(random) Sanitize random search results with DOMAIN_AVAILABILITY_INFORMATION join

Reduce socket and connection timeouts in HttpClient and adjust thread counts for job consumers

ROADMAP.md

@@ -38,14 +38,6 @@ associated with each language added, at least a models file or two, as well as s
 
 It would be very helpful to find a speaker of a large language other than English to help in the fine tuning.
 
-## Support for binary formats like PDF
-
-The crawler needs to be modified to retain them, and the conversion logic needs to parse them.  
-The documents database probably should have some sort of flag indicating it's a PDF as well.
-
-PDF parsing is known to be a bit of a security liability so some thought needs to be put in
-that direction as well.
-
 ## Custom ranking logic
 
 Stract does an interesting thing where they have configurable search filters.
@@ -66,6 +58,14 @@ One of the search engine's biggest limitations right now is that it does not ind
 
 # Completed
 
+## Support for binary formats like PDF (COMPLETED 2025-05)
+
+The crawler needs to be modified to retain them, and the conversion logic needs to parse them.  
+The documents database probably should have some sort of flag indicating it's a PDF as well.
+
+PDF parsing is known to be a bit of a security liability so some thought needs to be put in
+that direction as well.
+
 ## Web Design Overhaul (COMPLETED 2025-01)
 
 The design is kinda clunky and hard to maintain, and needlessly outdated-looking.  

build.gradle

@@ -5,7 +5,7 @@ plugins {
 
     // This is a workaround for a bug in the Jib plugin that causes it to stall randomly
     // https://github.com/GoogleContainerTools/jib/issues/3347
-    id 'com.google.cloud.tools.jib' version '3.4.4' apply(false)
+    id 'com.google.cloud.tools.jib' version '3.4.5' apply(false)
 }
 
 group 'marginalia'
@@ -43,12 +43,11 @@ subprojects.forEach {it ->
 }
 
 ext {
-    jvmVersion=23
+    jvmVersion = 24
-    dockerImageBase='container-registry.oracle.com/graalvm/jdk:23'
+    dockerImageBase='container-registry.oracle.com/graalvm/jdk:24'
     dockerImageTag='latest'
     dockerImageRegistry='marginalia'
-    jibVersion = '3.4.4'
+    jibVersion = '3.4.5'
-
 }
 
 idea {

code/common/config/java/nu/marginalia/UserAgent.java

@@ -1,3 +1,8 @@
 package nu.marginalia;
 
+/**
+ * A record representing a User Agent.
+ * @param uaString - the header value of the User Agent
+ * @param uaIdentifier - what we look for in robots.txt
+ */
 public record UserAgent(String uaString, String uaIdentifier) {}

code/common/db/resources/db/migration/V25_01_0_000__nsfw_domains.sql

@@ -0,0 +1,5 @@
+CREATE TABLE IF NOT EXISTS WMSA_prod.NSFW_DOMAINS (
+    ID INT NOT NULL AUTO_INCREMENT,
+    TIER INT NOT NULL,
+    PRIMARY KEY (ID)
+);

code/common/db/resources/db/migration/V25_01_0_001__ping_domains.sql

@@ -0,0 +1,213 @@
+
+-- Create metadata tables for domain ping status and security information
+
+-- These are not ICMP pings, but rather HTTP(S) pings to check the availability and security
+-- of web servers associated with domains, to assess uptime and changes in security configurations
+-- indicating ownership changes or security issues.
+
+-- Note: DOMAIN_ID and NODE_ID are used to identify the domain and the node that performed the ping.
+-- These are strictly speaking foreign keys to the EC_DOMAIN table, but as it
+-- is strictly append-only, we do not need to enforce foreign key constraints.
+
+CREATE TABLE IF NOT EXISTS DOMAIN_AVAILABILITY_INFORMATION (
+    DOMAIN_ID INT NOT NULL PRIMARY KEY,
+    NODE_ID INT NOT NULL,
+
+    SERVER_AVAILABLE BOOLEAN NOT NULL,  -- Indicates if the server is available (true) or not (false)
+    SERVER_IP VARBINARY(16),            -- IP address of the server (IPv4 or IPv6)
+    SERVER_IP_ASN INTEGER,              -- Autonomous System number
+
+    DATA_HASH BIGINT,                   -- Hash of the data for integrity checks
+    SECURITY_CONFIG_HASH BIGINT,        -- Hash of the security configuration for integrity checks
+
+    HTTP_SCHEMA ENUM('HTTP', 'HTTPS'),  -- HTTP or HTTPS protocol used
+    HTTP_ETAG VARCHAR(255),             -- ETag of the resource as per HTTP headers
+    HTTP_LAST_MODIFIED VARCHAR(255),    -- Last modified date of the resource as per HTTP headers
+    HTTP_STATUS INT,                    -- HTTP status code (e.g., 200, 404, etc.)
+    HTTP_LOCATION VARCHAR(255),         -- If the server redirects, this is the location of the redirect
+    HTTP_RESPONSE_TIME_MS SMALLINT UNSIGNED, -- Response time in milliseconds
+
+    ERROR_CLASSIFICATION ENUM('NONE', 'TIMEOUT', 'SSL_ERROR', 'DNS_ERROR', 'CONNECTION_ERROR', 'HTTP_CLIENT_ERROR', 'HTTP_SERVER_ERROR', 'UNKNOWN'), -- Classification of the error if the server is not available
+    ERROR_MESSAGE VARCHAR(255),         -- Error message if the server is not available
+
+    TS_LAST_PING TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP, -- Timestamp of the last ping
+    TS_LAST_AVAILABLE TIMESTAMP,        -- Timestamp of the last time the server was available
+    TS_LAST_ERROR TIMESTAMP,             -- Timestamp of the last error encountered
+
+    NEXT_SCHEDULED_UPDATE TIMESTAMP NOT NULL,
+    BACKOFF_CONSECUTIVE_FAILURES INT NOT NULL DEFAULT 0, -- Number of consecutive failures to ping the server
+    BACKOFF_FETCH_INTERVAL INT NOT NULL DEFAULT 60 -- Interval in seconds for the next scheduled ping
+) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin;
+
+CREATE INDEX IF NOT EXISTS DOMAIN_AVAILABILITY_INFORMATION__NODE_ID__DOMAIN_ID_IDX ON DOMAIN_AVAILABILITY_INFORMATION (NODE_ID, DOMAIN_ID);
+CREATE INDEX IF NOT EXISTS DOMAIN_AVAILABILITY_INFORMATION__NEXT_SCHEDULED_UPDATE_IDX ON DOMAIN_AVAILABILITY_INFORMATION (NODE_ID, NEXT_SCHEDULED_UPDATE);
+
+
+
+CREATE TABLE IF NOT EXISTS DOMAIN_SECURITY_INFORMATION (
+    DOMAIN_ID INT NOT NULL PRIMARY KEY,
+    NODE_ID INT NOT NULL,
+
+    ASN INTEGER,                     -- Autonomous System Number (ASN) of the server
+    HTTP_SCHEMA ENUM('HTTP', 'HTTPS'),  -- HTTP or HTTPS protocol used
+    HTTP_VERSION VARCHAR(10),           -- HTTP version used (e.g., HTTP/1.1, HTTP/2)
+    HTTP_COMPRESSION VARCHAR(50),       -- Compression method used (e.g., gzip, deflate, br)
+    HTTP_CACHE_CONTROL TEXT,            -- Cache control directives from HTTP headers
+
+    SSL_CERT_NOT_BEFORE TIMESTAMP,         -- Valid from date (usually same as issued)
+    SSL_CERT_NOT_AFTER TIMESTAMP,          -- Valid until date (usually same as expires)
+
+    SSL_CERT_ISSUER VARCHAR(255),         -- CA that issued the cert
+    SSL_CERT_SUBJECT VARCHAR(255),        -- Certificate subject/CN
+
+    SSL_CERT_PUBLIC_KEY_HASH BINARY(32),     -- SHA-256 hash of the public key
+    SSL_CERT_SERIAL_NUMBER VARCHAR(100),     -- Unique cert serial number
+    SSL_CERT_FINGERPRINT_SHA256 BINARY(32),  -- SHA-256 fingerprint for exact identification
+    SSL_CERT_SAN TEXT,                       -- Subject Alternative Names (JSON array)
+    SSL_CERT_WILDCARD BOOLEAN,               -- Wildcard certificate (*.example.com)
+
+    SSL_PROTOCOL VARCHAR(20),             -- TLS 1.2, TLS 1.3, etc.
+    SSL_CIPHER_SUITE VARCHAR(100),        -- e.g., TLS_AES_256_GCM_SHA384
+    SSL_KEY_EXCHANGE VARCHAR(50),         -- ECDHE, RSA, etc.
+    SSL_CERTIFICATE_CHAIN_LENGTH TINYINT, -- Number of certs in chain
+
+    SSL_CERTIFICATE_VALID BOOLEAN,        -- Valid cert chain
+
+    HEADER_CORS_ALLOW_ORIGIN TEXT,               -- Could be *, specific domains, or null
+    HEADER_CORS_ALLOW_CREDENTIALS BOOLEAN,       -- Credential handling
+    HEADER_CONTENT_SECURITY_POLICY_HASH INT,     -- CSP header, hash of the policy
+    HEADER_STRICT_TRANSPORT_SECURITY VARCHAR(255), -- HSTS header
+    HEADER_REFERRER_POLICY VARCHAR(50),          -- Referrer handling
+    HEADER_X_FRAME_OPTIONS VARCHAR(50),          -- Clickjacking protection
+    HEADER_X_CONTENT_TYPE_OPTIONS VARCHAR(50),   -- MIME sniffing protection
+    HEADER_X_XSS_PROTECTION VARCHAR(50),         -- XSS protection header
+
+    HEADER_SERVER VARCHAR(255),                 -- Server header (e.g., Apache, Nginx, etc.)
+    HEADER_X_POWERED_BY VARCHAR(255),           -- X-Powered-By header (if present)
+
+    TS_LAST_UPDATE TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP -- Timestamp of the last SSL check
+) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin;
+
+
+CREATE INDEX IF NOT EXISTS DOMAIN_SECURITY_INFORMATION__NODE_ID__DOMAIN_ID_IDX ON DOMAIN_SECURITY_INFORMATION (NODE_ID, DOMAIN_ID);
+
+CREATE TABLE IF NOT EXISTS DOMAIN_SECURITY_EVENTS (
+    CHANGE_ID BIGINT AUTO_INCREMENT PRIMARY KEY, -- Unique identifier for the change
+    DOMAIN_ID INT NOT NULL, -- Domain ID, used as a foreign key to EC_DOMAIN
+    NODE_ID INT NOT NULL,
+
+    TS_CHANGE TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP, -- Timestamp of the change
+
+    CHANGE_ASN BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to ASN (Autonomous System Number)
+    CHANGE_CERTIFICATE_FINGERPRINT BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to SSL certificate fingerprint
+    CHANGE_CERTIFICATE_PROFILE BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to SSL certificate profile (e.g., algorithm, exchange)
+    CHANGE_CERTIFICATE_SAN BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to SSL certificate SAN (Subject Alternative Name)
+    CHANGE_CERTIFICATE_PUBLIC_KEY BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to SSL certificate public key
+    CHANGE_SECURITY_HEADERS BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to security headers
+    CHANGE_IP_ADDRESS BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to IP address
+    CHANGE_SOFTWARE BOOLEAN  NOT NULL DEFAULT FALSE, -- Indicates if the change is related to the generator (e.g., web server software)
+    OLD_CERT_TIME_TO_EXPIRY INT, -- Time to expiry of the old certificate in hours, if applicable
+
+    SECURITY_SIGNATURE_BEFORE BLOB NOT NULL, -- Security signature before the change, gzipped json record
+    SECURITY_SIGNATURE_AFTER BLOB NOT NULL  -- Security signature after the change, gzipped json record
+) CHARACTER SET utf8mb4 COLLATE utf8mb4_bin;
+
+CREATE INDEX IF NOT EXISTS DOMAIN_SECURITY_EVENTS__NODE_ID__DOMAIN_ID_IDX ON DOMAIN_SECURITY_EVENTS (NODE_ID, DOMAIN_ID);
+CREATE INDEX IF NOT EXISTS DOMAIN_SECURITY_EVENTS__TS_CHANGE_IDX ON DOMAIN_SECURITY_EVENTS (TS_CHANGE);
+
+CREATE TABLE IF NOT EXISTS DOMAIN_AVAILABILITY_EVENTS (
+    DOMAIN_ID INT NOT NULL,
+    NODE_ID INT NOT NULL,
+
+    AVAILABLE BOOLEAN NOT NULL, -- True if the service is available, false if it is not
+    OUTAGE_TYPE ENUM('NONE', 'TIMEOUT', 'SSL_ERROR', 'DNS_ERROR', 'CONNECTION_ERROR', 'HTTP_CLIENT_ERROR', 'HTTP_SERVER_ERROR', 'UNKNOWN') NOT NULL,
+    HTTP_STATUS_CODE INT, -- HTTP status code if available (e.g., 200, 404, etc.)
+    ERROR_MESSAGE VARCHAR(255),       -- Specific error details
+
+    TS_CHANGE TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP, -- Timestamp of the last update
+
+    AVAILABILITY_RECORD_ID BIGINT AUTO_INCREMENT,
+    P_KEY_MONTH TINYINT NOT NULL DEFAULT MONTH(TS_CHANGE), -- Month of the change for partitioning
+    PRIMARY KEY (AVAILABILITY_RECORD_ID, P_KEY_MONTH)
+)
+CHARACTER SET utf8mb4 COLLATE utf8mb4_bin
+PARTITION BY RANGE (P_KEY_MONTH) (
+    PARTITION p0 VALUES LESS THAN (1),  -- January
+    PARTITION p1 VALUES LESS THAN (2),  -- February
+    PARTITION p2 VALUES LESS THAN (3),  -- March
+    PARTITION p3 VALUES LESS THAN (4),  -- April
+    PARTITION p4 VALUES LESS THAN (5),  -- May
+    PARTITION p5 VALUES LESS THAN (6),  -- June
+    PARTITION p6 VALUES LESS THAN (7),  -- July
+    PARTITION p7 VALUES LESS THAN (8),  -- August
+    PARTITION p8 VALUES LESS THAN (9),  -- September
+    PARTITION p9 VALUES LESS THAN (10), -- October
+    PARTITION p10 VALUES LESS THAN (11), -- November
+    PARTITION p11 VALUES LESS THAN (12)  -- December
+);
+
+CREATE INDEX DOMAIN_AVAILABILITY_EVENTS__DOMAIN_ID_TS_IDX ON DOMAIN_AVAILABILITY_EVENTS (DOMAIN_ID, TS_CHANGE);
+CREATE INDEX DOMAIN_AVAILABILITY_EVENTS__TS_CHANGE_IDX ON DOMAIN_AVAILABILITY_EVENTS (TS_CHANGE);
+
+CREATE TABLE IF NOT EXISTS DOMAIN_DNS_INFORMATION (
+    DNS_ROOT_DOMAIN_ID INT AUTO_INCREMENT PRIMARY KEY,
+    ROOT_DOMAIN_NAME VARCHAR(255) NOT NULL UNIQUE,
+    NODE_AFFINITY INT NOT NULL,              -- Node ID that performs the DNS check, assign randomly across nodes
+
+    DNS_A_RECORDS TEXT,                      -- JSON array of IPv4 addresses
+    DNS_AAAA_RECORDS TEXT,                   -- JSON array of IPv6 addresses
+    DNS_CNAME_RECORD VARCHAR(255),           -- Canonical name (if applicable)
+    DNS_MX_RECORDS TEXT,                     -- JSON array of mail exchange records
+    DNS_CAA_RECORDS TEXT,                    -- Certificate Authority Authorization
+    DNS_TXT_RECORDS TEXT,                    -- TXT records (SPF, DKIM, verification, etc.)
+    DNS_NS_RECORDS TEXT,                     -- Name servers (JSON array)
+    DNS_SOA_RECORD TEXT,                     -- Start of Authority (JSON object)
+
+    TS_LAST_DNS_CHECK TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP ON UPDATE CURRENT_TIMESTAMP,
+    TS_NEXT_DNS_CHECK TIMESTAMP NOT NULL,
+    DNS_CHECK_PRIORITY TINYINT DEFAULT 0    -- Priority of the DNS check, in case we want to schedule a refresh sooner
+) CHARACTER SET utf8mb4 COLLATE utf8mb4_unicode_ci;
+
+CREATE INDEX DOMAIN_DNS_INFORMATION__PRIORITY_NEXT_CHECK_IDX ON DOMAIN_DNS_INFORMATION (NODE_AFFINITY, DNS_CHECK_PRIORITY DESC, TS_NEXT_DNS_CHECK);
+
+CREATE TABLE IF NOT EXISTS DOMAIN_DNS_EVENTS (
+     DNS_ROOT_DOMAIN_ID INT NOT NULL,
+     NODE_ID INT NOT NULL,
+
+     TS_CHANGE TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+
+-- DNS change type flags
+     CHANGE_A_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,        -- IPv4 address changes
+     CHANGE_AAAA_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,     -- IPv6 address changes
+     CHANGE_CNAME BOOLEAN NOT NULL DEFAULT FALSE,            -- CNAME changes
+     CHANGE_MX_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,       -- Mail server changes
+     CHANGE_CAA_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,      -- Certificate authority changes
+     CHANGE_TXT_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,      -- TXT record changes (SPF, DKIM, etc.)
+     CHANGE_NS_RECORDS BOOLEAN NOT NULL DEFAULT FALSE,       -- Name server changes (big red flag!)
+     CHANGE_SOA_RECORD BOOLEAN NOT NULL DEFAULT FALSE,       -- Start of Authority changes
+
+     DNS_SIGNATURE_BEFORE BLOB NOT NULL,  -- Compressed JSON snapshot of DNS records before change
+     DNS_SIGNATURE_AFTER BLOB NOT NULL,    -- Compressed JSON snapshot of DNS records after change
+
+     DNS_EVENT_ID BIGINT AUTO_INCREMENT,
+     P_KEY_MONTH TINYINT NOT NULL DEFAULT MONTH(TS_CHANGE), -- Month of the change for partitioning
+     PRIMARY KEY (DNS_EVENT_ID, P_KEY_MONTH)
+)
+CHARACTER SET utf8mb4 COLLATE utf8mb4_bin
+PARTITION BY RANGE (P_KEY_MONTH) (
+    PARTITION p0 VALUES LESS THAN (1),  -- January
+    PARTITION p1 VALUES LESS THAN (2),  -- February
+    PARTITION p2 VALUES LESS THAN (3),  -- March
+    PARTITION p3 VALUES LESS THAN (4),  -- April
+    PARTITION p4 VALUES LESS THAN (5),  -- May
+    PARTITION p5 VALUES LESS THAN (6),  -- June
+    PARTITION p6 VALUES LESS THAN (7),  -- July
+    PARTITION p7 VALUES LESS THAN (8),  -- August
+    PARTITION p8 VALUES LESS THAN (9),  -- September
+    PARTITION p9 VALUES LESS THAN (10), -- October
+    PARTITION p10 VALUES LESS THAN (11), -- November
+    PARTITION p11 VALUES LESS THAN (12)  -- December
+);
+
+CREATE INDEX DOMAIN_DNS_EVENTS__DNS_ROOT_DOMAIN_ID_TS_IDX ON DOMAIN_DNS_EVENTS (DNS_ROOT_DOMAIN_ID, TS_CHANGE);
+CREATE INDEX DOMAIN_DNS_EVENTS__TS_CHANGE_IDX ON DOMAIN_DNS_EVENTS (TS_CHANGE);

