From 536f42a4e82eda6ccb3880dfe1547e86e2efe8e5 Mon Sep 17 00:00:00 2001
From: JMARyA <jmarya@hydrar.de>
Date: Fri, 3 Jan 2025 13:34:59 +0100
Subject: [PATCH] add per domain config

---
 config.toml      | 42 ++++++++++++++++++++++++++++
 src/archive.rs   | 71 ++++++++++++++++++++++++++++++++++++++----------
 src/blacklist.rs | 30 ++++++++++++++++++--
 src/conf.rs      | 39 +++++++++++++++++++++++++-
 4 files changed, 164 insertions(+), 18 deletions(-)

diff --git a/config.toml b/config.toml
index 6349bc0..349ff95 100644
--- a/config.toml
+++ b/config.toml
@@ -11,6 +11,48 @@ BLACKLIST_DOMAINS = [
     "^gitlab" # All domains starting with gitlab
 ]
 
+# Domain configuration (Example)
+[[websites.domains]]
+# The domain the config applies to
+domain = "example.com"
+
+# Blacklisted Path (Regexes)
+blacklist_paths = ["/.*"]
+
+# Exclude <audio> tags
+no_audio = false
+
+# Exclude <video> tags
+no_video = false
+
+# Exclude <img> tags
+no_image = false
+
+# Exclude CSS
+no_css = false
+
+# Exclude Javascript
+no_javascript = false
+
+# Exclude fonts
+no_fonts = false
+
+# Exclude iframes
+no_frames = false
+
+# User Agent
+user_agent = "Safari"
+
 [ai]
 # Ollama URL (Enables vector search)
 OLLAMA_URL="127.0.0.1:11434"
+
+# --- Website Config
+
+[[websites.domains]]
+domain = "developer.mozilla.org"
+no_javascript = true
+
+[[websites.domains]]
+domain = "github.com"
+no_javascript = true
diff --git a/src/archive.rs b/src/archive.rs
index fdd09ed..ccd6a55 100644
--- a/src/archive.rs
+++ b/src/archive.rs
@@ -3,7 +3,12 @@ use std::{collections::HashSet, io::Read, path::PathBuf};
 use based::{request::RequestContext, result::LogAndIgnore};
 use maud::html;
 
-use crate::{blacklist::check_blacklist, conf::get_config, favicon::download_fav_for, render_page};
+use crate::{
+    blacklist::{check_blacklist, check_blacklist_path},
+    conf::get_config,
+    favicon::download_fav_for,
+    render_page,
+};
 
 /// Read directory entries into `Vec<String>`
 pub fn read_dir(dir: &PathBuf) -> Vec<String> {
@@ -340,7 +345,6 @@ impl WebsiteArchive {
     ///
     /// This function downloads the content of the URL, processes it, and saves it to the archive.
     pub async fn archive_url(&self, url: &str) {
-        // TODO : refactor
         let parsed_url = url::Url::parse(url).unwrap();
 
         let domain = parsed_url.domain().unwrap().trim_start_matches("www.");
@@ -352,6 +356,10 @@ impl WebsiteArchive {
 
         let path = parsed_url.path();
 
+        if check_blacklist_path(domain, path) {
+            return;
+        }
+
         let mut folder_name = self.dir.join(domain);
 
         download_fav_for(domain).await;
@@ -369,21 +377,56 @@ impl WebsiteArchive {
 
         log::info!("Archiving {url} to {}", filename.to_str().unwrap());
 
-        run_command(&[
-            "monolith",
-            "-I",
-            "-o",
-            filename.to_str().unwrap(),
-            &format!("https://{}/{}", domain, path),
-        ]);
+        let conf = get_config()
+            .get_domain_config(domain)
+            .cloned()
+            .unwrap_or_default();
+
+        let mut cmd = vec!["monolith", "--isolate", "-o", filename.to_str().unwrap()];
+
+        if conf.no_audio {
+            cmd.push("--no-audio");
+        }
+
+        if conf.no_css {
+            cmd.push("--no-css");
+        }
+
+        if conf.no_frames {
+            cmd.push("--no-frames");
+        }
+
+        if conf.no_fonts {
+            cmd.push("--no-frames");
+        }
+
+        if conf.no_image {
+            cmd.push("--no-images");
+        }
+
+        if conf.no_javascript {
+            cmd.push("--no-js");
+            cmd.push("--unwrap-noscript");
+        }
+
+        if conf.no_video {
+            cmd.push("--no-video");
+        }
+
+        if let Some(ua) = &conf.user_agent {
+            cmd.push("--user-agent");
+            cmd.push(ua.as_str());
+        }
+
+        let mut url = url::Url::parse(&format!("https://{domain}")).unwrap();
+        url = url.join(path).unwrap();
+        let url = url.to_string();
+        cmd.push(&url);
+
+        run_command(&cmd);
     }
 }
 
