@optiprune/cli

🚀 OptiPrune

💎 The Vision

Most dead-code analyzers just guess. They look at import graphs and hope they don't miss anything critical. This leads to false positives, broken builds, and developer frustration.

OptiPrune is different. We use formal logic, isolated execution, and a high-performance engine to not just find dead code, but to mathematically prove it.

"Stop Guessing, Start Proving."

⚡ The 3 Pillars of Superiority

1. Raw Power: The Yuku Engine (Zig)

While other tools are throttled by the N-API bottleneck, OptiPrune utilizes the Yuku Engine written in Zig. It minimizes the overhead between native performance and the JavaScript runtime.

• The Result: Up to 3x faster parsing than traditional tools, even in massive monorepos.

2. Deep Intelligence: SMT & Z3 Solver

OptiPrune is the first analyzer to use a real SMT Solver (Z3). We don't just analyze if a function is exported; we analyze if the code inside the function is logically reachable.

• The Result: Detects dead logic paths (e.g., impossible if conditions) that are completely invisible to Knip.

3. Absolute Precision: WASM Sandbox Execution

Dynamic imports are the final boss of static analysis. OptiPrune solves this through a WASM-based QuickJS sandbox. We securely execute critical code snippets to resolve paths at runtime.

• The Result: Zero false alarms for dynamic paths. If OptiPrune says it's dead, it's dead.

🥊 OptiPrune vs. Knip: The Head-to-Head

Feature	Knip	OptiPrune
Engine	Babel / OXC (Standard)	Yuku / Zig (Hyper-Speed)
Logic Analysis	Heuristics (Guessing)	Z3 SMT Solver (Proving)
Dynamic Paths	Pattern Matching	WASM Sandbox Execution
Interface Audit	Ignores Members	Deep Member-Level Analysis
Framework Support	Plugins (Core-Level)	7-Layer Semantic Context
False Positives	High (in complex setups)	Near-Zero (Context Aware)

🏗️ The 7-Layer Architecture

OptiPrune operates in seven specialized layers to guarantee maximum accuracy:

• Layer 1: Discovery – Ultra-fast file scanning.
• Layer 2: Basic CFG – Detects standard dead code after terminal statements.
• Layer 3: SMT Logic – Mathematical path proofs with Z3.
• Layer 4: WASM Sandbox – Dynamic path resolution via execution.
• Layer 6: Schema Shield – Protection for Zod, Decorators & Contracts.
• Layer 6: Dependency Audit – Scans lockfiles & package.json scripts.
• Layer 7: Topology Engine – Understands NestJS DI & Event Buses.

📈 Benchmark Numbers (Real-World Test)

Tested on a NestJS project with 1000+ files.

• Knip Speed: 1.26s (with crash risks on complex types)
• OptiPrune Speed: 0.87s (Stable & Precise)
• Accuracy: OptiPrune found 15% more real dead code (unused interface properties & logical errors) that Knip completely missed.

Installation

Install Optiprune as a dev dependency via pnpm, npm, or yarn:

pnpm add -D @optiprune/core
# or
npm install --save-dev @optiprune/core
# or
yarn add -D @optiprune/core

---

## Usage

Run Optiprune from your project root:

```bash
npx @optiprune/cli

CLI Options (@optiprune/cli)

Flag	Description	Default
-r, --rootDir	Project root directory	process.cwd()
-e, --entry	Entry point patterns (glob)	[]
-i, --ignore	Patterns to ignore	[]
--no-report-unused-exports	Disable unused export reporting	false
--fail-on	Fail on confidence (high/medium/low/none)	high
--json	Output as JSON	false
--sarif	Output as SARIF	false
--skip-3	Skip Layer 3 (SMT Constraint Solver)	false
--skip-4	Skip Layer 4 (Concolic Execution Proofs)	false
--fix <targets...>	Required fix targets: files, exports, dependencies, devDependencies, or conditions	none
--confidence <level>	Minimum fix confidence: high, medium+, low+, or all	high
--force	Override the configured confidence safety boundary	false
--dry-run	Report planned fixes without modifying files	false

For example, to fix files, exports, dependencies, and development dependencies with low-confidence findings included, run:

npx @optiprune/cli analyze --fix files exports dependencies devDependencies --confidence low+

Use --force only when you explicitly accept fixes below the configured safety boundary:

npx @optiprune/cli analyze --fix exports --confidence high --force

--confidence, --force, and --dry-run require at least one --fix target. Unsupported or unknown targets are rejected before analysis begins.

🤝 Join the Revolution

OptiPrune isn't just a tool. It's a technical statement. Help us save the world from dirty code.

GitHub: DreamLongYT/optiprune Web: opti.drml.int.yt

Config

To setup OptiPrune, see config.md for more