code/common/model/java/nu/marginalia/model/DocumentFormat.java

@@ -0,0 +1,24 @@
+package nu.marginalia.model;
+
+public enum DocumentFormat {
+    PLAIN(0, 1, "text"),
+    PDF(0, 1, "pdf"),
+    UNKNOWN(0, 1, "???"),
+    HTML123(0, 1, "html"),
+    HTML4(-0.1, 1.05, "html"),
+    XHTML(-0.1, 1.05, "html"),
+    HTML5(0.5, 1.1, "html");
+
+    /** Used to tune quality score */
+    public final double offset;
+    /** Used to tune quality score */
+    public final double scale;
+    public final String shortFormat;
+
+    DocumentFormat(double offset, double scale, String shortFormat) {
+        this.offset = offset;
+        this.scale = scale;
+        this.shortFormat = shortFormat;
+    }
+
+}

code/common/model/java/nu/marginalia/model/EdgeDomain.java

@@ -14,7 +14,7 @@ public class EdgeDomain implements Serializable {
     @Nonnull
     public final String topDomain;
 
-    public EdgeDomain(String host) {
+    public EdgeDomain(@Nonnull String host) {
         Objects.requireNonNull(host, "domain name must not be null");
 
         host = host.toLowerCase();
@@ -61,6 +61,10 @@ public class EdgeDomain implements Serializable {
         this.topDomain = topDomain;
     }
 
+    public static String getTopDomain(String host) {
+        return new EdgeDomain(host).topDomain;
+    }
+
     private boolean looksLikeGovTld(String host) {
         if (host.length() < 8)
             return false;
@@ -108,32 +112,6 @@ public class EdgeDomain implements Serializable {
         return topDomain;
     }
 
-    public String getDomainKey() {
-        int cutPoint = topDomain.indexOf('.');
-        if (cutPoint < 0) {
-            return topDomain;
-        }
-        return topDomain.substring(0, cutPoint).toLowerCase();
-    }
-
-    public String getLongDomainKey() {
-        StringBuilder ret = new StringBuilder();
-
-        int cutPoint = topDomain.indexOf('.');
-        if (cutPoint < 0) {
-            ret.append(topDomain);
-        } else {
-            ret.append(topDomain, 0, cutPoint);
-        }
-
-        if (!subDomain.isEmpty() && !"www".equals(subDomain)) {
-            ret.append(":");
-            ret.append(subDomain);
-        }
-
-        return ret.toString().toLowerCase();
-    }
-
     /** If possible, try to provide an alias domain,
      * i.e. a domain name that is very likely to link to this one
      * */

code/common/model/java/nu/marginalia/model/EdgeUrl.java

@@ -1,16 +1,14 @@
 package nu.marginalia.model;
 
 import nu.marginalia.util.QueryParams;
+import org.apache.commons.lang3.StringUtils;
 
 import javax.annotation.Nullable;
 import java.io.Serializable;
-import java.net.MalformedURLException;
+import java.net.*;
-import java.net.URI;
+import java.nio.charset.StandardCharsets;
-import java.net.URISyntaxException;
-import java.net.URL;
 import java.util.Objects;
 import java.util.Optional;
-import java.util.regex.Pattern;
 
 public class EdgeUrl implements Serializable {
     public final String proto;
@@ -33,7 +31,7 @@ public class EdgeUrl implements Serializable {
 
     private static URI parseURI(String url) throws URISyntaxException {
         try {
-            return new URI(urlencodeFixer(url));
+            return EdgeUriFactory.parseURILenient(url);
         } catch (URISyntaxException ex) {
             throw new URISyntaxException("Failed to parse URI '" + url + "'", ex.getMessage());
         }
@@ -51,58 +49,6 @@ public class EdgeUrl implements Serializable {
         }
     }
 
-    private static Pattern badCharPattern = Pattern.compile("[ \t\n\"<>\\[\\]()',|]");
-
-    /* Java's URI parser is a bit too strict in throwing exceptions when there's an error.
-
-       Here on the Internet, standards are like the picture on the box of the frozen pizza,
-       and what you get is more like what's on the inside, we try to patch things instead,
-       just give it a best-effort attempt att cleaning out broken or unnecessary constructions
-       like bad or missing URLEncoding
-     */
-    public static String urlencodeFixer(String url) throws URISyntaxException {
-        var s = new StringBuilder();
-        String goodChars = "&.?:/-;+$#";
-        String hexChars = "0123456789abcdefABCDEF";
-
-        int pathIdx = findPathIdx(url);
-        if (pathIdx < 0) { // url looks like http://marginalia.nu
-            return url + "/";
-        }
-        s.append(url, 0, pathIdx);
-
-        // We don't want the fragment, and multiple fragments breaks the Java URIParser for some reason
-        int end = url.indexOf("#");
-        if (end < 0) end = url.length();
-
-        for (int i = pathIdx; i < end; i++) {
-            int c = url.charAt(i);
-
-            if (goodChars.indexOf(c) >= 0 || (c >= 'A' && c <= 'Z') || (c >= 'a' && c <= 'z') || (c >= '0' && c <= '9')) {
-                s.appendCodePoint(c);
-            } else if (c == '%' && i + 2 < end) {
-                int cn = url.charAt(i + 1);
-                int cnn = url.charAt(i + 2);
-                if (hexChars.indexOf(cn) >= 0 && hexChars.indexOf(cnn) >= 0) {
-                    s.appendCodePoint(c);
-                } else {
-                    s.append("%25");
-                }
-            } else {
-                s.append(String.format("%%%02X", c));
-            }
-        }
-
-        return s.toString();
-    }
-
-    private static int findPathIdx(String url) throws URISyntaxException {
-        int colonIdx = url.indexOf(':');
-        if (colonIdx < 0 || colonIdx + 2 >= url.length()) {
-            throw new URISyntaxException(url, "Lacking protocol");
-        }
-        return url.indexOf('/', colonIdx + 2);
-    }
 
     public EdgeUrl(URI URI) {
         try {
@@ -166,11 +112,32 @@ public class EdgeUrl implements Serializable {
             sb.append(port);
         }
 
+        EdgeUriFactory.urlencodePath(sb, path);
+
+        if (param != null) {
+            EdgeUriFactory.urlencodeQuery(sb, param);
+        }
+
+        return sb.toString();
+    }
+
+
+    public String toDisplayString() {
+        StringBuilder sb = new StringBuilder(256);
+
+        sb.append(proto);
+        sb.append("://");
+        sb.append(domain);
+
+        if (port != null) {
+            sb.append(':');
+            sb.append(port);
+        }
+
         sb.append(path);
 
         if (param != null) {
-            sb.append('?');
+            sb.append('?').append(param);
-            sb.append(param);
         }
 
         return sb.toString();
@@ -247,3 +214,244 @@ public class EdgeUrl implements Serializable {
     }
 
 }
+
+class EdgeUriFactory {
+    public static URI parseURILenient(String url) throws URISyntaxException {
+
+        if (shouldOmitUrlencodeRepair(url)) {
+            try {
+                return new URI(url);
+            }
+            catch (URISyntaxException ex) {
+                // ignore and run the lenient parser
+            }
+        }
+
+        var s = new StringBuilder(url.length()+8);
+
+        int pathIdx = findPathIdx(url);
+        if (pathIdx < 0) { // url looks like http://marginalia.nu
+            return new URI(url + "/");
+        }
+        s.append(url, 0, pathIdx);
+
+        // We don't want the fragment, and multiple fragments breaks the Java URIParser for some reason
+        int end = url.indexOf("#");
+        if (end < 0) end = url.length();
+
+        int queryIdx = url.indexOf('?');
+        if (queryIdx < 0) queryIdx = end;
+
+        urlencodePath(s, url.substring(pathIdx, queryIdx));
+        if (queryIdx < end) {
+            urlencodeQuery(s, url.substring(queryIdx + 1, end));
+        }
+        return new URI(s.toString());
+    }
+
+    /** Break apart the path element of an URI into its components, and then
+     * urlencode any component that needs it, and recombine it into a single
+     * path element again.
+     */
+    public static void urlencodePath(StringBuilder sb, String path) {
+        if (path == null || path.isEmpty()) {
+            return;
+        }
+
+        String[] pathParts = StringUtils.split(path, '/');
+        if (pathParts.length == 0) {
+            sb.append('/');
+            return;
+        }
+
+        boolean shouldUrlEncode = false;
+        for (String pathPart : pathParts) {
+            if (pathPart.isEmpty()) continue;
+
+            if (needsUrlEncode(pathPart)) {
+                shouldUrlEncode = true;
+                break;
+            }
+        }
+
+        for (String pathPart : pathParts) {
+            if (pathPart.isEmpty()) continue;
+
+            if (shouldUrlEncode) {
+                sb.append('/');
+                sb.append(URLEncoder.encode(pathPart, StandardCharsets.UTF_8).replace("+", "%20"));
+            } else {
+                sb.append('/');
+                sb.append(pathPart);
+            }
+        }
+
+        if (path.endsWith("/")) {
+            sb.append('/');
+        }
+
+    }
+
+    /** Break apart the query element of a URI into its components, and then
+     * urlencode any component that needs it, and recombine it into a single
+     * query element again.
+     */
+    public static void urlencodeQuery(StringBuilder sb, String param) {
+        if (param == null || param.isEmpty()) {
+            return;
+        }
+
+        String[] queryParts = StringUtils.split(param, '&');
+
+        boolean shouldUrlEncode = false;
+        for (String queryPart : queryParts) {
+            if (queryPart.isEmpty()) continue;
+
+            if (needsUrlEncode(queryPart)) {
+                shouldUrlEncode = true;
+                break;
+            }
+        }
+
+        boolean first = true;
+        for (String queryPart : queryParts) {
+            if (queryPart.isEmpty()) continue;
+
+            if (first) {
+                sb.append('?');
+                first = false;
+            } else {
+                sb.append('&');
+            }
+
+            if (shouldUrlEncode) {
+                int idx = queryPart.indexOf('=');
+                if (idx < 0) {
+                    sb.append(URLEncoder.encode(queryPart, StandardCharsets.UTF_8));
+                } else {
+                    sb.append(URLEncoder.encode(queryPart.substring(0, idx), StandardCharsets.UTF_8));
+                    sb.append('=');
+                    sb.append(URLEncoder.encode(queryPart.substring(idx + 1), StandardCharsets.UTF_8));
+                }
+            } else {
+                sb.append(queryPart);
+            }
+        }
+    }
+
+    /** Test if the url element needs URL encoding.
+     * <p></p>
+     * Note we may have been given an already encoded path element,
+     * so we include % and + in the list of good characters
+     */
+    static boolean needsUrlEncode(String urlElement) {
+        for (int i = 0; i < urlElement.length(); i++) {
+            char c = urlElement.charAt(i);
+
+            if (isUrlSafe(c)) continue;
+            if ("+".indexOf(c) >= 0) continue;
+            if (c == '%' && i + 2 < urlElement.length()) {
+                char c1 = urlElement.charAt(i + 1);
+                char c2 = urlElement.charAt(i + 2);
+                if (isHexDigit(c1) && isHexDigit(c2)) {
+                    i += 2;
+                    continue;
+                }
+            }
+
+            return true;
+        }
+
+        return false;
+    }
+
+
+    static boolean isUrlSafe(int c) {
+        if (c >= 'a' && c <= 'z') return true;
+        if (c >= 'A' && c <= 'Z') return true;
+        if (c >= '0' && c <= '9') return true;
+        if (c == '-' || c == '_' || c == '.' || c == '~') return true;
+
+        return false;
+    }
+
+    /** Test if the URL is a valid URL that does not need to be
+     * urlencoded.
+     * <p></p>
+     * This is a very simple heuristic test that does not guarantee
+     * that the URL is valid, but it will identify cases where we
+     * are fairly certain that the URL does not need encoding,
+     * so we can skip a bunch of allocations and string operations
+     * that would otherwise be needed to fix the URL.
+     */
+    static boolean shouldOmitUrlencodeRepair(String url) {
+        int idx = 0;
+        final int len = url.length();
+
+        // Validate the scheme
+        while (idx < len - 2) {
+            char c = url.charAt(idx++);
+            if (c == ':') break;
+            if (!isAsciiAlphabetic(c)) return false;
+        }
+        if (url.charAt(idx++) != '/') return false;
+        if (url.charAt(idx++) != '/') return false;
+
+        // Validate the authority
+        while (idx < len) {
+            char c = url.charAt(idx++);
+            if (c == '/') break;
+            if (c == ':') continue;
+            if (c == '@') continue;
+            if (!isUrlSafe(c)) return false;
+        }
+
+        // Validate the path
+        if (idx >= len) return true;
+
+        while (idx < len) {
+            char c = url.charAt(idx++);
+            if (c == '?') break;
+            if (c == '/') continue;
+            if (c == '#') return true;
+            if (!isUrlSafe(c)) return false;
+        }
+
+        if (idx >= len) return true;
+
+        // Validate the query
+        while (idx < len) {
+            char c = url.charAt(idx++);
+            if (c == '&') continue;
+            if (c == '=') continue;
+            if (c == '#') return true;
+            if (!isUrlSafe(c)) return false;
+        }
+
+        return true;
+    }
+
+
+    private static boolean isAsciiAlphabetic(int c) {
+        return (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F');
+    }
+
+    private static boolean isHexDigit(int c) {
+        return (c >= '0' && c <= '9') || (c >= 'a' && c <= 'f') || (c >= 'A' && c <= 'F');
+    }
+
+    /** Find the index of the path element in a URL.
+     * <p></p>
+     * The path element starts after the scheme and authority part of the URL,
+     * which is everything up to and including the first slash after the colon.
+     */
+    private static int findPathIdx(String url) throws URISyntaxException {
+        int colonIdx = url.indexOf(':');
+        if (colonIdx < 0 || colonIdx + 3 >= url.length()) {
+            throw new URISyntaxException(url, "Lacking scheme");
+        }
+        return url.indexOf('/', colonIdx + 3);
+    }
+
+
+}

