Getting started with ExcaliBrain - starting from an empty Obsidian Vault (Part 1 of 3)

ExcaliBrain turns an empty Obsidian vault into a visual argument map by linking markdown notes into a structured graph—issues, positions, supporting or opposing arguments, and evidence—so relationships become readable at a glance. The setup can feel intimidating because Excalibrain depends on multiple plugins and a set of styling and ontology rules, but the walkthrough breaks the process into concrete steps: create a new vault, install required dependencies, open Excalibrain, then start building a dialog map node-by-node.

The example begins with a fresh Obsidian vault named “XcalliBrain.” To make Excalibrain work, three plugins must be installed and enabled: “Xcalli Raw” (the foundation Excalibrain builds on), “Data View” (used to extract document structure), and “Xcalli Brain” itself. The walkthrough emphasizes that Excalibrain inherits behavior and some settings from Xcalli Raw, so later formatting and graph behavior may depend on configuration outside Excalibrain.

With the vault ready, the central issue is created as a markdown node titled “remote work policy.” Excalibrain is then opened via Obsidian’s command palette (using “brain” to launch “brain normal”). From there, the map is populated using the Issue-Based Information Systems (IBIS) structure: the issue is phrased as a question, and each position is added as a child node—here, “yes to remote work” and “no to remote work.” Clicking a generated wiki link creates the underlying markdown file for a node, while nodes that don’t yet exist in the vault can still appear in the graph, allowing connections to be planned before the notes are fully created.

A major portion of the setup focuses on styling. Excalibrain uses tags to determine how nodes render, so the walkthrough adds two tags—“question” and “position”—and then configures node appearance in Excalibrain’s settings. Customization is done through a base style inheritance model: new node types inherit default formatting unless explicitly overridden. The “position” node type is customized with a light-bulb emoji prefix, background color, opacity, and later text color for readability. The “question” node type is adjusted to display a question mark prefix.

When “no to remote work” initially lacks a corresponding markdown document, it doesn’t receive the same styling because the needed tags aren’t present. The fix is to create the missing markdown file by shift-clicking the node and choosing to create a new markdown document, then copying in the node’s content (including the position label and supporting argument structure). Finally, the layout problem—positions appearing in the wrong side of the central issue—is resolved by defining ontology relationships using Data View fields. Right-clicking a position line and adding it to the Excalibrain ontology as a “next friends” relationship makes Excalibrain place the position on the right side and changes the connecting line from dashed (inferred) to solid (explicit). The guide ends by recommending viewers replicate the steps and then continue with Part 2 for deeper customization and further graph building.