//! Deep Learning Artifact Detection with Explainability for MEG/EEG
//!
//! This crate provides transformer-based artifact detection using ONNX inference
//! with interpretable saliency maps to explain which channels and timepoints
//! contributed to artifact detection.
//!
//! # Features
//!
//! - **Multi-label artifact detection**: Eye blinks, muscle artifacts, heartbeat, etc.
//! - **ONNX inference**: High-performance inference with CPU/GPU/CoreML backends
//! - **Explainability**: Integrated Gradients, attention maps, SHAP values
//! - **Pre-trained models**: Ready-to-use models for common artifact types
//!
//! # Supported Artifact Types
//!
//! | Artifact | Description |
//! |----------|-------------|
//! | Eye Blink (EOG) | Blink artifacts in frontal channels |
//! | Eye Movement | Saccades and smooth pursuit artifacts |
//! | Muscle (EMG) | High-frequency muscle contamination |
//! | Heartbeat (ECG) | Cardiac artifact in MEG/EEG |
//! | Line Noise | 50/60 Hz power line interference |
//! | Movement | Head/body movement artifacts |
//!
//! # Architecture
//!
//! ```text
//! Input: [batch, channels, time]
//!            │
//!     ┌──────▼──────┐
//!     │ 1D Conv     │  Feature extraction
//!     │ Encoder     │
//!     └──────┬──────┘
//!            │
//!     ┌──────▼──────┐
//!     │ Transformer │  Self-attention over time
//!     │ Encoder     │
//!     └──────┬──────┘
//!            │
//!     ┌──────▼──────┐
//!     │ Multi-label │  Artifact classification
//!     │ Classifier  │
//!     └──────┬──────┘
//!            │
//! Output: [batch, n_artifact_types]
//! ```
//!
//! # Example
//!
//! ```rust,no_run
//! use rtx_neuro_artifacts::{ArtifactDetector, DetectorConfig, ArtifactType};
//!
//! fn main() -> Result<(), Box<dyn std::error::Error>> {
//!     // Create detector with default config
//!     let config = DetectorConfig::default();
//!     let detector = ArtifactDetector::new(config)?;
//!
//!     // Detect artifacts in EEG segment [channels x time]
//!     let eeg_data: Vec<Vec<f64>> = vec![vec![0.0; 1000]; 64];
//!     let result = detector.detect(&eeg_data)?;
//!
//!     // Check which artifacts were detected
//!     for (artifact_type, probability) in result.predictions() {
//!         if probability > 0.5 {
//!             println!("{:?} detected with probability {:.2}", artifact_type, probability);
//!         }
//!     }
//!
//!     Ok(())
//! }
//! ```

#![warn(missing_docs)]

pub mod detector;
pub mod error;
pub mod explain;
pub mod labels;
pub mod models;

pub use detector::{ArtifactDetector, DetectionBatch, DetectionResult, DetectorConfig};
pub use error::{ArtifactError, ArtifactResult};
pub use explain::{
    ArtifactExplainer, AttentionMap, ChannelImportance, ExplainerConfig, ExplanationResult,
    SaliencyMap,
};
pub use labels::{ArtifactLabel, ArtifactRegion, ArtifactType};
pub use models::{ModelInfo, ModelRegistry, ModelSource, download_model};

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn test_crate_compiles() {
        // Smoke test - crate structure is sound
        assert!(true);
    }
}