code/common/model/java/nu/marginalia/model/crawl/HtmlFeature.java

@@ -28,6 +28,8 @@ public enum HtmlFeature {
 
     GA_SPAM("special:gaspam"),
 
+    PDF("format:pdf"),
+
     /** For fingerprinting and ranking */
     OPENGRAPH("special:opengraph"),
     OPENGRAPH_IMAGE("special:opengraph:image"),

code/common/model/java/nu/marginalia/model/gson/GsonFactory.java

@@ -6,11 +6,20 @@ import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.EdgeUrl;
 
 import java.net.URISyntaxException;
+import java.time.Instant;
 
 public class GsonFactory {
     public static Gson get() {
         return new GsonBuilder()
                 .registerTypeAdapterFactory(RecordTypeAdapterFactory.builder().allowMissingComponentValues().create())
+                .registerTypeAdapter(Instant.class, (JsonSerializer<Instant>) (src, typeOfSrc, context) -> new JsonPrimitive(src.toEpochMilli()))
+                .registerTypeAdapter(Instant.class, (JsonDeserializer<Instant>) (json, typeOfT, context) -> {
+                    if (json.isJsonPrimitive() && json.getAsJsonPrimitive().isNumber()) {
+                        return Instant.ofEpochMilli(json.getAsLong());
+                    } else {
+                        throw new JsonParseException("Expected a number for Instant");
+                    }
+                })
                 .registerTypeAdapter(EdgeUrl.class, (JsonSerializer<EdgeUrl>) (src, typeOfSrc, context) -> new JsonPrimitive(src.toString()))
                 .registerTypeAdapter(EdgeDomain.class, (JsonSerializer<EdgeDomain>) (src, typeOfSrc, context) -> new JsonPrimitive(src.toString()))
                 .registerTypeAdapter(EdgeUrl.class, (JsonDeserializer<EdgeUrl>) (json, typeOfT, context) -> {

code/common/model/java/nu/marginalia/model/html/HtmlStandard.java

@@ -1,22 +0,0 @@
-package nu.marginalia.model.html;
-
-// This class really doesn't belong anywhere, but will squat here for now
-public enum HtmlStandard {
-    PLAIN(0, 1),
-    UNKNOWN(0, 1),
-    HTML123(0, 1),
-    HTML4(-0.1, 1.05),
-    XHTML(-0.1, 1.05),
-    HTML5(0.5, 1.1);
-
-    /** Used to tune quality score */
-    public final double offset;
-    /** Used to tune quality score */
-    public final double scale;
-
-    HtmlStandard(double offset, double scale) {
-        this.offset = offset;
-        this.scale = scale;
-    }
-
-}

code/common/model/java/nu/marginalia/model/idx/DocumentFlags.java

@@ -9,7 +9,7 @@ public enum DocumentFlags {
     GeneratorForum,
     GeneratorWiki,
     Sideloaded,
-    Unused7,
+    PdfFile,
     Unused8,
     ;
 

code/common/model/test/nu/marginalia/model/EdgeDomainTest.java

@@ -8,14 +8,6 @@ import static org.junit.jupiter.api.Assertions.assertEquals;
 
 class EdgeDomainTest {
 
-    @Test
-    public void testSkepdic() throws URISyntaxException {
-        var domain = new EdgeUrl("http://www.skepdic.com/astrology.html");
-        assertEquals("skepdic", domain.getDomain().getDomainKey());
-        var domain2 = new EdgeUrl("http://skepdic.com/astrology.html");
-        assertEquals("skepdic", domain2.getDomain().getDomainKey());
-    }
-
     @Test
     public void testHkDomain() throws URISyntaxException {
         var domain = new EdgeUrl("http://l7072i3.l7c.net");

code/common/model/test/nu/marginalia/model/EdgeUrlTest.java

@@ -1,6 +1,6 @@
 package nu.marginalia.model;
 
-import nu.marginalia.model.EdgeUrl;
+import org.junit.jupiter.api.Assertions;
 import org.junit.jupiter.api.Test;
 
 import java.net.URISyntaxException;
@@ -21,25 +21,70 @@ class EdgeUrlTest {
                 new EdgeUrl("https://memex.marginalia.nu/#here")
         );
     }
+
     @Test
-    public void testParam() throws URISyntaxException {
+    void testUriFromString() throws URISyntaxException {
-        System.out.println(new EdgeUrl("https://memex.marginalia.nu/index.php?id=1").toString());
+        // We test these URLs several times as we perform URLEncode-fixing both when parsing the URL and when
-        System.out.println(new EdgeUrl("https://memex.marginalia.nu/showthread.php?id=1&count=5&tracking=123").toString());
+        // converting it back to a string, we want to ensure there is no changes along the way.
-    }
+
-    @Test
+        Assertions.assertEquals("/", EdgeUriFactory.parseURILenient("https://www.example.com/").getPath());
-    void urlencodeFixer() throws URISyntaxException {
+        Assertions.assertEquals("https://www.example.com/", EdgeUriFactory.parseURILenient("https://www.example.com/").toString());
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/#heredoc"));
+        Assertions.assertEquals("https://www.example.com/", new EdgeUrl("https://www.example.com/").toString());
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/%-sign"));
+
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/%22-sign"));
+        Assertions.assertEquals("/", EdgeUriFactory.parseURILenient("https://www.example.com/#heredoc").getPath());
-        System.out.println(EdgeUrl.urlencodeFixer("https://www.example.com/\n \"huh\""));
+        Assertions.assertEquals("https://www.example.com/", EdgeUriFactory.parseURILenient("https://www.example.com/#heredoc").toString());
+        Assertions.assertEquals("https://www.example.com/", new EdgeUrl("https://www.example.com/#heredoc").toString());
+
+        Assertions.assertEquals("/trailingslash/", EdgeUriFactory.parseURILenient("https://www.example.com/trailingslash/").getPath());
+        Assertions.assertEquals("https://www.example.com/trailingslash/", EdgeUriFactory.parseURILenient("https://www.example.com/trailingslash/").toString());
+        Assertions.assertEquals("https://www.example.com/trailingslash/", new EdgeUrl("https://www.example.com/trailingslash/").toString());
+
+        Assertions.assertEquals("/%-sign", EdgeUriFactory.parseURILenient("https://www.example.com/%-sign").getPath());
+        Assertions.assertEquals("https://www.example.com/%25-sign", EdgeUriFactory.parseURILenient("https://www.example.com/%-sign").toString());
+        Assertions.assertEquals("https://www.example.com/%25-sign", new EdgeUrl("https://www.example.com/%-sign").toString());
+
+        Assertions.assertEquals("/%-sign/\"-sign", EdgeUriFactory.parseURILenient("https://www.example.com//%-sign/\"-sign").getPath());
+        Assertions.assertEquals("https://www.example.com/%25-sign/%22-sign", EdgeUriFactory.parseURILenient("https://www.example.com//%-sign/\"-sign").toString());
+        Assertions.assertEquals("https://www.example.com/%25-sign/%22-sign", new EdgeUrl("https://www.example.com//%-sign/\"-sign").toString());
+
+        Assertions.assertEquals("/\"-sign", EdgeUriFactory.parseURILenient("https://www.example.com/%22-sign").getPath());
+        Assertions.assertEquals("https://www.example.com/%22-sign", EdgeUriFactory.parseURILenient("https://www.example.com/%22-sign").toString());
+        Assertions.assertEquals("https://www.example.com/%22-sign", new EdgeUrl("https://www.example.com/%22-sign").toString());
+
+        Assertions.assertEquals("/\n \"huh\"", EdgeUriFactory.parseURILenient("https://www.example.com/\n \"huh\"").getPath());
+        Assertions.assertEquals("https://www.example.com/%0A%20%22huh%22", EdgeUriFactory.parseURILenient("https://www.example.com/\n \"huh\"").toString());
+        Assertions.assertEquals("https://www.example.com/%0A%20%22huh%22", new EdgeUrl("https://www.example.com/\n \"huh\"").toString());
+
+        Assertions.assertEquals("/wiki/Sámi", EdgeUriFactory.parseURILenient("https://en.wikipedia.org/wiki/Sámi").getPath());
+        Assertions.assertEquals("https://en.wikipedia.org/wiki/S%C3%A1mi", EdgeUriFactory.parseURILenient("https://en.wikipedia.org/wiki/Sámi").toString());
+        Assertions.assertEquals("https://en.wikipedia.org/wiki/S%C3%A1mi", new EdgeUrl("https://en.wikipedia.org/wiki/Sámi").toString());
+
+        Assertions.assertEquals("https://www.prijatelji-zivotinja.hr/index.en.php?id=2301k", new EdgeUrl("https://www.prijatelji-zivotinja.hr/index.en.php?id=2301k").toString());
     }
 
     @Test
     void testParms() throws URISyntaxException {
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?id=123"));
+        Assertions.assertEquals("id=123", new EdgeUrl("https://search.marginalia.nu/?id=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?t=123"));
+        Assertions.assertEquals("https://search.marginalia.nu/?id=123", new EdgeUrl("https://search.marginalia.nu/?id=123").toString());
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?v=123"));
+
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?m=123"));
+        Assertions.assertEquals("t=123", new EdgeUrl("https://search.marginalia.nu/?t=123").param);
-        System.out.println(new EdgeUrl("https://search.marginalia.nu/?follow=123"));
+        Assertions.assertEquals("https://search.marginalia.nu/?t=123", new EdgeUrl("https://search.marginalia.nu/?t=123").toString());
+
+        Assertions.assertEquals("v=123", new EdgeUrl("https://search.marginalia.nu/?v=123").param);
+        Assertions.assertEquals("https://search.marginalia.nu/?v=123", new EdgeUrl("https://search.marginalia.nu/?v=123").toString());
+
+        Assertions.assertEquals("id=1", new EdgeUrl("https://memex.marginalia.nu/showthread.php?id=1&count=5&tracking=123").param);
+        Assertions.assertEquals("https://memex.marginalia.nu/showthread.php?id=1",
+                new EdgeUrl("https://memex.marginalia.nu/showthread.php?id=1&count=5&tracking=123").toString());
+
+
+        Assertions.assertEquals("id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?id=1&t=5&tracking=123").param);
+        Assertions.assertEquals("https://memex.marginalia.nu/sh%C3%B6wthr%C3%ABad.php?id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?id=1&t=5&tracking=123").toString());
+
+        Assertions.assertEquals("id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?trëaking=123&id=1&t=5&").param);
+        Assertions.assertEquals("https://memex.marginalia.nu/sh%C3%B6wthr%C3%ABad.php?id=1&t=5", new EdgeUrl("https://memex.marginalia.nu/shöwthrëad.php?trëaking=123&id=1&t=5&").toString());
+
+        Assertions.assertNull(new EdgeUrl("https://search.marginalia.nu/?m=123").param);
+        Assertions.assertNull(new EdgeUrl("https://search.marginalia.nu/?follow=123").param);
     }
 }

code/common/service/java/nu/marginalia/process/control/ProcessAdHocTaskHeartbeatImpl.java

@@ -59,16 +59,13 @@ public class ProcessAdHocTaskHeartbeatImpl implements AutoCloseable, ProcessAdHo
      */
     @Override
     public void progress(String step, int stepProgress, int stepCount) {
+        int lastProgress = this.progress;
         this.step = step;
-
-
-        // off by one since we calculate the progress based on the number of steps,
-        // and Enum.ordinal() is zero-based (so the 5th step in a 5 step task is 4, not 5; resulting in the
-        // final progress being 80% and not 100%)
-
         this.progress = (int) Math.round(100. * stepProgress / (double) stepCount);
 
-        logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        if (this.progress / 10 != lastProgress / 10) {
+            logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        }
     }
 
     /** Wrap a collection to provide heartbeat progress updates as it's iterated through */

code/common/service/java/nu/marginalia/process/control/ProcessEventLog.java

@@ -0,0 +1,59 @@
+package nu.marginalia.process.control;
+
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
+import com.zaxxer.hikari.HikariDataSource;
+import nu.marginalia.process.ProcessConfiguration;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.sql.SQLException;
+import java.util.Objects;
+import java.util.UUID;
+
+@Singleton
+public class ProcessEventLog {
+    private final HikariDataSource dataSource;
+
+    private final Logger logger = LoggerFactory.getLogger(ProcessEventLog.class);
+
+    private final String serviceName;
+    private final UUID instanceUuid;
+    private final String serviceBase;
+
+    @Inject
+    public ProcessEventLog(HikariDataSource dataSource, ProcessConfiguration configuration) {
+        this.dataSource = dataSource;
+
+        this.serviceName = configuration.processName() + ":" + configuration.node();
+        this.instanceUuid = configuration.instanceUuid();
+        this.serviceBase = configuration.processName();
+
+        logger.info("Starting service {} instance {}", serviceName, instanceUuid);
+
+        logEvent("PCS-START", serviceName);
+    }
+
+    public void logEvent(Class<?> type, String message) {
+        logEvent(type.getSimpleName(), message);
+    }
+    public void logEvent(String type, String message) {
+
+        try (var conn = dataSource.getConnection();
+             var stmt = conn.prepareStatement("""
+                        INSERT INTO SERVICE_EVENTLOG(SERVICE_NAME, SERVICE_BASE, INSTANCE, EVENT_TYPE, EVENT_MESSAGE)
+                        VALUES (?, ?, ?, ?, ?)
+                     """)) {
+            stmt.setString(1, serviceName);
+            stmt.setString(2, serviceBase);
+            stmt.setString(3, instanceUuid.toString());
+            stmt.setString(4, type);
+            stmt.setString(5, Objects.requireNonNull(message, ""));
+
+            stmt.executeUpdate();
+        }
+        catch (SQLException ex) {
+            logger.error("Failed to log event {}:{}", type, message);
+        }
+    }
+}

code/common/service/java/nu/marginalia/service/control/ServiceAdHocTaskHeartbeatImpl.java

@@ -57,16 +57,13 @@ public class ServiceAdHocTaskHeartbeatImpl implements AutoCloseable, ServiceAdHo
      */
     @Override
     public void progress(String step, int stepProgress, int stepCount) {
+        int lastProgress = this.progress;
         this.step = step;
-
-
-        // off by one since we calculate the progress based on the number of steps,
-        // and Enum.ordinal() is zero-based (so the 5th step in a 5 step task is 4, not 5; resulting in the
-        // final progress being 80% and not 100%)
-
         this.progress = (int) Math.round(100. * stepProgress / (double) stepCount);
 
-        logger.info("ServiceTask {} progress: {}%", taskBase, progress);
+        if (this.progress / 10 != lastProgress / 10) {
+            logger.info("ProcessTask {} progress: {}%", taskBase, progress);
+        }
     }
 
     public void shutDown() {

code/common/service/java/nu/marginalia/service/discovery/ServiceRegistryIf.java

@@ -1,17 +1,23 @@
 package nu.marginalia.service.discovery;
 
-import nu.marginalia.service.discovery.monitor.*;
+import com.google.inject.ImplementedBy;
+import nu.marginalia.service.discovery.monitor.ServiceChangeMonitor;
+import nu.marginalia.service.discovery.monitor.ServiceMonitorIf;
 import nu.marginalia.service.discovery.property.ServiceEndpoint;
-import static nu.marginalia.service.discovery.property.ServiceEndpoint.*;
-
 import nu.marginalia.service.discovery.property.ServiceKey;
 
+import java.util.Collection;
 import java.util.List;
 import java.util.UUID;
+import java.util.function.BiConsumer;
+import java.util.function.Consumer;
+
+import static nu.marginalia.service.discovery.property.ServiceEndpoint.InstanceAddress;
 
 /** A service registry that allows services to register themselves and
  * be discovered by other services on the network.
  */
+@ImplementedBy(ZkServiceRegistry.class)
 public interface ServiceRegistryIf {
     /**
      * Register a service with the registry.
@@ -57,4 +63,9 @@ public interface ServiceRegistryIf {
      * </ul>
      * */
     void registerMonitor(ServiceMonitorIf monitor) throws Exception;
+
+    void registerProcess(String processName, int nodeId);
+    void deregisterProcess(String processName, int nodeId);
+    void watchProcess(String processName, int nodeId, Consumer<Boolean> callback) throws Exception;
+    void watchProcessAnyNode(String processName, Collection<Integer> nodes, BiConsumer<Boolean, Integer> callback) throws Exception;
 }

code/common/service/java/nu/marginalia/service/discovery/ZkServiceRegistry.java

@@ -13,11 +13,10 @@ import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import java.nio.charset.StandardCharsets;
-import java.util.ArrayList;
+import java.util.*;
-import java.util.List;
-import java.util.Random;
-import java.util.UUID;
 import java.util.concurrent.TimeUnit;
+import java.util.function.BiConsumer;
+import java.util.function.Consumer;
 
 import static nu.marginalia.service.discovery.property.ServiceEndpoint.InstanceAddress;
 
