ML-powered error classification oracle for bashrs.

Uses aprender Random Forest classifier (GPU-accelerated via trueno/wgpu) to:

• Classify shell errors into actionable categories (24 categories)
• Suggest fixes based on error patterns
• Detect error drift requiring model retraining

GPU Acceleration

Enable GPU feature for RTX 4090 acceleration via wgpu/trueno:

bashrs-oracle = { version = "*", features = ["gpu"] }

Performance Targets (from depyler-oracle)

• Accuracy: >90% (depyler achieved 97.73%)
• Training time: <1s
• Predictions/sec: >1000
• Model size: <1MB (with zstd compression)

bashrs-oracle

ML-powered error classification oracle for bashrs using aprender (GPU-accelerated).

Features

• GPU Acceleration: Uses aprender with trueno SIMD backend for fast inference
• Error Classification: Classifies shell script errors into actionable categories
• Fix Suggestions: Provides ML-based fix suggestions for common errors

Usage

use bashrs_oracle::{ErrorClassifier, ErrorCategory};

let classifier = ErrorClassifier::new()?;
let category = classifier.classify("unquoted variable expansion")?;

Categories

• Security - Security vulnerabilities (injection, etc.)
• Correctness - Logic errors and bugs
• Style - Code style issues
• Performance - Performance problems
• Portability - Cross-shell compatibility issues

License

MIT OR Apache-2.0