web_capture/

lib.rs

//! # web-capture
//!
//! A library and CLI/microservice to render web pages as HTML, Markdown, or PNG screenshots.
//!
//! ## Features
//!
//! - Fetch HTML content from URLs
//! - Convert HTML to Markdown
//! - Capture PNG screenshots of web pages
//! - Convert relative URLs to absolute URLs
//! - Support for headless browser rendering via browser-commander
//!
//! ## Example
//!
//! ```rust,no_run
//! use web_capture::{fetch_html, convert_html_to_markdown, capture_screenshot};
//!
//! #[tokio::main]
//! async fn main() -> anyhow::Result<()> {
//!     // Fetch HTML from a URL
//!     let html = fetch_html("https://example.com").await?;
//!     println!("HTML length: {}", html.len());
//!
//!     // Convert HTML to Markdown
//!     let markdown = convert_html_to_markdown(&html, Some("https://example.com"))?;
//!     println!("Markdown: {}", markdown);
//!
//!     // Capture a screenshot
//!     let screenshot = capture_screenshot("https://example.com").await?;
//!     println!("Screenshot size: {} bytes", screenshot.len());
//!
//!     Ok(())
//! }
//! ```

pub mod animation;
pub mod archive;
pub mod batch;
pub mod browser;
pub mod extract_images;
pub mod figures;
pub mod gdocs;
pub mod html;
pub mod kreuzberg;
pub mod latex;
pub mod localize_images;
pub mod markdown;
pub mod metadata;
pub mod postprocess;
pub mod search;
pub mod themed_image;
pub mod verify;

use thiserror::Error;

/// Version of the web-capture library
pub const VERSION: &str = env!("CARGO_PKG_VERSION");

/// Error types for web-capture operations
#[derive(Error, Debug)]
pub enum WebCaptureError {
    #[error("Failed to fetch URL: {0}")]
    FetchError(String),

    #[error("Failed to parse HTML: {0}")]
    ParseError(String),

    #[error("Failed to convert to Markdown: {0}")]
    MarkdownError(String),

    #[error("Failed to capture screenshot: {0}")]
    ScreenshotError(String),

    #[error("Browser error: {0}")]
    BrowserError(String),

    #[error("Invalid URL: {0}")]
    InvalidUrl(String),

    #[error("IO error: {0}")]
    IoError(#[from] std::io::Error),

    #[error("Request error: {0}")]
    RequestError(#[from] reqwest::Error),
}

/// Result type for web-capture operations
pub type Result<T> = std::result::Result<T, WebCaptureError>;

/// Fetch HTML content from a URL
///
/// This function makes a simple HTTP GET request to fetch the HTML content.
/// For JavaScript-heavy pages, use `render_html` instead.
///
/// # Arguments
///
/// * `url` - The URL to fetch
///
/// # Returns
///
/// The HTML content as a string
///
/// # Errors
///
/// Returns an error if the fetch fails or the response cannot be decoded
pub async fn fetch_html(url: &str) -> Result<String> {
    html::fetch_html(url).await
}

/// Render HTML content from a URL using a headless browser
///
/// This function uses browser-commander to launch a headless browser,
/// navigate to the URL, and return the rendered HTML content.
///
/// # Arguments
///
/// * `url` - The URL to render
///
/// # Returns
///
/// The rendered HTML content as a string
///
/// # Errors
///
/// Returns an error if browser operations fail
pub async fn render_html(url: &str) -> Result<String> {
    browser::render_html(url).await
}

/// Convert HTML content to Markdown
///
/// # Arguments
///
/// * `html` - The HTML content to convert
/// * `base_url` - Optional base URL for converting relative URLs to absolute
///
/// # Returns
///
/// The Markdown content as a string
///
/// # Errors
///
/// Returns an error if conversion fails
pub fn convert_html_to_markdown(html: &str, base_url: Option<&str>) -> Result<String> {
    markdown::convert_html_to_markdown(html, base_url)
}