@@ -256,6 +255,90 @@ public class ZkServiceRegistry implements ServiceRegistryIf {
                 .forPath("/running-instances");
     }
 
+    @Override
+    public void registerProcess(String processName, int nodeId) {
+        String path = "/process-locks/" + processName + "/" + nodeId;
+        try {
+            curatorFramework.create()
+                    .creatingParentsIfNeeded()
+                    .withMode(CreateMode.EPHEMERAL)
+                    .forPath(path);
+            livenessPaths.add(path);
+        }
+        catch (Exception ex) {
+            logger.error("Failed to register process {} on node {}", processName, nodeId, ex);
+        }
+    }
+
+    @Override
+    public void deregisterProcess(String processName, int nodeId) {
+        String path = "/process-locks/" + processName + "/" + nodeId;
+        try {
+            curatorFramework.delete().forPath(path);
+            livenessPaths.remove(path);
+        }
+        catch (Exception ex) {
+            logger.error("Failed to deregister process {} on node {}", processName, nodeId, ex);
+        }
+    }
+
+    @Override
+    public void watchProcess(String processName, int nodeId, Consumer<Boolean> callback) throws Exception {
+        String path = "/process-locks/" + processName + "/" + nodeId;
+
+        // first check if the path exists and call the callback accordingly
+
+        if (curatorFramework.checkExists().forPath(path) != null) {
+            callback.accept(true);
+        }
+        else {
+            callback.accept(false);
+        }
+
+        curatorFramework.watchers().add()
+                .usingWatcher((Watcher) change -> {
+                    Watcher.Event.EventType type = change.getType();
+
+                    if (type == Watcher.Event.EventType.NodeCreated) {
+                        callback.accept(true);
+                    }
+                    if (type == Watcher.Event.EventType.NodeDeleted) {
+                        callback.accept(false);
+                    }
+                })
+                .forPath(path);
+
+    }
+
+    @Override
+    public void watchProcessAnyNode(String processName, Collection<Integer> nodes, BiConsumer<Boolean, Integer> callback) throws Exception {
+
+        for (int node : nodes) {
+            String path = "/process-locks/" + processName + "/" + node;
+
+            // first check if the path exists and call the callback accordingly
+            if (curatorFramework.checkExists().forPath(path) != null) {
+                callback.accept(true, node);
+            }
+            else {
+                callback.accept(false, node);
+            }
+
+            curatorFramework.watchers().add()
+                    .usingWatcher((Watcher) change -> {
+                        Watcher.Event.EventType type = change.getType();
+
+                        if (type == Watcher.Event.EventType.NodeCreated) {
+                            callback.accept(true, node);
+                        }
+                        if (type == Watcher.Event.EventType.NodeDeleted) {
+                            callback.accept(false, node);
+                        }
+                    })
+                    .forPath(path);
+        }
+    }
+
     /* Exposed for tests */
     public synchronized void shutDown() {
         if (stopped)

code/common/service/java/nu/marginalia/service/server/JoobyService.java

@@ -122,6 +122,11 @@ public class JoobyService {
         // single digit percentage difference since HTML already compresses very well with level = 1.
         options.setCompressionLevel(1);
 
+        // Set a cap on the number of worker threads, as Jooby's default value does not seem to consider
+        // multi-tenant servers with high thread counts, and spins up an exorbitant number of threads in that
+        // scenario
+        options.setWorkerThreads(Math.min(128, options.getWorkerThreads()));
+
 
         jooby.setServerOptions(options);
 

code/common/service/java/nu/marginalia/service/server/RateLimiter.java

@@ -35,21 +35,8 @@ public class RateLimiter {
     }
 
 
-    public static RateLimiter forExpensiveRequest() {
-        return new RateLimiter(5, 10);
-    }
-
     public static RateLimiter custom(int perMinute) {
-        return new RateLimiter(perMinute, 60);
+        return new RateLimiter(4 * perMinute, perMinute);
-    }
-
-    public static RateLimiter forSpamBots() {
-        return new RateLimiter(120, 3600);
-    }
-
-
-    public static RateLimiter forLogin() {
-        return new RateLimiter(3, 15);
     }
 
     private void cleanIdleBuckets() {
@@ -62,7 +49,7 @@ public class RateLimiter {
     }
 
     private Bucket createBucket() {
-        var refill = Refill.greedy(1, Duration.ofSeconds(refillRate));
+        var refill = Refill.greedy(refillRate, Duration.ofSeconds(60));
         var bw = Bandwidth.classic(capacity, refill);
         return Bucket.builder().addLimit(bw).build();
     }

code/common/service/resources/log4j2-json.xml

@@ -3,8 +3,16 @@
         <Console name="Console" target="SYSTEM_OUT">
             <PatternLayout pattern="%d{HH:mm:ss,SSS} %style{%-8markerSimpleName}{FG_Cyan} %highlight{%-5level}{FATAL=red, ERROR=red, WARN=yellow} %-24t %-20c{1}  --  %msg%n"/>
             <Filters>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ProcessConsole" target="SYSTEM_OUT">
+            <PatternLayout pattern="%style{P}{FG_Cyan} %msg%n"/>
+            <Filters>
+                <MarkerFilter marker="PROCESS" onMatch="ALLOW" onMismatch="DENY" />
             </Filters>
         </Console>
         <RollingFile name="LogToFile" fileName="${env:WMSA_LOG_DIR:-/var/log/wmsa}/wmsa-${sys:service-name}-${env:WMSA_SERVICE_NODE:-0}.log" filePattern="/var/log/wmsa/wmsa-${sys:service-name}-${env:WMSA_SERVICE_NODE:-0}-log-%d{MM-dd-yy-HH-mm-ss}-%i.log.gz"
@@ -13,15 +21,29 @@
             <Filters>
                 <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
             </Filters>
             <SizeBasedTriggeringPolicy size="10MB" />
         </RollingFile>
+        <RollingFile name="LogToFile" fileName="${env:WMSA_LOG_DIR:-/var/log/wmsa}/crawler-audit-${env:WMSA_SERVICE_NODE:-0}.log" filePattern="/var/log/wmsa/crawler-audit-${env:WMSA_SERVICE_NODE:-0}-log-%d{MM-dd-yy-HH-mm-ss}-%i.log.gz"
+                     ignoreExceptions="false">
+            <PatternLayout>
+                <Pattern>%d{yyyy-MM-dd HH:mm:ss,SSS}: %msg{nolookups}%n</Pattern>
+            </PatternLayout>
+            <SizeBasedTriggeringPolicy size="100MB" />
+            <Filters>
+                <MarkerFilter marker="CRAWLER" onMatch="ALLOW" onMismatch="DENY" />
+            </Filters>
+        </RollingFile>
     </Appenders>
     <Loggers>
         <Logger name="org.apache.zookeeper" level="WARN" />
-
+        <Logger name="org.apache.pdfbox" level="ERROR" />
+        <Logger name="org.apache.fontbox.ttf" level="ERROR" />
         <Root level="info">
             <AppenderRef ref="Console"/>
+            <AppenderRef ref="ProcessConsole"/>
             <AppenderRef ref="LogToFile"/>
         </Root>
     </Loggers>

code/common/service/resources/log4j2-prod.xml

@@ -1,10 +1,49 @@
 <Configuration xmlns="http://logging.apache.org/log4j/2.0/config" >
     <Appenders>
-        <Console name="Console" target="SYSTEM_OUT">
+        <Console name="ConsoleInfo" target="SYSTEM_OUT">
-            <PatternLayout pattern="%d{HH:mm:ss,SSS} %style{%-8markerSimpleName}{FG_Cyan} %highlight{%-5level}{FATAL=red, ERROR=red, WARN=yellow} %-24t %-20c{1}  --  %msg%n"/>
+            <PatternLayout pattern="- %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
             <Filters>
+                <LevelMatchFilter level="INFO" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleWarn" target="SYSTEM_OUT">
+            <PatternLayout pattern="⚠ %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="WARN" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleError" target="SYSTEM_OUT">
+            <PatternLayout pattern="🔥 %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="ERROR" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleFatal" target="SYSTEM_OUT">
+            <PatternLayout pattern="💀 %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="FATAL" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ProcessConsole" target="SYSTEM_OUT">
+            <PatternLayout pattern="%style{%msg%n}{FG_Cyan}"/>
+            <Filters>
+                <MarkerFilter marker="PROCESS" onMatch="ALLOW" onMismatch="DENY" />
             </Filters>
         </Console>
         <RollingFile name="LogToFile" fileName="${env:WMSA_LOG_DIR:-/var/log/wmsa}/wmsa-${sys:service-name}-${env:WMSA_SERVICE_NODE:-0}.log" filePattern="/var/log/wmsa/wmsa-${sys:service-name}-${env:WMSA_SERVICE_NODE:-0}-log-%d{MM-dd-yy-HH-mm-ss}-%i.log.gz"
@@ -17,14 +56,30 @@
                 <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="QUERY" onMatch="DENY" onMismatch="NEUTRAL" />
                 <MarkerFilter marker="HTTP" onMatch="DENY" onMismatch="NEUTRAL" />
+                <MarkerFilter marker="CRAWLER" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </RollingFile>
+        <RollingFile name="LogToFile" fileName="${env:WMSA_LOG_DIR:-/var/log/wmsa}/crawler-audit-${env:WMSA_SERVICE_NODE:-0}.log" filePattern="/var/log/wmsa/crawler-audit-${env:WMSA_SERVICE_NODE:-0}-log-%d{MM-dd-yy-HH-mm-ss}-%i.log.gz"
+                     ignoreExceptions="false">
+            <PatternLayout>
+                <Pattern>%d{yyyy-MM-dd HH:mm:ss,SSS}: %msg{nolookups}%n</Pattern>
+            </PatternLayout>
+            <SizeBasedTriggeringPolicy size="100MB" />
+            <Filters>
+                <MarkerFilter marker="CRAWLER" onMatch="ALLOW" onMismatch="DENY" />
             </Filters>
         </RollingFile>
     </Appenders>
     <Loggers>
         <Logger name="org.apache.zookeeper" level="WARN" />
-
+        <Logger name="org.apache.pdfbox" level="ERROR" />
+        <Logger name="org.apache.fontbox.ttf" level="ERROR" />
         <Root level="info">
-            <AppenderRef ref="Console"/>
+            <AppenderRef ref="ConsoleInfo"/>
+            <AppenderRef ref="ConsoleWarn"/>
+            <AppenderRef ref="ConsoleError"/>
+            <AppenderRef ref="ConsoleFatal"/>
+            <AppenderRef ref="ProcessConsole"/>
             <AppenderRef ref="LogToFile"/>
         </Root>
     </Loggers>

code/common/service/resources/log4j2-test.xml

@@ -1,15 +1,50 @@
 <Configuration xmlns="http://logging.apache.org/log4j/2.0/config" >
     <Appenders>
-        <Console name="Console" target="SYSTEM_OUT">
+        <Console name="ConsoleInfo" target="SYSTEM_OUT">
-            <PatternLayout pattern="%d{HH:mm:ss,SSS} %style{%-8markerSimpleName}{FG_Cyan} %highlight{%-5level}{FATAL=red, ERROR=red, WARN=yellow} %-24t %-20c{1}  --  %msg%n"/>
+            <PatternLayout pattern="- %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="INFO" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleWarn" target="SYSTEM_OUT">
+            <PatternLayout pattern="⚠ %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="WARN" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleError" target="SYSTEM_OUT">
+            <PatternLayout pattern="🔥 %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="ERROR" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ConsoleFatal" target="SYSTEM_OUT">
+            <PatternLayout pattern="💀 %d{HH:mm:ss,SSS} %-20c{1} -- %msg%n"/>
+            <Filters>
+                <LevelMatchFilter level="FATAL" onMatch="ALLOW" onMismatch="DENY"/>
+                <MarkerFilter marker="PROCESS" onMatch="DENY" onMismatch="NEUTRAL" />
+            </Filters>
+        </Console>
+        <Console name="ProcessConsole" target="SYSTEM_OUT">
+            <PatternLayout pattern="%style{%msg%n}{FG_Cyan}"/>
+            <Filters>
+                <MarkerFilter marker="PROCESS" onMatch="ALLOW" onMismatch="DENY" />
+            </Filters>
         </Console>
     </Appenders>
     <Loggers>
         <Logger name="org.apache.zookeeper" level="WARN" />
-
+        <Logger name="org.apache.pdfbox" level="ERROR" />
+        <Logger name="org.apache.fontbox.ttf" level="ERROR" />
         <Root level="info">
-            <AppenderRef ref="Console"/>
+            <AppenderRef ref="ConsoleInfo"/>
-            <AppenderRef ref="LogToFile"/>
+            <AppenderRef ref="ConsoleWarn"/>
+            <AppenderRef ref="ConsoleError"/>
+            <AppenderRef ref="ConsoleFatal"/>
+            <AppenderRef ref="ProcessConsole"/>
         </Root>
     </Loggers>
 </Configuration>

code/common/service/test/nu/marginalia/service/discovery/ZkServiceRegistryTest.java

@@ -25,7 +25,7 @@ import static org.mockito.Mockito.when;
 class ZkServiceRegistryTest {
     private static final int ZOOKEEPER_PORT = 2181;
     private static final GenericContainer<?> zookeeper =
-            new GenericContainer<>("zookeeper:3.8.0")
+            new GenericContainer<>("zookeeper:3.8")
                     .withExposedPorts(ZOOKEEPER_PORT);
 
     List<ZkServiceRegistry> registries = new ArrayList<>();

code/execution/api/java/nu/marginalia/executor/client/ExecutorExportClient.java

@@ -48,12 +48,13 @@ public class ExecutorExportClient {
         return msgId;
     }
 
-    public void exportSampleData(int node, FileStorageId fid, int size, String name) {
+    public void exportSampleData(int node, FileStorageId fid, int size, String ctFilter, String name) {
         channelPool.call(ExecutorExportApiBlockingStub::exportSampleData)
                 .forNode(node)
                 .run(RpcExportSampleData.newBuilder()
                         .setFileStorageId(fid.id())
                         .setSize(size)
+                        .setCtFilter(ctFilter)
                         .setName(name)
                         .build());
     }

code/execution/api/src/main/protobuf/executor-api.proto

@@ -100,6 +100,7 @@ message RpcExportSampleData {
   int64 fileStorageId = 1;
   int32 size = 2;
   string name = 3;
+  string ctFilter = 4;
 }
 message RpcDownloadSampleData {
   string sampleSet = 1;

code/execution/build.gradle

@@ -19,6 +19,7 @@ dependencies {
     implementation project(':code:processes:crawling-process')
     implementation project(':code:processes:live-crawling-process')
     implementation project(':code:processes:loading-process')
+    implementation project(':code:processes:ping-process')
     implementation project(':code:processes:converting-process')
     implementation project(':code:processes:index-constructor-process')
 
@@ -37,6 +38,7 @@ dependencies {
     implementation project(':code:functions:link-graph:api')
     implementation project(':code:functions:live-capture:api')
     implementation project(':code:functions:search-query')
+    implementation project(':code:functions:nsfw-domain-filter')
     implementation project(':code:execution:api')
 
     implementation project(':code:processes:crawling-process:model')

code/execution/java/nu/marginalia/actor/ExecutorActor.java

@@ -6,11 +6,13 @@ import java.util.Set;
 
 public enum ExecutorActor {
     PREC_EXPORT_ALL(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
+    SYNC_NSFW_LISTS(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
 
