Language Detection

Language Detection

Automatically detect project languages and apply language-specific prevention rules.

The Rule

Know what language you're working in and apply its safety patterns.

Language detection is determined by:

1. Config files (requirements.txt, go.mod, Cargo.toml, tsconfig.json, etc.)
2. Source file extensions (.py, .ts, .go, .rs)
3. Project structure indicators (node_modules/, venv/, etc.)

Supported Languages

Language	Config Signals	Rules
Python	requirements.txt, pyproject.toml, setup.py, Pipfile	eval/exec, subprocess shell=True, bare except, mutable defaults, pickle, hardcoded secrets, SQL injection, type: ignore
TypeScript/JS	tsconfig.json, package.json	any type, non-null assertion, eval/Function, innerHTML, console.log, document.write, hardcoded secrets
Go	go.mod, go.sum	ignored errors, panic, SQL concat, os/exec, Close without defer, goroutine without context
Rust	Cargo.toml, Cargo.lock	unsafe blocks, unwrap, panic!, todo!/unimplemented!, clone on large types, raw pointer deref

How It Works

1. Detection: Scans project for config files, source extensions, and directory indicators
2. Rule Loading: Loads language-specific patterns from .guardrails/prevention-rules/languages/
3. Checking: When guardrail_check_pattern is called, both generic and language-specific rules are applied
4. File Filtering: Rules are filtered by filePatterns — Python rules only fire on .py files, etc.

Adding New Languages

Create a JSON file at .guardrails/prevention-rules/languages/<language>.json:

{
  "$schema": "./language-rules.schema.json",
  "language": "java",
  "version": "1.0.0",
  "detectors": ["*.java", "pom.xml", "build.gradle"],
  "rules": [
    {
      "id": "java-reflection",
      "name": "Reflection usage",
      "pattern": "\\.getDeclaredMethod\\s*\\(",
      "severity": "warning",
      "message": "Reflection bypasses compile-time checking — verify usage is intentional",
      "filePatterns": ["\\.java$"]
    }
  ]
}

When to Use

• Before writing code in any language — check which patterns to avoid
• After code generation — scan for language-specific antipatterns
• During code review — validate against language-specific safety rules

Pi Enforcement

When using @architectit/pi-guardrails:

• guardrail_detect_language — Scan a project directory and return detected languages
• guardrail_get_language_profile — Get detected languages + available rules with descriptions
• guardrail_check_pattern — Check code content against all loaded rules (generic + language-specific)

Language rules are automatically loaded when PatternRuleEngine.loadRules() is called with a LanguageDetector configured. No manual setup needed — detection happens on first rule load.