/// Capture a PNG screenshot of a URL
///
/// This function uses browser-commander to launch a headless browser,
/// navigate to the URL, and capture a screenshot.
///
/// # Arguments
///
/// * `url` - The URL to capture
///
/// # Returns
///
/// The PNG image data as bytes
///
/// # Errors
///
/// Returns an error if browser operations fail
pub async fn capture_screenshot(url: &str) -> Result<Vec<u8>> {
    browser::capture_screenshot(url).await
}

/// Convert relative URLs to absolute URLs in HTML content
///
/// # Arguments
///
/// * `html` - The HTML content to process
/// * `base_url` - The base URL to use for resolving relative URLs
///
/// # Returns
///
/// The HTML content with absolute URLs
#[must_use]
pub fn convert_relative_urls(html: &str, base_url: &str) -> String {
    html::convert_relative_urls(html, base_url)
}

/// Convert HTML content to UTF-8 encoding
///
/// Detects the current encoding from meta tags and converts to UTF-8 if needed.
///
/// # Arguments
///
/// * `html` - The HTML content to convert
///
/// # Returns
///
/// The UTF-8 encoded HTML content
#[must_use]
pub fn convert_to_utf8(html: &str) -> String {
    html::convert_to_utf8(html)
}

/// Options for enhanced HTML-to-Markdown conversion.
#[allow(clippy::struct_excessive_bools)]
#[derive(Debug, Clone)]
pub struct EnhancedOptions {
    /// Extract LaTeX formulas from img.formula, `KaTeX`, `MathJax` elements.
    pub extract_latex: bool,
    /// Extract article metadata (author, date, hubs, tags).
    pub extract_metadata: bool,
    /// Apply post-processing (unicode normalization, LaTeX spacing, etc.).
    pub post_process: bool,
    /// Detect and correct code block languages.
    pub detect_code_language: bool,
    /// CSS selector used to scope Markdown conversion.
    pub content_selector: Option<String>,
    /// CSS selector for article body Markdown; prepends the selected article title when available.
    pub body_selector: Option<String>,
}

impl Default for EnhancedOptions {
    fn default() -> Self {
        Self {
            extract_latex: true,
            extract_metadata: true,
            post_process: true,
            detect_code_language: true,
            content_selector: None,
            body_selector: None,
        }
    }
}

/// Result of enhanced HTML-to-Markdown conversion.
#[derive(Debug, Clone)]
pub struct EnhancedMarkdownResult {
    pub markdown: String,
    pub metadata: Option<metadata::ArticleMetadata>,
}

/// Convert HTML to Markdown with enhanced options.
///
/// Supports LaTeX formula extraction, metadata extraction, and
/// post-processing pipeline matching the JavaScript implementation.
///
/// # Arguments
///
/// * `html` - The HTML content to convert
/// * `base_url` - Optional base URL for resolving relative URLs
/// * `options` - Enhanced conversion options
///
/// # Returns
///
/// Enhanced result with markdown text and optional metadata
///
/// # Errors
///
/// Returns an error if base conversion fails
pub fn convert_html_to_markdown_enhanced(
    html: &str,
    base_url: Option<&str>,
    options: &EnhancedOptions,
) -> Result<EnhancedMarkdownResult> {
    let mut html_for_markdown = scope_html_with_selectors(html, options);

    if options.extract_latex {
        html_for_markdown = replace_latex_formula_elements(&html_for_markdown);
    }

    if options.detect_code_language {
        html_for_markdown = correct_code_languages(&html_for_markdown);
    }

    // Start with basic markdown conversion
    let mut md = markdown::convert_html_to_markdown(&html_for_markdown, base_url)?;

    // Extract metadata if requested
    let extracted_metadata = if options.extract_metadata {
        let meta = metadata::extract_metadata(html);
        // Prepend metadata block
        let header_lines = metadata::format_metadata_block(&meta);
        if !header_lines.is_empty() {
            let header = header_lines.join("\n");
            // Insert after the first heading
            if let Some(pos) = md.find("\n\n") {
                md = format!("{}\n\n{}\n{}", &md[..pos], header, &md[pos + 2..]);
            } else {
                md = format!("{header}\n\n{md}");
            }
        }
        // Append footer block
        let footer_lines = metadata::format_footer_block(&meta);
        if !footer_lines.is_empty() {
            md.push_str("\n\n");
            md.push_str(&footer_lines.join("\n"));
        }
        Some(meta)
    } else {
        None
    };