     CRAWL(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
     RECRAWL(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
     RECRAWL_SINGLE_DOMAIN(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
     PROC_CRAWLER_SPAWNER(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
+    PROC_PING_SPAWNER(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED, NodeProfile.REALTIME),
     PROC_EXPORT_TASKS_SPAWNER(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
     ADJACENCY_CALCULATION(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
     EXPORT_DATA(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED),
@@ -35,7 +37,8 @@ public enum ExecutorActor {
     LIVE_CRAWL(NodeProfile.REALTIME),
     PROC_LIVE_CRAWL_SPAWNER(NodeProfile.REALTIME),
     SCRAPE_FEEDS(NodeProfile.REALTIME),
-    UPDATE_RSS(NodeProfile.REALTIME);
+    UPDATE_RSS(NodeProfile.REALTIME)
+    ;
 
     public String id() {
         return "fsm:" + name().toLowerCase();

code/execution/java/nu/marginalia/actor/ExecutorActorControlService.java

@@ -49,6 +49,7 @@ public class ExecutorActorControlService {
                                        RecrawlSingleDomainActor recrawlSingleDomainActor,
                                        RestoreBackupActor restoreBackupActor,
                                        ConverterMonitorActor converterMonitorFSM,
+                                       PingMonitorActor pingMonitorActor,
                                        CrawlerMonitorActor crawlerMonitorActor,
                                        LiveCrawlerMonitorActor liveCrawlerMonitorActor,
                                        LoaderMonitorActor loaderMonitor,
@@ -68,6 +69,7 @@ public class ExecutorActorControlService {
                                        ExecutorActorStateMachines stateMachines,
                                        MigrateCrawlDataActor migrateCrawlDataActor,
                                        ExportAllPrecessionActor exportAllPrecessionActor,
+                                       UpdateNsfwFiltersActor updateNsfwFiltersActor,
                                        UpdateRssActor updateRssActor) throws SQLException {
         this.messageQueueFactory = messageQueueFactory;
         this.eventLog = baseServiceParams.eventLog;
@@ -88,6 +90,7 @@ public class ExecutorActorControlService {
         register(ExecutorActor.PROC_CONVERTER_SPAWNER, converterMonitorFSM);
         register(ExecutorActor.PROC_LOADER_SPAWNER, loaderMonitor);
         register(ExecutorActor.PROC_CRAWLER_SPAWNER, crawlerMonitorActor);
+        register(ExecutorActor.PROC_PING_SPAWNER, pingMonitorActor);
         register(ExecutorActor.PROC_LIVE_CRAWL_SPAWNER, liveCrawlerMonitorActor);
         register(ExecutorActor.PROC_EXPORT_TASKS_SPAWNER, exportTasksMonitorActor);
 
@@ -109,6 +112,7 @@ public class ExecutorActorControlService {
         register(ExecutorActor.UPDATE_RSS, updateRssActor);
 
         register(ExecutorActor.MIGRATE_CRAWL_DATA, migrateCrawlDataActor);
+        register(ExecutorActor.SYNC_NSFW_LISTS, updateNsfwFiltersActor);
 
         if (serviceConfiguration.node() == 1) {
             register(ExecutorActor.PREC_EXPORT_ALL, exportAllPrecessionActor);

code/execution/java/nu/marginalia/actor/proc/PingMonitorActor.java

@@ -0,0 +1,186 @@
+package nu.marginalia.actor.proc;
+
+import com.google.gson.Gson;
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
+import nu.marginalia.actor.prototype.RecordActorPrototype;
+import nu.marginalia.actor.state.ActorResumeBehavior;
+import nu.marginalia.actor.state.ActorStep;
+import nu.marginalia.actor.state.Resume;
+import nu.marginalia.actor.state.Terminal;
+import nu.marginalia.mq.MqMessageState;
+import nu.marginalia.mq.persistence.MqMessageHandlerRegistry;
+import nu.marginalia.mq.persistence.MqPersistence;
+import nu.marginalia.mqapi.ProcessInboxNames;
+import nu.marginalia.mqapi.ping.PingRequest;
+import nu.marginalia.nodecfg.NodeConfigurationService;
+import nu.marginalia.nodecfg.model.NodeProfile;
+import nu.marginalia.process.ProcessService;
+import nu.marginalia.service.module.ServiceConfiguration;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.sql.SQLException;
+import java.util.Set;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicBoolean;
+
+@Singleton
+public class PingMonitorActor extends RecordActorPrototype {
+
+    private final MqPersistence persistence;
+    private final ProcessService processService;
+
+    private final Logger logger = LoggerFactory.getLogger(getClass());
+
+    public static final int MAX_ATTEMPTS = 3;
+    private final String inboxName;
+    private final ProcessService.ProcessId processId;
+    private final ExecutorService executorService = Executors.newSingleThreadExecutor();
+    private final int node;
+    private final boolean isPrimaryNode;
+    private final Gson gson;
+
+    public record Initial() implements ActorStep {}
+    @Resume(behavior = ActorResumeBehavior.RETRY)
+    public record Monitor(int errorAttempts) implements ActorStep {}
+    @Resume(behavior = ActorResumeBehavior.RESTART)
+    public record Run(int attempts) implements ActorStep {}
+    @Terminal
+    public record Aborted() implements ActorStep {}
+
+    @Override
+    public ActorStep transition(ActorStep self) throws Exception {
+        return switch (self) {
+            case Initial i -> {
+                PingRequest request = new PingRequest(isPrimaryNode ? "primary": "secondary");
+
+                persistence.sendNewMessage(inboxName, null, null,
+                        "PingRequest",
+                        gson.toJson(request),
+                        null);
+
+                yield new Monitor(0);
+            }
+            case Monitor(int errorAttempts) -> {
+                for (;;) {
+                    var messages = persistence.eavesdrop(inboxName, 1);
+
+                    if (messages.isEmpty() && !processService.isRunning(processId)) {
+                        synchronized (processId) {
+                            processId.wait(5000);
+                        }
+
+                        if (errorAttempts > 0) { // Reset the error counter if there is silence in the inbox
+                            yield new Monitor(0);
+                        }
+                        // else continue
+                    } else {
+                        // Special: Associate this thread with the message so that we can get tracking
+                        MqMessageHandlerRegistry.register(messages.getFirst().msgId());
+
+                        yield new Run(0);
+                    }
+                }
+            }
+            case Run(int attempts) -> {
+                try {
+                    long startTime = System.currentTimeMillis();
+                    var exec = new TaskExecution();
+                    long endTime = System.currentTimeMillis();
+
+                    if (exec.isError()) {
+                        if (attempts < MAX_ATTEMPTS)
+                            yield new Run(attempts + 1);
+                        else
+                            yield new Error();
+                    }
+                    else if (endTime - startTime < TimeUnit.SECONDS.toMillis(1)) {
+                        // To avoid boot loops, we transition to error if the process
+                        // didn't run for longer than 1 seconds.  This might happen if
+                        // the process crashes before it can reach the heartbeat and inbox
+                        // stages of execution.  In this case it would not report having acted
+                        // on its message, and the process would be restarted forever without
+                        // the attempts counter incrementing.
+                        yield new Error("Process terminated within 1 seconds of starting");
+                    }
+                }
+                catch (InterruptedException ex) {
+                    // We get this exception when the process is cancelled by the user
+
+                    processService.kill(processId);
+                    setCurrentMessageToDead();
+
+                    yield new Aborted();
+                }
+
+                yield new Monitor(attempts);
+            }
+            default -> new Error();
+        };
+    }
+
+    public String describe() {
+        return "Spawns a(n) " + processId +  " process and monitors its inbox for messages";
+    }
+
+    @Inject
+    public PingMonitorActor(Gson gson,
+                                       NodeConfigurationService nodeConfigurationService,
+                                       ServiceConfiguration configuration,
+                                       MqPersistence persistence,
+                                       ProcessService processService) throws SQLException {
+        super(gson);
+        this.gson = gson;
+        this.node = configuration.node();
+        this.persistence = persistence;
+        this.processService = processService;
+        this.inboxName = ProcessInboxNames.PING_INBOX + ":" + node;
+        this.processId = ProcessService.ProcessId.PING;
+
+        this.isPrimaryNode = Set.of(NodeProfile.BATCH_CRAWL, NodeProfile.MIXED)
+                .contains(nodeConfigurationService.get(node).profile());
+    }
+
+    /** Sets the message to dead in the database to avoid
+     * the service respawning on the same task when we
+     * re-enable this actor */
+    private void setCurrentMessageToDead() {
+        try {
+            var messages = persistence.eavesdrop(inboxName, 1);
+
+            if (messages.isEmpty()) // Possibly a race condition where the task is already finished
+                return;
+
+            var theMessage = messages.iterator().next();
+            persistence.updateMessageState(theMessage.msgId(), MqMessageState.DEAD);
+        }
+        catch (SQLException ex) {
+            logger.error("Tried but failed to set the message for " + processId + " to dead", ex);
+        }
+    }
+
+    /** Encapsulates the execution of the process in a separate thread so that
+     * we can interrupt the thread if the process is cancelled */
+    private class TaskExecution {
+        private final AtomicBoolean error = new AtomicBoolean(false);
+        public TaskExecution() throws ExecutionException, InterruptedException {
+            // Run this call in a separate thread so that this thread can be interrupted waiting for it
+            executorService.submit(() -> {
+                try {
+                    processService.trigger(processId);
+                } catch (Exception e) {
+                    logger.warn("Error in triggering process", e);
+                    error.set(true);
+                }
+            }).get(); // Wait for the process to start
+        }
+
+        public boolean isError() {
+            return error.get();
+        }
+    }
+}

code/execution/java/nu/marginalia/actor/task/DownloadSampleActor.java

@@ -8,6 +8,7 @@ import nu.marginalia.actor.state.ActorResumeBehavior;
 import nu.marginalia.actor.state.ActorStep;
 import nu.marginalia.actor.state.Resume;
 import nu.marginalia.service.control.ServiceEventLog;
+import nu.marginalia.service.control.ServiceHeartbeat;
 import nu.marginalia.storage.FileStorageService;
 import nu.marginalia.storage.model.FileStorage;
 import nu.marginalia.storage.model.FileStorageId;
@@ -19,6 +20,7 @@ import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
 import java.io.*;
+import java.net.HttpURLConnection;
 import java.net.MalformedURLException;
 import java.net.URI;
 import java.net.URL;
@@ -32,6 +34,7 @@ public class DownloadSampleActor extends RecordActorPrototype {
 
     private final FileStorageService storageService;
     private final ServiceEventLog eventLog;
+    private final ServiceHeartbeat heartbeat;
     private final Logger logger = LoggerFactory.getLogger(getClass());
 
     @Resume(behavior = ActorResumeBehavior.ERROR)
@@ -66,15 +69,39 @@ public class DownloadSampleActor extends RecordActorPrototype {
 
                 Files.deleteIfExists(Path.of(tarFileName));
 
-                try (var is = new BufferedInputStream(new URI(downloadURI).toURL().openStream());
+                HttpURLConnection urlConnection = (HttpURLConnection) new URI(downloadURI).toURL().openConnection();
-                     var os = new BufferedOutputStream(Files.newOutputStream(Path.of(tarFileName), StandardOpenOption.CREATE))) {
+
-                    is.transferTo(os);
+                try (var hb = heartbeat.createServiceAdHocTaskHeartbeat("Downloading sample")) {
+                    long size = urlConnection.getContentLengthLong();
+                    byte[] buffer = new byte[8192];
+
+                    try (var is = new BufferedInputStream(urlConnection.getInputStream());
+                         var os = new BufferedOutputStream(Files.newOutputStream(Path.of(tarFileName), StandardOpenOption.CREATE))) {
+                        long copiedSize = 0;
+
+                        while (copiedSize < size) {
+                            int read = is.read(buffer);
+
+                            if (read < 0) // We've been promised a file of length 'size'
+                                throw new IOException("Unexpected end of stream");
+
+                            os.write(buffer, 0, read);
+                            copiedSize += read;
+
+                            // Update progress bar
+                            hb.progress(String.format("%d MB", copiedSize / 1024 / 1024), (int) (copiedSize / 1024), (int) (size / 1024));
+                        }
+                    }
+
                 }
                 catch (Exception ex) {
                     eventLog.logEvent(DownloadSampleActor.class, "Error downloading sample");
                     logger.error("Error downloading sample", ex);
                     yield new Error();
                 }
+                finally {
+                    urlConnection.disconnect();
+                }
 
                 eventLog.logEvent(DownloadSampleActor.class, "Download complete");
                 yield new Extract(fileStorageId, tarFileName);
@@ -170,11 +197,12 @@ public class DownloadSampleActor extends RecordActorPrototype {
     @Inject
     public DownloadSampleActor(Gson gson,
                                FileStorageService storageService,
-                               ServiceEventLog eventLog)
+                               ServiceEventLog eventLog, ServiceHeartbeat heartbeat)
     {
         super(gson);
         this.storageService = storageService;
         this.eventLog = eventLog;
+        this.heartbeat = heartbeat;
     }
 
 }

code/execution/java/nu/marginalia/actor/task/ExportSampleDataActor.java

@@ -26,32 +26,32 @@ public class ExportSampleDataActor extends RecordActorPrototype {
     private final MqOutbox exportTasksOutbox;
     private final Logger logger = LoggerFactory.getLogger(getClass());
 
-    public record Export(FileStorageId crawlId, int size, String name) implements ActorStep {}
+    public record Export(FileStorageId crawlId, int size, String ctFilter, String name) implements ActorStep {}
-    public record Run(FileStorageId crawlId, FileStorageId destId, int size, String name, long msgId) implements ActorStep {
+    public record Run(FileStorageId crawlId, FileStorageId destId, int size, String ctFilter, String name, long msgId) implements ActorStep {
-        public Run(FileStorageId crawlId, FileStorageId destId, int size, String name) {
+        public Run(FileStorageId crawlId, FileStorageId destId, int size, String name, String ctFilter) {
-            this(crawlId, destId, size, name, -1);
+            this(crawlId, destId, size, name, ctFilter,-1);
         }
     }
 
     @Override
     public ActorStep transition(ActorStep self) throws Exception {
         return switch(self) {
-            case Export(FileStorageId crawlId, int size, String name) -> {
+            case Export(FileStorageId crawlId, int size, String ctFilter, String name) -> {
                 var storage = storageService.allocateStorage(FileStorageType.EXPORT,
                         "crawl-sample-export",
                         "Crawl Data Sample " + name + "/" + size + " " + LocalDateTime.now()
                 );
 
                 if (storage == null) yield new Error("Bad storage id");
-                yield new Run(crawlId, storage.id(), size, name);
+                yield new Run(crawlId, storage.id(), size, ctFilter, name);
             }
-            case Run(FileStorageId crawlId, FileStorageId destId, int size, String name, long msgId) when msgId < 0 -> {
+            case Run(FileStorageId crawlId, FileStorageId destId, int size, String ctFilter, String name, long msgId) when msgId < 0 -> {
                 storageService.setFileStorageState(destId, FileStorageState.NEW);
 
-                long newMsgId = exportTasksOutbox.sendAsync(ExportTaskRequest.sampleData(crawlId, destId, size, name));
+                long newMsgId = exportTasksOutbox.sendAsync(ExportTaskRequest.sampleData(crawlId, destId, ctFilter, size, name));
-                yield new Run(crawlId, destId, size, name, newMsgId);
+                yield new Run(crawlId, destId, size, ctFilter, name, newMsgId);
             }
-            case Run(_, FileStorageId destId, _, _, long msgId) -> {
+            case Run(_, FileStorageId destId, _, _, _, long msgId) -> {
                 var rsp = processWatcher.waitResponse(exportTasksOutbox, ProcessService.ProcessId.EXPORT_TASKS, msgId);
 
                 if (rsp.state() != MqMessageState.OK) {
@@ -70,7 +70,7 @@ public class ExportSampleDataActor extends RecordActorPrototype {
 
     @Override
     public String describe() {
-        return "Export RSS/Atom feeds from crawl data";
+        return "Export sample crawl data";
     }
 
