentations and documents processed through this provider leave the self-hosted environment. Use a local provider when your data-handling policy requires processing to remain on your infrastructure.
</Warning>

## Supported image providers

Pexels, Pixabay, Gemini Flash, Nano Banana Pro, DALL·E 3, GPT Image 1.5, ComfyUI, Open WebUI, and OpenAI-compatible image endpoints.

## Supported web search

Native provider search, self-hosted SearXNG, Tavily, and Exa.

| Your priority  | Good starting point                                                        |
| -------------- | -------------------------------------------------------------------------- |
| Easiest setup  | Connect Presenton Cloud, or use a hosted text model with Pexels or Pixabay |
| Local privacy  | Ollama or LM Studio with ComfyUI or no generated images                    |
| One gateway    | LiteLLM or OpenRouter                                                      |
| Private search | SearXNG                                                                    |

<Card title="Set up a local model" icon="laptop-code" href="/hosting/providers/ollama">
  Connect Presenton to Ollama step by step.
</Card>

<CardGroup cols={2}>
  <Card title="Configure text providers" icon="microchip" href="/self-hosted/configuration/text-providers">
    Find exact environment variables and examples for every supported text provider.
  </Card>

  <Card title="Configure images and search" icon="images" href="/self-hosted/configuration/images-and-search">
    Set stock, generated-image, compatible API, and web research providers.
  </Card>
</CardGroup>