    // Apply post-processing if requested
    if options.post_process {
        md = postprocess::post_process_markdown(&md, &postprocess::PostProcessOptions::default());
    }

    if options.extract_latex {
        md = normalize_extracted_latex_markdown(&md);
    }

    Ok(EnhancedMarkdownResult {
        markdown: md,
        metadata: extracted_metadata,
    })
}

/// Convert HTML to Markdown using the kreuzberg html-to-markdown library.
///
/// Returns a structured result with content, metadata, tables, images, and warnings.
/// This is a high-performance alternative to `convert_html_to_markdown` using the
/// same Rust core that powers the kreuzberg ecosystem.
///
/// # Arguments
///
/// * `html` - The HTML content to convert
/// * `base_url` - Optional base URL for converting relative URLs to absolute
///
/// # Returns
///
/// A `KreuzbergResult` with structured conversion output
///
/// # Errors
///
/// Returns an error if conversion fails
pub fn convert_with_kreuzberg(
    html: &str,
    base_url: Option<&str>,
) -> Result<kreuzberg::KreuzbergResult> {
    kreuzberg::convert_with_kreuzberg(html, base_url)
}

/// Convert HTML to Markdown using kreuzberg after applying enhanced scoping options.
///
/// This keeps the alternate converter compatible with the same `contentSelector`
/// and `bodySelector` controls used by the default enhanced converter.
///
/// # Errors
///
/// Returns an error if conversion fails.
pub fn convert_with_kreuzberg_enhanced(
    html: &str,
    base_url: Option<&str>,
    options: &EnhancedOptions,
) -> Result<kreuzberg::KreuzbergResult> {
    let scoped_html = scope_html_with_selectors(html, options);
    kreuzberg::convert_with_kreuzberg(&scoped_html, base_url)
}

fn normalize_extracted_latex_markdown(markdown: &str) -> String {
    let re = regex::Regex::new(r"\$([^$\n]+)\$").expect("valid regex");
    re.replace_all(markdown, |caps: &regex::Captures<'_>| {
        let formula = caps.get(1).map_or("", |m| m.as_str()).replace(r"\\", r"\");
        format!("${formula}$")
    })
    .into_owned()
}

fn scope_html_with_selectors(html: &str, options: &EnhancedOptions) -> String {
    if let Some(body_selector) = options.body_selector.as_deref() {
        let body_html = markdown::select_html(html, body_selector);
        let title_selector = options
            .content_selector
            .as_deref()
            .map_or_else(|| "h1".to_string(), |selector| format!("{selector} h1, h1"));
        let title_html = markdown::select_html(html, &title_selector);
        return match (title_html, body_html) {
            (Some(title), Some(body)) => format!("{title}\n{body}"),
            (None, Some(body)) => body,
            _ => html.to_string(),
        };
    }

    options
        .content_selector
        .as_deref()
        .and_then(|selector| markdown::select_html(html, selector))
        .unwrap_or_else(|| html.to_string())
}

fn replace_latex_formula_elements(html: &str) -> String {
    let mut result = html.to_string();

    let img_formula_re = regex::Regex::new(r"(?is)<img\b[^>]*>").expect("valid regex");
    result = img_formula_re
        .replace_all(&result, |caps: &regex::Captures<'_>| {
            let tag = caps.get(0).map_or("", |m| m.as_str());
            if is_formula_img_tag(tag) {
                extract_attr(tag, "source")
                    .or_else(|| extract_attr(tag, "alt"))
                    .map_or_else(
                        || tag.to_string(),
                        |latex| format!("${}$", normalize_latex_for_html(&latex)),
                    )
            } else {
                tag.to_string()
            }
        })
        .into_owned();

    let math_attr_re = regex::Regex::new(
        r"(?is)<(?P<tag>mjx-container|span|div)\b(?P<attrs>[^>]*)>.*?</(?P<tag_close>mjx-container|span|div)>",
    )
    .expect("valid regex");
    math_attr_re
        .replace_all(&result, |caps: &regex::Captures<'_>| {
            let full = caps.get(0).map_or("", |m| m.as_str());
            let attrs = caps.name("attrs").map_or("", |m| m.as_str());
            let tag = caps
                .name("tag")
                .map_or("", |m| m.as_str())
                .to_ascii_lowercase();
            let tag_close = caps
                .name("tag_close")
                .map_or("", |m| m.as_str())
                .to_ascii_lowercase();