     @Inject

code/execution/java/nu/marginalia/actor/task/UpdateNsfwFiltersActor.java

@@ -0,0 +1,53 @@
+package nu.marginalia.actor.task;
+
+import com.google.gson.Gson;
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
+import nu.marginalia.actor.prototype.RecordActorPrototype;
+import nu.marginalia.actor.state.ActorStep;
+import nu.marginalia.nsfw.NsfwDomainFilter;
+import nu.marginalia.service.module.ServiceConfiguration;
+
+@Singleton
+public class UpdateNsfwFiltersActor extends RecordActorPrototype {
+    private final ServiceConfiguration serviceConfiguration;
+    private final NsfwDomainFilter nsfwDomainFilter;
+
+    public record Initial() implements ActorStep {}
+    public record Run() implements ActorStep {}
+
+    @Override
+    public ActorStep transition(ActorStep self) throws Exception {
+        return switch(self) {
+            case Initial() -> {
+                if (serviceConfiguration.node() != 1) {
+                    yield new Error("This actor can only run on node 1");
+                }
+                else {
+                    yield new Run();
+                }
+            }
+            case Run() -> {
+                nsfwDomainFilter.fetchLists();
+                yield new End();
+            }
+            default -> new Error();
+        };
+    }
+
+    @Override
+    public String describe() {
+        return "Sync NSFW filters";
+    }
+
+    @Inject
+    public UpdateNsfwFiltersActor(Gson gson,
+                                  ServiceConfiguration serviceConfiguration,
+                                  NsfwDomainFilter nsfwDomainFilter)
+    {
+        super(gson);
+        this.serviceConfiguration = serviceConfiguration;
+        this.nsfwDomainFilter = nsfwDomainFilter;
+    }
+
+}

code/execution/java/nu/marginalia/execution/ExecutorExportGrpcService.java

@@ -49,6 +49,7 @@ public class ExecutorExportGrpcService
                     new ExportSampleDataActor.Export(
                             FileStorageId.of(request.getFileStorageId()),
                             request.getSize(),
+                            request.getCtFilter(),
                             request.getName()
                     )
             );

code/execution/java/nu/marginalia/process/ProcessService.java

@@ -8,6 +8,7 @@ import nu.marginalia.crawl.CrawlerMain;
 import nu.marginalia.index.IndexConstructorMain;
 import nu.marginalia.livecrawler.LiveCrawlerMain;
 import nu.marginalia.loading.LoaderMain;
+import nu.marginalia.ping.PingMain;
 import nu.marginalia.service.control.ServiceEventLog;
 import nu.marginalia.service.server.BaseServiceParams;
 import nu.marginalia.task.ExportTasksMain;
@@ -41,6 +42,7 @@ public class ProcessService {
         return switch (id) {
             case "converter" -> ProcessId.CONVERTER;
             case "crawler" -> ProcessId.CRAWLER;
+            case "ping" -> ProcessId.PING;
             case "loader" -> ProcessId.LOADER;
             case "export-tasks" -> ProcessId.EXPORT_TASKS;
             case "index-constructor" -> ProcessId.INDEX_CONSTRUCTOR;
@@ -50,6 +52,7 @@ public class ProcessService {
 
     public enum ProcessId {
         CRAWLER(CrawlerMain.class),
+        PING(PingMain.class),
         LIVE_CRAWLER(LiveCrawlerMain.class),
         CONVERTER(ConverterMain.class),
         LOADER(LoaderMain.class),
@@ -68,6 +71,7 @@ public class ProcessService {
                 case LIVE_CRAWLER -> "LIVE_CRAWLER_PROCESS_OPTS";
                 case CONVERTER -> "CONVERTER_PROCESS_OPTS";
                 case LOADER -> "LOADER_PROCESS_OPTS";
+                case PING -> "PING_PROCESS_OPTS";
                 case INDEX_CONSTRUCTOR -> "INDEX_CONSTRUCTION_PROCESS_OPTS";
                 case EXPORT_TASKS -> "EXPORT_TASKS_PROCESS_OPTS";
             };

code/features-search/random-websites/java/nu/marginalia/browse/DbBrowseDomainsRandom.java

@@ -27,10 +27,12 @@ public class DbBrowseDomainsRandom {
     public List<BrowseResult> getRandomDomains(int count, DomainBlacklist blacklist, int set) {
 
         final String q = """
-                SELECT DOMAIN_ID, DOMAIN_NAME, INDEXED
+                SELECT EC_RANDOM_DOMAINS.DOMAIN_ID, DOMAIN_NAME, INDEXED
                 FROM EC_RANDOM_DOMAINS
                 INNER JOIN EC_DOMAIN ON EC_DOMAIN.ID=DOMAIN_ID
+                LEFT JOIN DOMAIN_AVAILABILITY_INFORMATION DAI ON DAI.DOMAIN_ID=EC_RANDOM_DOMAINS.DOMAIN_ID
                 WHERE STATE<2
+                AND SERVER_AVAILABLE
                 AND DOMAIN_SET=?
                 AND DOMAIN_ALIAS IS NULL
                 ORDER BY RAND()

code/functions/live-capture/build.gradle

@@ -25,9 +25,9 @@ dependencies {
 
     implementation project(':code:execution:api')
     implementation project(':code:processes:crawling-process:ft-content-type')
+    implementation project(':third-party:rssreader')
 
     implementation libs.jsoup
-    implementation project(':third-party:rssreader')
     implementation libs.opencsv
     implementation libs.slop
     implementation libs.sqlite
@@ -57,8 +57,6 @@ dependencies {
     implementation libs.bundles.gson
     implementation libs.bundles.mariadb
 
-
-
     testImplementation libs.bundles.slf4j.test
     testImplementation libs.bundles.junit
     testImplementation libs.mockito

code/functions/live-capture/java/nu/marginalia/domsample/DomSampleService.java

@@ -0,0 +1,126 @@
+package nu.marginalia.domsample;
+
+import com.google.inject.Inject;
+import com.zaxxer.hikari.HikariDataSource;
+import jakarta.inject.Named;
+import nu.marginalia.domsample.db.DomSampleDb;
+import nu.marginalia.livecapture.BrowserlessClient;
+import nu.marginalia.service.module.ServiceConfiguration;
+import org.apache.commons.lang3.StringUtils;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.net.URI;
+import java.net.URISyntaxException;
+import java.time.Duration;
+import java.util.HashSet;
+import java.util.Set;
+import java.util.concurrent.TimeUnit;
+
+public class DomSampleService {
+    private final DomSampleDb db;
+    private final HikariDataSource mariadbDataSource;
+    private final URI browserlessURI;
+
+    private static final Logger logger = LoggerFactory.getLogger(DomSampleService.class);
+
+    @Inject
+    public DomSampleService(DomSampleDb db,
+                            HikariDataSource mariadbDataSource,
+                            @Named("browserless-uri") String browserlessAddress,
+                            ServiceConfiguration serviceConfiguration)
+            throws URISyntaxException
+    {
+        this.db = db;
+        this.mariadbDataSource = mariadbDataSource;
+
+        if (StringUtils.isEmpty(browserlessAddress) || serviceConfiguration.node() > 1) {
+            logger.warn("Live capture service will not run");
+            browserlessURI = null;
+        }
+        else {
+            browserlessURI = new URI(browserlessAddress);
+        }
+    }
+
+    public void start() {
+        if (browserlessURI == null) {
+            logger.warn("DomSampleService is not enabled due to missing browserless URI or multi-node configuration");
+            return;
+        }
+
+        Thread.ofPlatform().daemon().start(this::run);
+    }
+
+    public void syncDomains() {
+        Set<String> dbDomains = new HashSet<>();
+
+        logger.info("Fetching domains from database...");
+
+        try (var conn = mariadbDataSource.getConnection();
+            var stmt = conn.prepareStatement("""
+                SELECT DOMAIN_NAME 
+                FROM EC_DOMAIN 
+                WHERE NODE_AFFINITY>0
+                """)
+        ) {
+            var rs = stmt.executeQuery();
+            while (rs.next()) {
+                dbDomains.add(rs.getString("DOMAIN_NAME"));
+            }
+        } catch (Exception e) {
+            throw new RuntimeException("Failed to sync domains", e);
+        }
+
+        logger.info("Found {} domains in database", dbDomains.size());
+
+        db.syncDomains(dbDomains);
+
+        logger.info("Synced domains to sqlite");
+    }
+
+    public void run() {
+
+        try (var client = new BrowserlessClient(browserlessURI)) {
+
+            while (!Thread.currentThread().isInterrupted()) {
+
+                try {
+                    // Grace sleep in case we're operating on an empty domain list
+                    TimeUnit.SECONDS.sleep(15);
+
+                    syncDomains();
+                    var domains = db.getScheduledDomains();
+
+                    for (var domain : domains) {
+                        updateDomain(client, domain);
+                    }
+                } catch (InterruptedException e) {
+                    Thread.currentThread().interrupt();
+                    logger.info("DomSampleService interrupted, stopping...");
+                    return;
+                } catch (Exception e) {
+                    logger.error("Error in DomSampleService run loop", e);
+                }
+            }
+
+        }
+    }
+
+    private void updateDomain(BrowserlessClient client, String domain) {
+        var rootUrl = "https://" + domain + "/";
+        try {
+            var content = client.annotatedContent(rootUrl, new BrowserlessClient.GotoOptions("load", Duration.ofSeconds(10).toMillis()));
+
+            if (content.isPresent()) {
+                db.saveSample(domain, rootUrl, content.get());
+            }
+        } catch (Exception e) {
+            logger.error("Failed to process domain: " + domain, e);
+        }
+        finally {
+            db.flagDomainAsFetched(domain);
+        }
+    }
+
+}

code/functions/live-capture/java/nu/marginalia/domsample/db/DomSampleDb.java

@@ -0,0 +1,174 @@
+package nu.marginalia.domsample.db;
+
+import nu.marginalia.WmsaHome;
+import org.jsoup.Jsoup;
+
+import java.nio.file.Path;
+import java.sql.Connection;
+import java.sql.DriverManager;
+import java.sql.SQLException;
+import java.util.*;
+
+public class DomSampleDb implements AutoCloseable {
+    private static final String dbFileName = "dom-sample.db";
+    private final Connection connection;
+
+    public DomSampleDb() throws SQLException{
+        this(WmsaHome.getDataPath().resolve(dbFileName));
+    }
+
+    public DomSampleDb(Path dbPath) throws SQLException {
+        String dbUrl = "jdbc:sqlite:" + dbPath.toAbsolutePath();
+
+        connection = DriverManager.getConnection(dbUrl);
+
+        try (var stmt = connection.createStatement()) {
+            stmt.executeUpdate("CREATE TABLE IF NOT EXISTS samples (url TEXT PRIMARY KEY, domain TEXT, sample BLOB, requests BLOB, accepted_popover BOOLEAN DEFAULT FALSE)");
+            stmt.executeUpdate("CREATE INDEX IF NOT EXISTS domain_index ON samples (domain)");
+            stmt.executeUpdate("CREATE TABLE IF NOT EXISTS schedule (domain TEXT PRIMARY KEY, last_fetch TIMESTAMP DEFAULT NULL)");
+            stmt.execute("PRAGMA journal_mode=WAL");
+        }
+
+    }
+
+    public void syncDomains(Set<String> domains) {
+        Set<String> currentDomains = new HashSet<>();
+        try (var stmt = connection.prepareStatement("SELECT domain FROM schedule")) {
+            var rs = stmt.executeQuery();
+            while (rs.next()) {
+                currentDomains.add(rs.getString("domain"));
+            }
+        } catch (SQLException e) {
+            throw new RuntimeException("Failed to sync domains", e);
+        }
+
+        Set<String> toRemove = new HashSet<>(currentDomains);
+        Set<String> toAdd = new HashSet<>(domains);
+
+        toRemove.removeAll(domains);
+        toAdd.removeAll(currentDomains);
+
+        try (var removeStmt = connection.prepareStatement("DELETE FROM schedule WHERE domain = ?");
+                var addStmt = connection.prepareStatement("INSERT OR IGNORE INTO schedule (domain) VALUES (?)")
+        ) {
+            for (String domain : toRemove) {
+                removeStmt.setString(1, domain);
+                removeStmt.executeUpdate();
+            }
+
+            for (String domain : toAdd) {
+                addStmt.setString(1, domain);
+                addStmt.executeUpdate();
+            }
+        } catch (SQLException e) {
+            throw new RuntimeException("Failed to remove domains", e);
+        }
+    }
+
+    public List<String> getScheduledDomains() {
+        List<String> domains = new ArrayList<>();
+        try (var stmt = connection.prepareStatement("SELECT domain FROM schedule ORDER BY last_fetch IS NULL DESC, last_fetch ASC")) {
+            var rs = stmt.executeQuery();
+            while (rs.next()) {
+                domains.add(rs.getString("domain"));
+            }
+        } catch (SQLException e) {
+            throw new RuntimeException("Failed to get scheduled domains", e);
+        }
+        return domains;
+    }
+
+    public void flagDomainAsFetched(String domain) {
+        try (var stmt = connection.prepareStatement("INSERT OR REPLACE INTO schedule (domain, last_fetch) VALUES (?, CURRENT_TIMESTAMP)")) {
+            stmt.setString(1, domain);
+            stmt.executeUpdate();
+        } catch (SQLException e) {
+            throw new RuntimeException("Failed to flag domain as fetched", e);
+        }
+    }
+
+
+    public record Sample(String url, String domain, String sample, String requests, boolean acceptedPopover) {}
+
+    public List<Sample> getSamples(String domain) throws SQLException {
+        List<Sample> samples = new ArrayList<>();
+
+        try (var stmt = connection.prepareStatement("""
+                SELECT url, sample, requests, accepted_popover
+                FROM samples 
+                WHERE domain = ?
+                """))
+        {
+            stmt.setString(1, domain);
+            var rs = stmt.executeQuery();
+            while (rs.next()) {
+                samples.add(
+                        new Sample(
+                                rs.getString("url"),
+                                domain,
+                                rs.getString("sample"),
+                                rs.getString("requests"),
+                                rs.getBoolean("accepted_popover")
+                        )
+                );
+            }
+        }
+        return samples;
+    }
+
+    public void saveSample(String domain, String url, String rawContent) throws SQLException {
+        var doc = Jsoup.parse(rawContent);
+
+        var networkRequests = doc.getElementById("marginalia-network-requests");
+
+        boolean acceptedPopover = false;
+
+        StringBuilder requestTsv = new StringBuilder();
+        if (networkRequests != null) {
+
+            acceptedPopover = !networkRequests.getElementsByClass("marginalia-agreed-cookies").isEmpty();
+
+            for (var request : networkRequests.getElementsByClass("network-request")) {
+                String method = request.attr("data-method");
+                String urlAttr = request.attr("data-url");
+                String timestamp = request.attr("data-timestamp");
+
+                requestTsv
+                        .append(method)
+                        .append('\t')
+                        .append(timestamp)
+                        .append('\t')
+                        .append(urlAttr.replace('\n', ' '))
+                        .append("\n");
+            }
+
+            networkRequests.remove();
+        }
+
+        doc.body().removeAttr("id");
+
+        String sample = doc.html();
+
+        saveSampleRaw(domain, url, sample, requestTsv.toString().trim(), acceptedPopover);
+
+    }
+
+    public void saveSampleRaw(String domain, String url, String sample, String requests, boolean acceptedPopover) throws SQLException {
+        try (var stmt = connection.prepareStatement("""
+                INSERT OR REPLACE 
+                INTO samples (domain, url, sample, requests, accepted_popover) 
+                VALUES (?, ?, ?, ?, ?)
+                """)) {
+            stmt.setString(1, domain);
+            stmt.setString(2, url);
+            stmt.setString(3, sample);
+            stmt.setString(4, requests);
+            stmt.setBoolean(5, acceptedPopover);
+            stmt.executeUpdate();
+        }
+    }
+
+    public void close() throws SQLException {
+        connection.close();
+    }
+}

code/functions/live-capture/java/nu/marginalia/livecapture/BrowserlessClient.java

@@ -8,10 +8,13 @@ import org.slf4j.LoggerFactory;
 
 import java.io.IOException;
 import java.net.URI;
+import java.net.URLEncoder;
 import java.net.http.HttpClient;
 import java.net.http.HttpRequest;
 import java.net.http.HttpResponse;
+import java.nio.charset.StandardCharsets;
 import java.time.Duration;
+import java.util.List;
 import java.util.Map;
 import java.util.Optional;
 
@@ -60,6 +63,42 @@ public class BrowserlessClient implements AutoCloseable {
         return Optional.of(rsp.body());
     }
 
+    /** Fetches content with a marginalia hack extension loaded that decorates the DOM with attributes for
+     * certain CSS attributes, to be able to easier identify popovers and other nuisance elements.
+     */
+    public Optional<String> annotatedContent(String url, GotoOptions gotoOptions) throws IOException, InterruptedException {
+        Map<String, Object> requestData = Map.of(
+                "url", url,
+                "userAgent", userAgent,
+                "gotoOptions", gotoOptions,
+                "waitForSelector", Map.of("selector", "#marginaliahack", "timeout", 15000)
+        );
+
+        // Launch parameters for the browserless instance to load the extension
+        Map<String, Object> launchParameters = Map.of(
+                "args", List.of("--load-extension=/dom-export")
+        );
+
+        String launchParametersStr = URLEncoder.encode(gson.toJson(launchParameters), StandardCharsets.UTF_8);
+
+        var request = HttpRequest.newBuilder()
+                .uri(browserlessURI.resolve("/content?token="+BROWSERLESS_TOKEN+"&launch="+launchParametersStr))
+                .method("POST", HttpRequest.BodyPublishers.ofString(
+                        gson.toJson(requestData)
+                ))
+                .header("Content-type", "application/json")
+                .build();
+
+        var rsp = httpClient.send(request, HttpResponse.BodyHandlers.ofString());
+
+        if (rsp.statusCode() >= 300) {
+            logger.info("Failed to fetch annotated content for {}, status {}", url, rsp.statusCode());
+            return Optional.empty();
+        }
+
+        return Optional.of(rsp.body());
+    }
+
     public byte[] screenshot(String url, GotoOptions gotoOptions, ScreenshotOptions screenshotOptions)
             throws IOException, InterruptedException {
 

code/functions/live-capture/java/nu/marginalia/livecapture/LiveCaptureGrpcService.java

