# AI

The AI features let users generate text suggestions for posts, bios, stories and streams, and can also generate profile avatars and covers when the selected provider supports images.

**Important**: Most hosted AI APIs are usage-based, meaning costs can accrue if you or your users make many requests. Monitor usage in the provider dashboard and set billing limits where your provider supports them.

## General Setup

1. Go to `Admin > Settings > AI`.
2. In the **General** tab, enable **AI text generation**, **AI image generation**, or both.
3. In the **Text** tab, choose the text provider and model used for text suggestions. You can also adjust the default max tokens and temperature.
4. In the **Images** tab, choose the image provider and model used for avatars and covers.
5. In the **Providers** tab, add the API key or base URL required by your selected providers, then save the settings.

## Supported Drivers

| Driver | Text | Images | Quick setup |
| --- | --- | --- | --- |
| OpenAI | ✓ | ✓ | Create an API key in the OpenAI dashboard, add it under `OpenAI API key` , and leave the base URL as `https://api.openai.com/v1` unless you use a compatible gateway. |
| Ollama | ✓ | ✗ | Install and run Ollama on a server reachable by the app, pull the model you want to use, then set `Ollama base URL` , for example `http://127.0.0.1:11434` . |
| Anthropic | ✓ | ✗ | Create an Anthropic API key, add it under `Anthropic API key` , and keep the default base URL `https://api.anthropic.com` . |
| DeepSeek | ✓ | ✗ | Create a DeepSeek API key, add it under `DeepSeek API key` , and keep the default base URL `https://api.deepseek.com` . |
| Google (Gemini) | ✓ | ✓ | Create a Google AI Studio API key, add it under `Google API key` , and keep the default base URL `https://generativelanguage.googleapis.com` . |
| xAI | ✓ | ✓ | Create an xAI API key, add it under `xAI API key` , and keep the default base URL `https://api.x.ai/v1` . |

Once configured, users can generate text suggestions from the supported creation screens. If image generation is enabled, users can also generate profile avatars and covers from profile settings. If you see errors, confirm the selected provider has valid credentials, the chosen model is available, and the account has enough credits or quota.