            if tag != tag_close || !is_math_attrs(&tag, attrs) {
                return full.to_string();
            }

            extract_attr(attrs, "data-tex")
                .or_else(|| extract_attr(attrs, "data-latex"))
                .or_else(|| extract_annotation_tex(full))
                .map_or_else(
                    || full.to_string(),
                    |latex| format!("${}$", normalize_latex_for_html(&latex)),
                )
        })
        .into_owned()
}

fn correct_code_languages(html: &str) -> String {
    let code_re = regex::Regex::new(r"(?is)<code\b(?P<attrs>[^>]*)>(?P<body>.*?)</code>")
        .expect("valid regex");

    code_re
        .replace_all(html, |caps: &regex::Captures<'_>| {
            let full = caps.get(0).map_or("", |m| m.as_str());
            let attrs = caps.name("attrs").map_or("", |m| m.as_str());
            let body = caps.name("body").map_or("", |m| m.as_str());

            if !has_matlab_language(attrs) || !looks_like_coq(body) {
                return full.to_string();
            }

            let updated_attrs = attrs
                .replace("language-matlab", "language-coq")
                .replace(r#"class="matlab""#, r#"class="coq""#)
                .replace("class='matlab'", "class='coq'");

            format!("<code{updated_attrs}>{body}</code>")
        })
        .into_owned()
}

fn is_formula_img_tag(tag: &str) -> bool {
    extract_attr(tag, "source").is_some()
        || extract_attr(tag, "class").is_some_and(|classes| classes.contains("formula"))
}

fn is_math_attrs(tag: &str, attrs: &str) -> bool {
    tag == "mjx-container"
        || extract_attr(attrs, "class").is_some_and(|classes| {
            classes.contains("katex") || classes.contains("math") || classes.contains("MathJax")
        })
}

fn has_matlab_language(attrs: &str) -> bool {
    extract_attr(attrs, "class").is_some_and(|classes| {
        classes
            .split_whitespace()
            .any(|class| class == "language-matlab" || class == "matlab")
    })
}

fn looks_like_coq(text: &str) -> bool {
    let decoded = crate::html::decode_html_entities(text);
    [
        "Require Import",
        "Definition",
        "Fixpoint",
        "Lemma",
        "Theorem",
        "Proof",
        "Qed",
        "Notation",
        "Inductive",
    ]
    .iter()
    .any(|needle| decoded.contains(needle))
}

fn normalize_latex_for_html(latex: &str) -> String {
    latex.trim().replace('\\', "&#92;")
}

fn extract_annotation_tex(html: &str) -> Option<String> {
    let re = regex::Regex::new(
        r#"(?is)<annotation\b[^>]*encoding\s*=\s*["']application/x-tex["'][^>]*>(.*?)</annotation>"#,
    )
    .ok()?;

    re.captures(html).and_then(|caps| {
        let text = caps.get(1)?.as_str().trim();
        (!text.is_empty()).then(|| crate::html::decode_html_entities(text))
    })
}

fn extract_attr(tag: &str, attr: &str) -> Option<String> {
    let re = regex::Regex::new(&format!(
        r#"(?is)\b{}\s*=\s*(?:"([^"]*)"|'([^']*)'|([^\s"'=<>`]+))"#,
        regex::escape(attr)
    ))
    .ok()?;

    re.captures(tag).and_then(|caps| {
        let value = caps
            .get(1)
            .or_else(|| caps.get(2))
            .or_else(|| caps.get(3))?
            .as_str()
            .trim();
        (!value.is_empty()).then(|| crate::html::decode_html_entities(value))
    })
}

// Re-export commonly used types
pub use browser::BrowserEngine;
pub use search::{
    search, SearchDiagnostics, SearchResult, SearchResultItem, DEFAULT_LIMIT, DEFAULT_PROVIDER,
    SEARCH_PROVIDERS,
};