@@ -126,7 +126,6 @@ public class LiveCaptureGrpcService
                 }
                 else {
                     EdgeDomain domain = domainNameOpt.get();
-                    String domainNameStr = domain.toString();
 
                     if (!isValidDomainForCapture(domain)) {
                         ScreenshotDbOperations.flagDomainAsFetched(conn, domain);

code/functions/live-capture/java/nu/marginalia/rss/svc/FeedFetcherService.java

@@ -33,6 +33,7 @@ import java.sql.SQLException;
 import java.time.*;
 import java.time.format.DateTimeFormatter;
 import java.util.*;
+import java.util.concurrent.ExecutorService;
 import java.util.concurrent.Executors;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicInteger;
@@ -71,7 +72,7 @@ public class FeedFetcherService {
     public enum UpdateMode {
         CLEAN,
         REFRESH
-    };
+    }
 
     public void updateFeeds(UpdateMode updateMode) throws IOException {
         if (updating) // Prevent concurrent updates
@@ -87,6 +88,7 @@ public class FeedFetcherService {
                 .followRedirects(HttpClient.Redirect.NORMAL)
                 .version(HttpClient.Version.HTTP_2)
                 .build();
+             ExecutorService fetchExecutor = Executors.newCachedThreadPool();
              FeedJournal feedJournal = FeedJournal.create();
              var heartbeat = serviceHeartbeat.createServiceAdHocTaskHeartbeat("Update Rss Feeds")
         ) {
@@ -131,7 +133,7 @@ public class FeedFetcherService {
 
                         FetchResult feedData;
                         try (DomainLocks.DomainLock domainLock = domainLocks.lockDomain(new EdgeDomain(feed.domain()))) {
-                            feedData = fetchFeedData(feed, client, ifModifiedSinceDate, ifNoneMatchTag);
+                            feedData = fetchFeedData(feed, client, fetchExecutor, ifModifiedSinceDate, ifNoneMatchTag);
                         } catch (Exception ex) {
                             feedData = new FetchResult.TransientError();
                         }
@@ -211,6 +213,7 @@ public class FeedFetcherService {
 
     private FetchResult fetchFeedData(FeedDefinition feed,
                                       HttpClient client,
+                                      ExecutorService executorService,
                                       @Nullable String ifModifiedSinceDate,
                                       @Nullable String ifNoneMatchTag)
     {
@@ -226,18 +229,27 @@ public class FeedFetcherService {
                     .timeout(Duration.ofSeconds(15))
                     ;
 
-            if (ifModifiedSinceDate != null) {
+            // Set the If-Modified-Since or If-None-Match headers if we have them
+            // though since there are certain idiosyncrasies in server implementations,
+            // we avoid setting both at the same time as that may turn a 304 into a 200.
+            if (ifNoneMatchTag != null) {
+                requestBuilder.header("If-None-Match", ifNoneMatchTag);
+            } else if (ifModifiedSinceDate != null) {
                 requestBuilder.header("If-Modified-Since", ifModifiedSinceDate);
             }
 
-            if (ifNoneMatchTag != null) {
-                requestBuilder.header("If-None-Match", ifNoneMatchTag);
-            }
 
             HttpRequest getRequest = requestBuilder.build();
 
             for (int i = 0; i < 3; i++) {
-                HttpResponse<byte[]> rs = client.send(getRequest, HttpResponse.BodyHandlers.ofByteArray());
+
+                /* Note we need to use an executor to time-limit the send() method in HttpClient, as
+                 * its support for timeouts only applies to the time until response starts to be received,
+                 * and does not catch the case when the server starts to send data but then hangs.
+                 */
+                HttpResponse<byte[]> rs = executorService.submit(
+                        () -> client.send(getRequest, HttpResponse.BodyHandlers.ofByteArray()))
+                                .get(15, TimeUnit.SECONDS);
 
                 if (rs.statusCode() == 429) { // Too Many Requests
                     int retryAfter = Integer.parseInt(rs.headers().firstValue("Retry-After").orElse("2"));

code/functions/live-capture/java/nu/marginalia/rss/svc/SimpleFeedParser.java

@@ -79,9 +79,17 @@ public class SimpleFeedParser {
                 if (!link.isBlank())
                     break;
                 var tag = element.getElementsByTag(attr).first();
+
                 if (tag != null) {
-                    link = tag.text();
+                    String linkText = tag.text();
+
+                    if (linkText.isBlank()) {
+                        linkText = tag.attr("href");
+                    }
+
+                    link = linkText;
                 }
+
             }
 
             ret.add(new ItemData(title, description, link, pubDate));

code/functions/live-capture/test/nu/marginalia/domsample/db/DomSampleDbTest.java

@@ -0,0 +1,113 @@
+package nu.marginalia.domsample.db;
+
+import org.junit.jupiter.api.AfterEach;
+import org.junit.jupiter.api.BeforeEach;
+import org.junit.jupiter.api.Test;
+import org.testcontainers.shaded.org.apache.commons.io.FileUtils;
+
+import java.io.IOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.*;
+
+import static org.junit.jupiter.api.Assertions.*;
+
+class DomSampleDbTest {
+    Path tempDir;
+
+    @BeforeEach
+    void setUp() throws Exception {
+        tempDir = Files.createTempDirectory("test");
+    }
+
+    @AfterEach
+    void tearDown() throws IOException {
+        FileUtils.deleteDirectory(tempDir.toFile());
+    }
+
+    @Test
+    public void testSetUp() {
+        var dbPath = tempDir.resolve("test.db");
+        try (var db = new DomSampleDb(dbPath)) {
+        }
+        catch (Exception e) {
+            fail("Failed to set up database: " + e.getMessage());
+        }
+    }
+
+    @Test
+    public void testSyncDomains() {
+        var dbPath = tempDir.resolve("test.db");
+        try (var db = new DomSampleDb(dbPath)) {
+
+            db.syncDomains(Set.of("example.com", "test.com", "foobar.com"));
+            assertEquals(Set.of("example.com", "test.com", "foobar.com"), new HashSet<>(db.getScheduledDomains()));
+            db.syncDomains(Set.of("example.com", "test.com"));
+            assertEquals(Set.of("example.com", "test.com"), new HashSet<>(db.getScheduledDomains()));
+            db.syncDomains(Set.of("foobar.com", "test.com"));
+            assertEquals(Set.of("foobar.com", "test.com"), new HashSet<>(db.getScheduledDomains()));
+        }
+        catch (Exception e) {
+            fail("Failed to sync domains: " + e.getMessage());
+        }
+    }
+
+    @Test
+    public void testFetchDomains() {
+        var dbPath = tempDir.resolve("test.db");
+        try (var db = new DomSampleDb(dbPath)) {
+
+            db.syncDomains(Set.of("example.com", "test.com", "foobar.com"));
+            db.flagDomainAsFetched("example.com");
+            db.flagDomainAsFetched("test.com");
+            db.flagDomainAsFetched("foobar.com");
+            assertEquals(List.of("example.com", "test.com", "foobar.com"), db.getScheduledDomains());
+            db.flagDomainAsFetched("test.com");
+            assertEquals(List.of("example.com", "foobar.com", "test.com"), db.getScheduledDomains());
+        }
+        catch (Exception e) {
+            fail("Failed to sync domains: " + e.getMessage());
+        }
+    }
+
+    @Test
+    public void saveLoadSingle() {
+        var dbPath = tempDir.resolve("test.db");
+        try (var db = new DomSampleDb(dbPath)) {
+            db.saveSampleRaw("example.com", "http://example.com/sample", "sample data", "requests data", true);
+            var samples = db.getSamples("example.com");
+            assertEquals(1, samples.size());
+            var sample = samples.getFirst();
+            assertEquals("example.com", sample.domain());
+            assertEquals("http://example.com/sample", sample.url());
+            assertEquals("sample data", sample.sample());
+            assertEquals("requests data", sample.requests());
+            assertTrue(sample.acceptedPopover());
+        }
+        catch (Exception e) {
+            fail("Failed to save/load sample: " + e.getMessage());
+        }
+    }
+
+    @Test
+    public void saveLoadTwo() {
+        var dbPath = tempDir.resolve("test.db");
+        try (var db = new DomSampleDb(dbPath)) {
+            db.saveSampleRaw("example.com", "http://example.com/sample", "sample data", "r1", true);
+            db.saveSampleRaw("example.com", "http://example.com/sample2", "sample data2", "r2", false);
+            var samples = db.getSamples("example.com");
+            assertEquals(2, samples.size());
+
+            Map<String, String> samplesByUrl = new HashMap<>();
+            for (var sample : samples) {
+                samplesByUrl.put(sample.url(), sample.sample());
+            }
+
+            assertEquals("sample data", samplesByUrl.get("http://example.com/sample"));
+            assertEquals("sample data2", samplesByUrl.get("http://example.com/sample2"));
+        }
+        catch (Exception e) {
+            fail("Failed to save/load sample: " + e.getMessage());
+        }
+    }
+}

code/functions/live-capture/test/nu/marginalia/livecapture/BrowserlessClientTest.java

@@ -3,17 +3,21 @@ package nu.marginalia.livecapture;
 import com.github.tomakehurst.wiremock.WireMockServer;
 import com.github.tomakehurst.wiremock.core.WireMockConfiguration;
 import nu.marginalia.WmsaHome;
+import nu.marginalia.domsample.db.DomSampleDb;
 import nu.marginalia.service.module.ServiceConfigurationModule;
 import org.junit.jupiter.api.Assertions;
 import org.junit.jupiter.api.BeforeAll;
 import org.junit.jupiter.api.Tag;
 import org.junit.jupiter.api.Test;
 import org.testcontainers.containers.GenericContainer;
+import org.testcontainers.images.PullPolicy;
 import org.testcontainers.junit.jupiter.Testcontainers;
 import org.testcontainers.utility.DockerImageName;
 
 import java.io.IOException;
 import java.net.URI;
+import java.nio.file.Files;
+import java.nio.file.Path;
 import java.util.Map;
 
 import static com.github.tomakehurst.wiremock.client.WireMock.*;
@@ -22,9 +26,14 @@ import static com.github.tomakehurst.wiremock.client.WireMock.*;
 @Testcontainers
 @Tag("slow")
 public class BrowserlessClientTest {
-    static GenericContainer<?> container = new GenericContainer<>(DockerImageName.parse("browserless/chrome"))
+    // Run gradle docker if this image is not available
+    static GenericContainer<?> container = new GenericContainer<>(DockerImageName.parse("marginalia-browserless"))
             .withEnv(Map.of("TOKEN", "BROWSERLESS_TOKEN"))
+            .withImagePullPolicy(PullPolicy.defaultPolicy())
             .withNetworkMode("bridge")
+            .withLogConsumer(frame -> {
+                System.out.print(frame.getUtf8String());
+            })
             .withExposedPorts(3000);
 
     static WireMockServer wireMockServer =
@@ -34,6 +43,7 @@ public class BrowserlessClientTest {
     static String localIp;
 
     static URI browserlessURI;
+    static URI browserlessWssURI;
 
     @BeforeAll
     public static void setup() throws IOException {
@@ -44,6 +54,12 @@ public class BrowserlessClientTest {
                 container.getMappedPort(3000))
         );
 
+        browserlessWssURI = URI.create(String.format("ws://%s:%d/?token=BROWSERLESS_TOKEN",
+                container.getHost(),
+                container.getMappedPort(3000))
+        );
+
+
         wireMockServer.start();
         wireMockServer.stubFor(get("/").willReturn(aResponse().withStatus(200).withBody("Ok")));
 
@@ -85,6 +101,30 @@ public class BrowserlessClientTest {
         }
     }
 
+    @Test
+    public void testAnnotatedContent() throws Exception {
+
+        try (var client = new BrowserlessClient(browserlessURI);
+             DomSampleDb dbop = new DomSampleDb(Path.of("/tmp/dom-sample.db"))
+        ) {
+            var content = client.annotatedContent("https://marginalia.nu/", BrowserlessClient.GotoOptions.defaultValues()).orElseThrow();
+            dbop.saveSample("marginalia.nu", "https://marginalia.nu/", content);
+            System.out.println(content);
+            Assertions.assertFalse(content.isBlank(), "Content should not be empty");
+
+            dbop.getSamples("marginalia.nu").forEach(sample -> {
+                System.out.println("Sample URL: " + sample.url());
+                System.out.println("Sample Content: " + sample.sample());
+                System.out.println("Sample Requests: " + sample.requests());
+                System.out.println("Accepted Popover: " + sample.acceptedPopover());
+            });
+        }
+        finally {
+            Files.deleteIfExists(Path.of("/tmp/dom-sample.db"));
+        }
+
+    }
+
     @Test
     public void testScreenshot() throws Exception {
         try (var client = new BrowserlessClient(browserlessURI)) {

code/functions/nsfw-domain-filter/build.gradle

@@ -0,0 +1,43 @@
+plugins {
+    id 'java'
+    id 'jvm-test-suite'
+}
+
+java {
+    toolchain {
+        languageVersion.set(JavaLanguageVersion.of(rootProject.ext.jvmVersion))
+    }
+}
+
+apply from: "$rootProject.projectDir/srcsets.gradle"
+
+dependencies {
+
+    implementation project(':code:common:config')
+    implementation project(':code:common:model')
+    implementation project(':code:common:db')
+
+
+    implementation libs.bundles.slf4j
+    implementation libs.prometheus
+    implementation libs.guava
+    implementation libs.commons.lang3
+    implementation dependencies.create(libs.guice.get()) {
+        exclude group: 'com.google.guava'
+    }
+    implementation libs.notnull
+    implementation libs.fastutil
+    implementation libs.bundles.mariadb
+
+
+    testImplementation libs.bundles.slf4j.test
+    testImplementation libs.bundles.junit
+    testImplementation libs.mockito
+
+    testImplementation platform('org.testcontainers:testcontainers-bom:1.17.4')
+    testImplementation libs.commons.codec
+    testImplementation project(':code:common:service')
+    testImplementation 'org.testcontainers:mariadb:1.17.4'
+    testImplementation 'org.testcontainers:junit-jupiter:1.17.4'
+    testImplementation project(':code:libraries:test-helpers')
+}

