neural-shared

Shared utilities for Neural Garage code analysis tools

Overview

neural-shared provides common functionality for the Neural Garage suite of code analysis tools. It includes:

• Parser Module - Tree-sitter-based AST parsing for multiple languages
• Scanner Module - File system traversal with .gitignore support
• Report Module - Generic reporting framework for analysis results

Features

Parser Module

Language-agnostic parsing infrastructure:

use neural_shared::parser::{Language, PythonParser, Parser};

let parser = PythonParser::new()?;
let source = "def hello(): pass";
let parsed = parser.parse(source, Path::new("example.py"))?;

// Access definitions, usages, and entry points
for def in parsed.definitions {
    println!("Found: {} at {}:{}", def.name, def.location.line, def.location.column);
}

Supported languages:

• Python
• TypeScript
• JavaScript

Scanner Module

Efficient file system scanning:

use neural_shared::Scanner;

let scanner = Scanner::new(Path::new("./src"));
let files = scanner.scan()?;  // Respects .gitignore

Features:

• .gitignore support via the ignore crate
• Parallel file scanning
• Language-specific file filtering

Report Module

Generic reporting framework with Finding trait:

use neural_shared::report::{Finding, Reporter, JsonReporter};
use serde::Serialize;

#[derive(Serialize)]
struct MyFinding {
    // Your finding data
}

impl Finding for MyFinding {
    fn kind(&self) -> String { /* */ }
    fn name(&self) -> String { /* */ }
    // ... other methods
}

let reporter = JsonReporter;
let json = reporter.report(&findings)?;

Built-in reporters:

• JsonReporter - LLM-friendly JSON output
• MarkdownReporter - Human-readable markdown

Part of Neural Garage 🧠🔧

This library is part of the Neural Garage monorepo.

Tools using neural-shared:

• bury - Dead code detector
• complexity (coming soon) - Code complexity analyzer

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contributing

See the main repository for contribution guidelines.