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

# AI Agent

> Natural language interface for scientific data analysis.

# AI Agent

Relay's AI agent is a conversational interface that lets you analyze data, run models, generate pipelines, organize files, and execute custom Python code — all through natural language.

## Opening the AI panel

Click the **AI icon** in the sidebar, or use the prompt bar at the bottom of the screen. The AI panel opens as a chat interface alongside your viewer.

## File references

Reference files in your prompts using the `@` mention syntax:

1. Type `@` in the prompt bar
2. A file picker popover appears showing your files and folders
3. Select a file or type to filter
4. The file appears as an inline badge in your prompt

You can reference multiple files in a single prompt. The AI agent will use the referenced files as context for your request.

## Smart suggestions

When you drag and drop a file onto a pane, Relay suggests relevant actions based on the file type:

| File Type                            | Suggested Actions                                            |
| ------------------------------------ | ------------------------------------------------------------ |
| **Scientific TIFF** (.tif, .ome.tif) | SAM3 segmentation, PlantCV analysis, Suite2p calcium imaging |
| **Microscopy** (.czi, .nd2, .lif)    | SAM3 segmentation, PlantCV analysis, Suite2p calcium imaging |
| **Standard images** (.png, .jpg)     | Segmentation, PlantCV, SExtractor                            |
| **Video** (.mp4, .avi, .mov)         | DeepLabCut tracking, SAM3 video segmentation                 |
| **DICOM** (.dcm)                     | Segmentation, description                                    |
| **Neural data** (.nwb, .h5, .bin)    | SpikeInterface spike sorting                                 |
| **Astronomy** (.fits)                | SExtractor source extraction                                 |
| **All files**                        | Generate a processing pipeline                               |

## Chat features

* **Markdown rendering** — responses are formatted with headings, lists, code blocks, and tables
* **Inline images** — analysis results appear as images in the chat
* **Action buttons** — suggested follow-up actions appear as clickable buttons
* **Analysis badges** — results like cell counts display as inline badges
* **Chart visualizations** — bar charts and histograms render directly in chat
* **Segmentation overlays** — segmentation masks appear on the canvas

## LLM providers

Relay supports multiple LLM backends:

| Provider      | Default Model | Notes                     |
| ------------- | ------------- | ------------------------- |
| **OpenAI**    | gpt-4o-mini   | Default provider          |
| **Anthropic** | Claude Sonnet | Direct API integration    |
| **Ollama**    | Llama 3.1     | Self-hosted, local models |

The provider is selected automatically based on configured API keys, or can be set explicitly.