MLensIT

MLensIT is an offline Static Code Intelligence Platform for TypeScript/JavaScript. It parses your codebase, builds a structural understanding of it (symbols, dependencies, call graph), generates and runs tests, and produces actionable reports — entirely through static analysis.

No network calls, no AI inference, no telemetry: every result is deterministic and reproducible from your source tree alone.

Full project docs and source live at github.com/mit-0044/mlensit.

Features

• Static analysis — cyclomatic complexity hotspots, duplicate code, dead code, code smells (long methods, deep nesting, large classes), and a transparent 0–100 architecture score.
• Test generation — Vitest tests for functions, React/React Native components, custom hooks, and service modules, generated from static analysis.
• Test execution — runs generated tests via your project's own Vitest, measures coverage, and supports --watch mode that reruns only the tests affected by what changed.
• Reporting — console, JSON, Markdown, and HTML reports, plus a local interactive dashboard mlensit report opens immediately with live progress as the pipeline runs.
• Plugins — extend analyze with third-party AnalyzerPlugins (e.g. @mlensit/react) without forking MLensIT.
• Optional AI suggestions — @mlensit/ai is an opt-in plugin that uses your own API key to add one-sentence refactor suggestions to the highest-complexity functions. It never runs unless you declare it, and no key is ever shipped with MLensIT.

Installation

npm install -g @mit-0044/mlensit

Or run without installing:

npx @mit-0044/mlensit analyze .

Requires Node.js >=18. The package is a single self-contained bundle — no extra dependencies are installed alongside it.

Quick Start

Run from your project's root (path defaults to .):

cd my-project
mlensit analyze     # scan → build symbol/dependency/call graph → print diagnostics
mlensit generate    # analyze, then generate Vitest tests from the graph
mlensit test        # analyze, generate, persist, and run the tests
mlensit report      # full pipeline — dashboard URL printed immediately, live progress in browser

Example output:

$ mlensit analyze
Analyzing .
Scanning .
Detected typescript, npm, framework: none, tests: vitest
Analysis of . completed
Found 3 diagnostic(s)
Analyzed 42 file(s), 118 symbol(s), 67 import edge(s), 203 call edge(s)
Architecture score: 94/100 (avg. complexity 2.1, max complexity 11)
Exported symbol 'legacyHelper' is not used by any other analyzed file (src/utils/legacy.ts)
'processOrder' has cyclomatic complexity 11 (threshold: 10) — consider refactoring (src/orders/process.ts)
'Form' takes 9 props (threshold: 8) — consider grouping them into a single props object (src/components/Form.tsx)

CLI Reference

Every command accepts an optional [path] argument (the project to operate on), defaulting to ..

| Command | Description | | ------- | ----------- | | mlensit analyze [path] | Build the symbol/dependency/call graph; print diagnostics and architecture score. | | mlensit analyze [path] --ai-provider <provider> --ai-key <key> | Same, plus AI-powered refactor suggestions on the highest-complexity functions (groq | openai | gemini). | | mlensit generate [path] | Run analyze, then generate Vitest tests from the graph. | | mlensit test [path] | Run generate, then persist and execute the generated tests. | | mlensit test [path] --coverage | Also measure coverage and list files with no generated tests. | | mlensit test [path] --watch | Rerun only the tests affected by changed files until interrupted (Ctrl+C). | | mlensit report [path] | Full pipeline — dashboard URL printed immediately, live progress in browser, then the final interactive dashboard. Also writes console, JSON, Markdown, and HTML reports to <path>/.mlensit/. | | mlensit --help | List all commands and options. | | mlensit --version | Print the installed version. |

Configuration & Plugins

Declare plugins in an mlensit.config.json at your project root:

{
  "plugins": ["@mlensit/react", "./local-plugin.mjs"]
}

Each entry is either a bare package name (resolved from your project's node_modules) or a .//-prefixed path (resolved relative to the project root). A plugin that fails to load or throws is reported as a warning and never aborts the run.

AI Suggestions (@mlensit/ai)

@mlensit/ai is an optional, opt-in plugin that asks a bring-your-own-key provider for one-sentence refactor suggestions on your project's highest-complexity functions. It never runs unless declared, and no API key is ever shipped with MLensIT.

Via mlensit.config.json + environment variables:

{ "plugins": ["@mlensit/ai"] }

export MLENSIT_AI_PROVIDER=groq            # groq | openai | gemini
export MLENSIT_GROQ_API_KEY=your-key-here  # or MLENSIT_OPENAI_API_KEY / MLENSIT_GEMINI_API_KEY
mlensit analyze

Via CLI flags (no config file change needed):

mlensit analyze --ai-provider groq --ai-key your-key-here

With no provider configured, every command stays fully offline and deterministic.

License

MIT