KASAUTI — Code Quality Visualizer

कसौटी is the touchstone a goldsmith rubs metal against to judge its purity. This extension does the same for code: it measures your codebase against published industry metrics, gives every file a grade from A to E, and shows the whole project as a 3D pyramid you can walk around — updating live as you type.

Everything runs offline inside VS Code. No account, no server, no telemetry, no network calls.

What it measures

Every number comes from a published, documented metric — nothing invented:

How the grade is calculated

KASAUTI uses SonarQube's publicly documented technical debt ratio model, so any grade can be defended in a review or a presentation:

debt ratio = remediation minutes ÷ (lines of code × 30 min)
A ≤ 5%    B ≤ 10%    C ≤ 20%    D ≤ 50%    E > 50%
score     = 100 × max(0, 1 − ratio ÷ 0.5)

Folder and workspace grades sum debt and lines before dividing, so a hundred tiny clean files cannot hide one terrible one.

Language support

Every source file in your workspace is analyzed. How deeply depends on whether a parser exists for its language:

Full analysis — 24 languages Bash · C · C++ · C# · Dart · Elixir · Go · Java · JavaScript · Kotlin · Lua · Objective-C · OCaml · PHP · Python (and Jupyter notebooks) · ReScript · Ruby · Rust · Scala · Solidity · Swift · TypeScript · TSX · Zig

Per-function complexity, Halstead, maintainability, nesting, parameters, classes, duplication.

Partial analysis — every other language Pascal, Fortran, Haskell, R, Julia, Perl, Clojure, Erlang, F#, Groovy, PowerShell, SQL, VHDL, Verilog, COBOL, Ada, MATLAB, GDScript, CUDA, GLSL and more.

File length, line length, indentation-based nesting, duplication and TODO markers. These files are clearly labelled "partial analysis" everywhere they appear, so a shallow check is never mistaken for a full one. You can exclude them from scores with one setting.

Skipped: generated, minified, binary, and oversized files — with the reason shown.

Using it

1. Click the KASAUTI icon in the Activity Bar, then Scan workspace.
2. Watch the pyramid build itself, one block per file.
3. Click a block to see its functions and findings; double-click to open the file.
4. Fix something. The block recolours and the score moves as you type — no save needed.

Views — 3D pyramid (floors = folder depth, blocks = files, colour = grade, height = the metric you pick), 2D treemap, and a sortable list. The treemap and list are full replacements, not afterthoughts: they work without a GPU and are fully keyboard navigable.

In the editor — squiggles and Problems-panel entries, a CodeLens above every function showing its grade and complexity, and a status-bar item with the current file and workspace grade.

Export — a self-contained HTML report (opens offline in any browser, good for submissions), SARIF 2.1.0 (validated against the OASIS schema; works with GitHub code scanning), and JSON with every metric.

Rules

Every limit is configurable, and every finding explains why it matters and how to fix it. Full pages are bundled in media/rules/.

Silencing a finding

When a rule is wrong about your code, say so in the code itself:

# kasauti-disable-next-line CQ001
def deliberately_complex(state):
    ...

kasauti-disable-line and kasauti-disable-file also work, and the rule ids are optional — leave them off to silence every rule on that line. A suppressed finding costs no technical debt, so your grade reflects the decision you made. The lightbulb on any KASAUTI diagnostic will write the comment for you in the right syntax for the language.

Project configuration

Run KASAUTI: Create Project Configuration File, or write .kasauti.toml in the workspace root:

[thresholds]
cyclomatic = 10
functionLength = 50
lineLength = 120

exclude = ["**/generated/**"]

# The CI gate is OFF unless this block exists.
# [gate]
# minGrade = "C"
# maxDebtRatio = 20.0
# failOn = ["CQ001"]

Commit it and everyone on the team grades the same way. The same settings can live under a kasauti key in package.json instead. A VS Code setting you explicitly changed still wins over the file; a setting left at its default does not, so committing a config actually has an effect.

Running in CI

npx kasauti scan .                    # report only, always exits 0
npx kasauti scan . --format sarif     # upload to GitHub code scanning
npx kasauti scan . --min-grade C      # exits 1 below grade C

The gate stays off until you add a [gate] block or pass --min-grade. This is deliberate: installing a quality tool should never break a build you did not ask it to judge. The CLI uses the same engine as the editor, and a test asserts the two produce the same grade for the same code.

Settings

The ones worth knowing:

.gitignore and dependency folders (node_modules, venv, dist, target, …) are excluded automatically.

Accessibility

Colour is never the only signal: every block, row and tooltip carries its grade letter, and severities have distinct icons. The list view mirrors the pyramid for screen readers, all views are keyboard navigable, an Okabe–Ito colour-blind palette is one setting away, and animations respect reduce motion.

Privacy

KASAUTI makes no network requests, ever. Your code is parsed, never executed, and never leaves your machine. The analysis cache lives in VS Code's own storage folder, not in your repository.

Built by

Kanak Prabhakar — BCA student, AISECT University, Hazaribagh.

Licensed MIT.