# In-App Assistant

:::caution
Memside uses an externally managed provider model. Model usage and billing are
handled by the selected provider.
:::

## Overview

The in-app assistant provides a chat surface inside Memside. Open it with the
floating **AI** button on the dashboard.

The button can be moved. The assistant window can also be repositioned and
resized.

## Connect a provider

Open **Settings > LLM Configuration**.

Memside supports:

- **Custom (OpenAI-compatible)** providers;
- **Local (Ollama)** for models running on the same computer.

### Custom provider

1. Select **Custom (OpenAI-compatible)**.
2. Choose a provider example when available.
3. Review the provider label, base URL, and model.
4. Enter the provider API key.
5. Select **Save**, then **Test**.

### Local Ollama

1. Start Ollama or the local model server.
2. Select **Local (Ollama)**.
3. Review the base URL and model.
4. No API key is required for Ollama.
5. Select **Save**, then **Test**.

The assistant is ready when the test succeeds.

## Costs

Memside does not bill for the external model. Usage costs and limits are
controlled by the connected provider.

## Assistant modes

The assistant starts in **Chat** mode. Quick Actions provide:

- **Chat** for normal discussion;
- **Create** to draft a Memory;
- **Update** to suggest changes to a Memory;
- **Fetch** to find saved Memories for the conversation.

Modes are selected manually. Create and Update always require review before
changes are saved.

## Assistant controls

- **Attach Bundle** adds the current Memory Bundle to the message.
- **Clear Chat** removes the current assistant conversation.
- **Close** hides the assistant.

The assistant is a convenient way to use a connected model with saved context.
It remains inactive until a provider is configured.

## Next steps

- [Settings](/product/settings/)
- [Memory Bundles](/product/memory-bundles/)
- [Security and Privacy](/help/security-and-privacy/)