car-browser 0.1.1

//! Perception pipeline trait and basic implementation.

use async_trait::async_trait;
use thiserror::Error;

use crate::models::{A11yNode, Viewport};
use super::ax_converter::AxConverter;
use super::signals::SignalDetector;
use super::ui_map::{TextBlock, UiMap};

#[derive(Error, Debug)]
pub enum PerceptionError {
    #[error("Conversion failed: {0}")]
    ConversionFailed(String),
    #[error("Signal detection failed: {0}")]
    SignalDetectionFailed(String),
}

/// Trait for perception pipelines that convert raw browser state to UiMap.
#[async_trait]
pub trait PerceptionPipeline: Send + Sync {
    async fn perceive(
        &self,
        screenshot: &[u8],
        a11y_nodes: &[A11yNode],
        url: &str,
        viewport: Viewport,
    ) -> Result<UiMap, PerceptionError>;
}

/// Basic perception pipeline using accessibility tree only (no visual analysis).
pub struct BasicPerceptionPipeline {
    converter: AxConverter,
    signal_detector: SignalDetector,
}

impl BasicPerceptionPipeline {
    pub fn new() -> Self {
        Self {
            converter: AxConverter::new(),
            signal_detector: SignalDetector::new(),
        }
    }
}

impl Default for BasicPerceptionPipeline {
    fn default() -> Self {
        Self::new()
    }
}

#[async_trait]
impl PerceptionPipeline for BasicPerceptionPipeline {
    async fn perceive(
        &self,
        _screenshot: &[u8],
        a11y_nodes: &[A11yNode],
        url: &str,
        viewport: Viewport,
    ) -> Result<UiMap, PerceptionError> {
        let elements = self.converter.convert(a11y_nodes);
        let text_blocks = self.extract_text_blocks(a11y_nodes);
        let page_signals = self.signal_detector.detect(a11y_nodes);

        Ok(UiMap::new(
            url.to_string(),
            elements,
            text_blocks,
            page_signals,
            viewport,
            String::new(),
        ))
    }
}

impl BasicPerceptionPipeline {
    fn extract_text_blocks(&self, nodes: &[A11yNode]) -> Vec<TextBlock> {
        nodes
            .iter()
            .filter(|n| {
                let role = n.role.to_lowercase();
                matches!(role.as_str(), "statictext" | "label" | "heading")
                    && n.name.is_some()
                    && n.bounds.width > 0.0
                    && n.bounds.height > 0.0
            })
            .map(|n| {
                TextBlock::from_ax(
                    n.name.clone().unwrap_or_default(),
                    n.bounds,
                )
            })
            .collect()
    }
}

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

    #[tokio::test]
    async fn test_basic_pipeline() {
        let pipeline = BasicPerceptionPipeline::new();
        let nodes = vec![
            A11yNode {
                node_id: "n0".to_string(),
                role: "button".to_string(),
                name: Some("Submit".to_string()),
                value: None,
                bounds: Bounds::new(100.0, 100.0, 80.0, 30.0),
                children: vec![],
                focusable: true,
                focused: false,
                disabled: false,
            },
            A11yNode {
                node_id: "n1".to_string(),
                role: "statictext".to_string(),
                name: Some("Welcome".to_string()),
                value: None,
                bounds: Bounds::new(0.0, 0.0, 200.0, 20.0),
                children: vec![],
                focusable: false,
                focused: false,
                disabled: false,
            },
        ];
        let viewport = Viewport { width: 1280, height: 720, device_pixel_ratio: 2.0 };

        let ui_map = pipeline.perceive(&[], &nodes, "https://example.com", viewport).await.unwrap();
        assert_eq!(ui_map.elements.len(), 2);
        assert_eq!(ui_map.text_blocks.len(), 1);
        assert_eq!(ui_map.text_blocks[0].text, "Welcome");
    }
}