-// full text search
-// add new sites?
-// transparent auto page downloading
-// redownload after threshold
-
 fn run_command(cmd: &[&str]) {
     let mut cmd_setup = std::process::Command::new(cmd[0]);
     let cmd_setup = cmd_setup
diff --git a/src/blacklist.rs b/src/blacklist.rs
index 8ff228f..26d6ebf 100644
--- a/src/blacklist.rs
+++ b/src/blacklist.rs
@@ -7,13 +7,37 @@ pub fn check_blacklist(domain: &str) -> bool {
     let conf = get_config();
     let conf = conf.websites.as_ref();
 
+    // TODO : Block IPs
+    // Test SSRF
+
     let blacklisted_domains = conf
         .map(|x| x.BLACKLIST_DOMAINS.as_ref())
         .unwrap_or_default();
 
-    for domain_regex in blacklisted_domains.unwrap_or(&Vec::new()) {
-        let rgx = regex::Regex::new(domain_regex).unwrap();
-        if rgx.is_match(domain) {
+    check_regex(domain, blacklisted_domains.unwrap_or(&Vec::new()))
+}
+
+pub fn check_blacklist_path(domain: &str, path: &str) -> bool {
+    let conf = get_config();
+    let conf = conf.websites.as_ref();
+
+    if let Some(website) = conf {
+        let empty = Vec::new();
+        let domain_conf = website.domains.as_ref().unwrap_or(&empty);
+        if let Some(domain_conf) = domain_conf.iter().find(|x| x.domain == domain) {
+            let empty = Vec::new();
+            let blacklist = domain_conf.blacklist_paths.as_ref().unwrap_or(&empty);
+            return check_regex(path, blacklist);
+        }
+    }
+
+    false
+}
+
+pub fn check_regex(input: &str, regexes: &Vec<String>) -> bool {
+    for regex in regexes {
+        let rgx = regex::Regex::new(regex).unwrap();
+        if rgx.is_match(input) {
             return true;
         }
     }
diff --git a/src/conf.rs b/src/conf.rs
index 3b043d6..6fedc50 100644
--- a/src/conf.rs
+++ b/src/conf.rs
@@ -38,6 +38,19 @@ pub struct Config {
     pub websites: Option<WebsiteConfig>,
 }
 
+impl Config {
+    pub fn get_domain_config(&self, domain: &str) -> Option<&DomainConfig> {
+        if let Some(websites) = &self.websites {
+            if let Some(domains) = &websites.domains {
+                let domain = domains.iter().find(|x| x.domain == domain);
+                return domain;
+            }
+        }
+
+        None
+    }
+}
+
 #[allow(non_snake_case)]
 #[derive(Debug, Clone, Deserialize)]
 pub struct AIConfig {
@@ -53,9 +66,33 @@ pub struct WebsiteConfig {
 
 #[derive(Debug, Clone, Deserialize)]
 pub struct DomainConfig {
-    // TODO : Domain specific config
+    pub domain: String,
     pub blacklist_paths: Option<Vec<String>>,
+    pub no_audio: bool,
+    pub no_video: bool,
+    pub no_image: bool,
+    pub no_css: bool,
     pub no_javascript: bool,
+    pub no_fonts: bool,
+    pub no_frames: bool,
+    pub user_agent: Option<String>,
+}
+
+impl Default for DomainConfig {
+    fn default() -> Self {
+        Self {
+            domain: String::new(),
+            blacklist_paths: None,
+            no_audio: false,
+            no_video: false,
+            no_image: false,
+            no_css: false,
+            no_javascript: false,
+            no_fonts: false,
+            no_frames: false,
+            user_agent: None,
+        }
+    }
 }
 
 impl Default for Config {