Message Types & Formats

Ontheia represents information in the chat history in different ways.

1. Text & Markdown

AI responses are rendered as formatted Markdown. This includes:

• Bold text, lists, and tables.
• Code Blocks: With syntax highlighting, a copy button, and a pencil icon that loads the content into the artifact panel as a draft (see Artifacts).
• Mathematical formulas (LaTeX).

2. Diagrams (Mermaid)

A ```mermaid-code block is rendered as a diagram in the chat — in agent responses as well as in your own messages. Simply ask an agent: “Draw the workflow as a mermaid flowchart.”

• While streaming, the source stays visible; once the diagram is complete, the block automatically flips to the graphic.
• Toolbar on the block: zoom in/out/reset, fullscreen (overlay with its own zoom, close via Escape or clicking the backdrop), toggle diagram ↔ source, copy the source, and edit in panel (see Artifacts).
• Invalid mermaid code is shown unchanged as a code block.

All mermaid diagram types are supported (flowchart, sequence, class, ER, Gantt, and more). Rendering happens entirely locally in the browser.

3. Images & Visual Content

Images can be attached directly to a message in the Composer. Supported formats: JPEG, PNG, GIF, WebP.

• The image is sent to the AI model together with the message text (vision input).
• Supported by multimodal providers (e.g. Claude, GPT-5).
• Use cases: screenshots, diagrams, documents, photos — the agent can describe, analyze, or extract data from them.

Note: Image support requires a provider with vision capability. If the selected model does not support images, the attachment is ignored.

4. File Cards

When an agent reads or writes a file, a file card appears instead of the content. Clicking it opens the file in the artifact panel for viewing and editing; PDFs open in a viewer with selectable text. Details in Artifacts.

5. Tool Cards (Permissions)

When an agent wants to use a tool (e.g., access a file), a Tool Card appears:

• Details: Shows which server and tool are to be called and which arguments are being sent.
• Allow Once: Executes the current call.
• Always Allow: The agent may use this tool for the rest of the chat without further inquiry.
• Decline: Refuses access (the agent receives a corresponding error message).

6. Status & Error Messages

Technical events are presented compactly:

• System Hints: Inform about the start of Chains or the loading of memory.
• Errors: If a provider is unreachable or a tool crashes, this is displayed marked in red.