Skip to content
Dennis Lee edited this page May 27, 2026 · 1 revision

title: swark radar_quadrant: Tools radar_ring: Assess radar_position: inner created: 2026-05-26 last_updated: 2026-05-26 tags: [architecture, diagrams, llm, documentation, vscode, mermaid] source_url: https://github.com/swark-io/swark

swark

A VS Code extension that generates architecture diagrams automatically from source code using LLMs. It reads the codebase and produces Mermaid diagrams without requiring manual annotation, a DSL, or a pre-written model.

How It Works

swark scans the open workspace, feeds relevant source files to an LLM, and outputs a Mermaid diagram representing the inferred component relationships, service boundaries, and data flows. The diagram is generated fresh each time, reflecting the current state of the code rather than a manually maintained model.

Differentiation from Similar Tools

Three tools on the radar address architecture visualisation:

  • C4 Model and Structurizr: requires writing and maintaining a DSL model; produces precise, multi-level diagrams; best for documented, stable architecture.
  • GitDiagram: works from a public GitHub URL; no local install; broad repo-level overview.
  • swark: fully automated from local source; no model to write; best for quick comprehension of unfamiliar codebases or for generating a starting-point diagram.

swark trades precision for automation. The output requires review — LLM inference is imperfect — but removes the blank-page problem for documentation and onboarding.

Use Cases

  • Onboarding to an unfamiliar codebase
  • Generating a first-draft architecture diagram for documentation
  • Quickly visualising component relationships before a refactoring session

Radar Assessment

Placed in Tools / Assess / inner. The LLM-from-code diagram generation angle is distinct from existing tools. Inner position reflects low friction (VS Code extension install) and direct applicability to any active project needing architecture documentation. Trial gate: one architecture diagram generated from a real codebase, reviewed for accuracy, and used in a team onboarding or design discussion.

Clone this wiki locally