code/functions/nsfw-domain-filter/java/nu/marginalia/nsfw/NsfwDomainFilter.java

@@ -0,0 +1,192 @@
+package nu.marginalia.nsfw;
+
+import com.google.inject.Inject;
+import com.google.inject.Singleton;
+import com.google.inject.name.Named;
+import com.zaxxer.hikari.HikariDataSource;
+import it.unimi.dsi.fastutil.ints.IntOpenHashSet;
+import org.apache.commons.lang3.StringUtils;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.BufferedReader;
+import java.io.ByteArrayInputStream;
+import java.io.InputStreamReader;
+import java.net.http.HttpClient;
+import java.net.http.HttpRequest;
+import java.net.http.HttpResponse;
+import java.sql.SQLException;
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.List;
+import java.util.concurrent.TimeUnit;
+import java.util.zip.GZIPInputStream;
+
+@Singleton
+public class NsfwDomainFilter {
+    private final HikariDataSource dataSource;
+
+    private final List<String> dangerLists;
+    private final List<String> smutLists;
+
+    private volatile IntOpenHashSet blockedDomainIdsTier1 = new IntOpenHashSet();
+    private volatile IntOpenHashSet blockedDomainIdsTier2 = new IntOpenHashSet();
+
+    private static final Logger logger = LoggerFactory.getLogger(NsfwDomainFilter.class);
+
+    public static final int NSFW_DISABLE = 0;
+    public static final int NSFW_BLOCK_DANGER = 1;
+    public static final int NSFW_BLOCK_SMUT = 2;
+
+    @Inject
+    public NsfwDomainFilter(HikariDataSource dataSource,
+                            @Named("nsfw.dangerLists") List<String> dangerLists,
+                            @Named("nsfw.smutLists") List<String> smutLists
+                            ) {
+        this.dataSource = dataSource;
+
+        this.dangerLists = dangerLists;
+        this.smutLists = smutLists;
+
+        Thread.ofPlatform().daemon().name("NsfwDomainFilterSync").start(() -> {
+            while (true) {
+                sync();
+                try {
+                    TimeUnit.HOURS.sleep(1);
+                } catch (InterruptedException e) {
+                    Thread.currentThread().interrupt();
+                    break; // Exit the loop if interrupted
+                }
+            }
+        });
+    }
+
+    public boolean isBlocked(int domainId, int tier) {
+        if (tier == 0)
+            return false;
+
+        if (tier >= 1 && blockedDomainIdsTier1.contains(domainId))
+            return true;
+        if (tier >= 2 && blockedDomainIdsTier2.contains(domainId))
+            return true;
+
+        return false;
+    }
+
+    private synchronized void sync() {
+        try (var conn = dataSource.getConnection();
+             var stmt = conn.prepareStatement("SELECT ID, TIER FROM NSFW_DOMAINS")
+        ) {
+            var rs = stmt.executeQuery();
+            IntOpenHashSet tier1 = new IntOpenHashSet();
+            IntOpenHashSet tier2 = new IntOpenHashSet();
+
+            while (rs.next()) {
+                int domainId = rs.getInt("ID");
+                int tier = rs.getInt("TIER");
+
+                switch (tier) {
+                    case 1 -> tier1.add(domainId);
+                    case 2 -> tier2.add(domainId);
+                }
+            }
+
+            this.blockedDomainIdsTier1 = tier1;
+            this.blockedDomainIdsTier2 = tier2;
+
+            logger.info("NSFW domain filter synced: {} tier 1, {} tier 2", tier1.size(), tier2.size());
+
+        }
+        catch (SQLException ex) {
+            logger.error("Failed to sync NSFW domain filter", ex);
+        }
+    }
+
+    public synchronized void fetchLists() {
+        try (var conn = dataSource.getConnection();
+             HttpClient client = HttpClient.newBuilder()
+                     .followRedirects(HttpClient.Redirect.ALWAYS)
+                     .build();
+             var stmt = conn.createStatement();
+             var insertStmt = conn.prepareStatement("INSERT IGNORE INTO NSFW_DOMAINS_TMP (ID, TIER) SELECT ID, ? FROM EC_DOMAIN WHERE DOMAIN_NAME = ?")) {
+
+            stmt.execute("DROP TABLE IF EXISTS NSFW_DOMAINS_TMP");
+            stmt.execute("CREATE TABLE NSFW_DOMAINS_TMP LIKE NSFW_DOMAINS");
+
+            List<String> combinedDangerList = new ArrayList<>(10_000);
+            for (var dangerListUrl : dangerLists) {
+                combinedDangerList.addAll(fetchList(client, dangerListUrl));
+            }
+
+            for (String domain : combinedDangerList) {
+                insertStmt.setInt(1, NSFW_BLOCK_DANGER);
+                insertStmt.setString(2, domain);
+                insertStmt.execute();
+            }
+
+            List<String> combinedSmutList = new ArrayList<>(10_000);
+            for (var smutListUrl : smutLists) {
+                combinedSmutList.addAll(fetchList(client, smutListUrl));
+            }
+
+            for (String domain : combinedSmutList) {
+                insertStmt.setInt(1, NSFW_BLOCK_SMUT);
+                insertStmt.setString(2, domain);
+                insertStmt.addBatch();
+                insertStmt.execute();
+            }
+
+            stmt.execute("""
+                    DROP TABLE IF EXISTS NSFW_DOMAINS
+                    """);
+            stmt.execute("""
+                    RENAME TABLE NSFW_DOMAINS_TMP TO NSFW_DOMAINS
+                    """);
+            sync();
+        }
+        catch (SQLException ex) {
+            logger.error("Failed to fetch NSFW domain lists", ex);
+        }
+     }
+
+     public List<String> fetchList(HttpClient client, String url) {
+
+        logger.info("Fetching NSFW domain list from {}", url);
+
+        var request = HttpRequest.newBuilder()
+                .uri(java.net.URI.create(url))
+                .build();
+
+        try {
+            if (url.endsWith(".gz")) {
+                var response = client.send(request, HttpResponse.BodyHandlers.ofByteArray());
+
+                byte[] body = response.body();
+
+                try (var reader = new BufferedReader(new InputStreamReader(new GZIPInputStream(new ByteArrayInputStream(body))))) {
+                    return reader.lines()
+                            .filter(StringUtils::isNotEmpty)
+                            .toList();
+                } catch (Exception e) {
+                    logger.error("Error reading GZIP response from {}", url, e);
+                }
+            } else {
+                var response = client.send(request, HttpResponse.BodyHandlers.ofString());
+                if (response.statusCode() == 200) {
+
+                    return Arrays.stream(StringUtils.split(response.body(), "\n"))
+                            .filter(StringUtils::isNotEmpty)
+                            .toList();
+                } else {
+                    logger.warn("Failed to fetch list from {}: HTTP {}", url, response.statusCode());
+                }
+            }
+        }
+        catch (Exception e) {
+            logger.error("Error fetching NSFW domain list from {}", url, e);
+        }
+
+
+        return List.of();
+     }
+}

code/functions/nsfw-domain-filter/java/nu/marginalia/nsfw/NsfwFilterModule.java

@@ -0,0 +1,30 @@
+package nu.marginalia.nsfw;
+
+import com.google.inject.AbstractModule;
+import com.google.inject.Provides;
+import jakarta.inject.Named;
+
+import java.util.List;
+
+public class NsfwFilterModule extends AbstractModule {
+
+    @Provides
+    @Named("nsfw.dangerLists")
+    public List<String> nsfwDomainLists1() {
+        return List.of(
+                "https://raw.githubusercontent.com/olbat/ut1-blacklists/refs/heads/master/blacklists/cryptojacking/domains",
+                "https://raw.githubusercontent.com/olbat/ut1-blacklists/refs/heads/master/blacklists/malware/domains",
+                "https://raw.githubusercontent.com/olbat/ut1-blacklists/refs/heads/master/blacklists/phishing/domains"
+        );
+    }
+    @Provides
+    @Named("nsfw.smutLists")
+    public List<String> nsfwDomainLists2() {
+        return List.of(
+                "https://github.com/olbat/ut1-blacklists/raw/refs/heads/master/blacklists/adult/domains.gz",
+                "https://raw.githubusercontent.com/olbat/ut1-blacklists/refs/heads/master/blacklists/gambling/domains"
+        );
+    }
+
+    public void configure() {}
+}

code/functions/nsfw-domain-filter/test/nu/marginalia/nsfw/NsfwDomainFilterTest.java

@@ -0,0 +1,108 @@
+package nu.marginalia.nsfw;
+
+
+import com.google.inject.AbstractModule;
+import com.google.inject.Guice;
+import com.google.inject.Provides;
+import com.zaxxer.hikari.HikariConfig;
+import com.zaxxer.hikari.HikariDataSource;
+import jakarta.inject.Named;
+import nu.marginalia.test.TestMigrationLoader;
+import org.junit.jupiter.api.BeforeAll;
+import org.junit.jupiter.api.Tag;
+import org.junit.jupiter.api.Test;
+import org.testcontainers.containers.MariaDBContainer;
+import org.testcontainers.junit.jupiter.Container;
+import org.testcontainers.junit.jupiter.Testcontainers;
+
+import java.io.IOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.util.List;
+
+import static org.junit.jupiter.api.Assertions.assertFalse;
+import static org.junit.jupiter.api.Assertions.assertTrue;
+
+
+@Tag("slow")
+@Testcontainers
+class NsfwDomainFilterTest extends AbstractModule {
+
+    @Container
+    static MariaDBContainer<?> mariaDBContainer = new MariaDBContainer<>("mariadb")
+            .withDatabaseName("WMSA_prod")
+            .withUsername("wmsa")
+            .withPassword("wmsa")
+            .withNetworkAliases("mariadb");
+
+    static HikariDataSource dataSource;
+    static Path tempDir;
+
+    @BeforeAll
+    public static void setUpDb() throws IOException {
+        tempDir = Files.createTempDirectory(NsfwDomainFilterTest.class.getSimpleName());
+
+        System.setProperty("system.homePath", tempDir.toString());
+
+        HikariConfig config = new HikariConfig();
+        config.setJdbcUrl(mariaDBContainer.getJdbcUrl());
+        config.setUsername("wmsa");
+        config.setPassword("wmsa");
+
+        dataSource = new HikariDataSource(config);
+
+        TestMigrationLoader.flywayMigration(dataSource);
+
+        try (var conn = dataSource.getConnection();
+             var stmt = conn.prepareStatement("INSERT INTO EC_DOMAIN (DOMAIN_NAME, DOMAIN_TOP, NODE_AFFINITY) VALUES (?, ?, 1)")
+        ) {
+
+            // Ensure the database is ready
+            conn.createStatement().execute("SELECT 1");
+
+            stmt.setString(1, "www.google.com");
+            stmt.setString(2, "google.com");
+            stmt.executeUpdate();
+            stmt.setString(1, "www.bing.com");
+            stmt.setString(2, "bing.com");
+            stmt.executeUpdate();
+        } catch (Exception e) {
+            throw new RuntimeException("Failed to connect to the database", e);
+        }
+    }
+
+    @Provides
+    @Named("nsfw.dangerLists")
+    public List<String> nsfwDomainLists1() {
+        return List.of(
+                "https://downloads.marginalia.nu/test/list1"
+        );
+    }
+
+    @Provides
+    @Named("nsfw.smutLists")
+    public List<String> nsfwDomainLists2() {
+        return List.of(
+                "https://downloads.marginalia.nu/test/list2.gz"
+        );
+    }
+
+    public void configure() {
+        bind(HikariDataSource.class).toInstance(dataSource);
+    }
+
+    @Test
+    public void test() {
+        var filter = Guice
+                .createInjector(this)
+                .getInstance(NsfwDomainFilter.class);
+
+        filter.fetchLists();
+
+        assertTrue(filter.isBlocked(1, NsfwDomainFilter.NSFW_BLOCK_DANGER));
+        assertTrue(filter.isBlocked(1, NsfwDomainFilter.NSFW_BLOCK_SMUT));
+        assertFalse(filter.isBlocked(2, NsfwDomainFilter.NSFW_BLOCK_DANGER));
+        assertTrue(filter.isBlocked(2, NsfwDomainFilter.NSFW_BLOCK_SMUT));
+    }
+
+}

code/functions/search-query/api/java/nu/marginalia/api/searchquery/QueryProtobufCodec.java

@@ -1,9 +1,6 @@
 package nu.marginalia.api.searchquery;
 
-import nu.marginalia.api.searchquery.model.query.ProcessedQuery;
+import nu.marginalia.api.searchquery.model.query.*;
-import nu.marginalia.api.searchquery.model.query.QueryParams;
-import nu.marginalia.api.searchquery.model.query.QueryResponse;
-import nu.marginalia.api.searchquery.model.query.SearchSpecification;
 import nu.marginalia.api.searchquery.model.results.DecoratedSearchResultItem;
 import nu.marginalia.api.searchquery.model.results.PrototypeRankingParameters;
 import nu.marginalia.api.searchquery.model.results.SearchResultItem;
@@ -32,6 +29,8 @@ public class QueryProtobufCodec {
         builder.setSearchSetIdentifier(query.specs.searchSetIdentifier);
         builder.setHumanQuery(request.getHumanQuery());
 
+        builder.setNsfwFilterTierValue(request.getNsfwFilterTierValue());
+
         builder.setQuality(IndexProtobufCodec.convertSpecLimit(query.specs.quality));
         builder.setYear(IndexProtobufCodec.convertSpecLimit(query.specs.year));
         builder.setSize(IndexProtobufCodec.convertSpecLimit(query.specs.size));
@@ -78,6 +77,8 @@ public class QueryProtobufCodec {
         builder.setSearchSetIdentifier(query.specs.searchSetIdentifier);
         builder.setHumanQuery(humanQuery);
 
+        builder.setNsfwFilterTier(RpcIndexQuery.NSFW_FILTER_TIER.DANGER);
+
         builder.setQuality(IndexProtobufCodec.convertSpecLimit(query.specs.quality));
         builder.setYear(IndexProtobufCodec.convertSpecLimit(query.specs.year));
         builder.setSize(IndexProtobufCodec.convertSpecLimit(query.specs.size));
@@ -112,6 +113,7 @@ public class QueryProtobufCodec {
                 request.getSearchSetIdentifier(),
                 QueryStrategy.valueOf(request.getQueryStrategy()),
                 RpcTemporalBias.Bias.valueOf(request.getTemporalBias().getBias().name()),
+                NsfwFilterTier.fromCodedValue(request.getNsfwFilterTierValue()),
                 request.getPagination().getPage()
         );
     }
@@ -327,6 +329,7 @@ public class QueryProtobufCodec {
                 .setRank(IndexProtobufCodec.convertSpecLimit(params.rank()))
                 .setSearchSetIdentifier(params.identifier())
                 .setQueryStrategy(params.queryStrategy().name())
+                .setNsfwFilterTierValue(params.filterTier().getCodedValue())
                 .setTemporalBias(RpcTemporalBias.newBuilder()
                         .setBias(RpcTemporalBias.Bias.valueOf(params.temporalBias().name()))
                         .build())

code/functions/search-query/api/java/nu/marginalia/api/searchquery/model/query/NsfwFilterTier.java

@@ -0,0 +1,26 @@
+package nu.marginalia.api.searchquery.model.query;
+
+public enum NsfwFilterTier {
+    OFF(0),
+    DANGER(1),
+    PORN_AND_GAMBLING(2);
+
+    private final int codedValue; // same as ordinal() for now, but can be changed later if needed
+
+    NsfwFilterTier(int codedValue) {
+        this.codedValue = codedValue;
+    }
+
+    public static NsfwFilterTier fromCodedValue(int codedValue) {
+        for (NsfwFilterTier tier : NsfwFilterTier.values()) {
+            if (tier.codedValue == codedValue) {
+                return tier;
+            }
+        }
+        throw new IllegalArgumentException("Invalid coded value for NsfwFilterTirer: " + codedValue);
+    }
+
+    public int getCodedValue() {
+        return codedValue;
+    }
+}

code/functions/search-query/api/java/nu/marginalia/api/searchquery/model/query/QueryParams.java

@@ -25,10 +25,11 @@ public record QueryParams(
         String identifier,
         QueryStrategy queryStrategy,
         RpcTemporalBias.Bias temporalBias,
+        NsfwFilterTier filterTier,
         int page
 )
 {
-    public QueryParams(String query, RpcQueryLimits limits, String identifier) {
+    public QueryParams(String query, RpcQueryLimits limits, String identifier, NsfwFilterTier filterTier) {
         this(query, null,
                 List.of(),
                 List.of(),
@@ -43,6 +44,7 @@ public record QueryParams(
                 identifier,
                 QueryStrategy.AUTO,
                 RpcTemporalBias.Bias.NONE,
+                filterTier,
                 1 // page
                 );
     }

code/functions/search-query/api/java/nu/marginalia/api/searchquery/model/results/DecoratedSearchResultItem.java

@@ -1,6 +1,7 @@
 package nu.marginalia.api.searchquery.model.results;
 
 import nu.marginalia.api.searchquery.model.results.debug.ResultRankingDetails;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import org.jetbrains.annotations.NotNull;
 
@@ -161,4 +162,14 @@ public class DecoratedSearchResultItem implements Comparable<DecoratedSearchResu
     public String toString() {
         return "DecoratedSearchResultItem(rawIndexResult=" + this.getRawIndexResult() + ", url=" + this.getUrl() + ", title=" + this.getTitle() + ", description=" + this.getDescription() + ", urlQuality=" + this.getUrlQuality() + ", format=" + this.getFormat() + ", features=" + this.getFeatures() + ", pubYear=" + this.getPubYear() + ", dataHash=" + this.getDataHash() + ", wordsTotal=" + this.getWordsTotal() + ", bestPositions=" + this.getBestPositions() + ", rankingScore=" + this.getRankingScore() + ", resultsFromDomain=" + this.getResultsFromDomain() + ", rankingDetails=" + this.getRankingDetails() + ")";
     }
+
+    public String getShortFormat() {
+        try {
+            var df = DocumentFormat.valueOf(format);
+            return df.shortFormat;
+        }
+        catch (IllegalArgumentException e) {
+            return DocumentFormat.UNKNOWN.shortFormat;
+        }
+    }
 }

code/functions/search-query/api/src/main/protobuf/query-api.proto

@@ -32,6 +32,14 @@ message RpcQsQuery {
   RpcTemporalBias temporalBias = 16;
 
   RpcQsQueryPagination pagination = 17;
+
+  NSFW_FILTER_TIER nsfwFilterTier = 18;
+
+  enum NSFW_FILTER_TIER {
+    NONE = 0;
+    DANGER = 1;
+    PORN_AND_GAMBLING = 2;
+  };
 }
 
 /* Query service query response */
@@ -78,8 +86,17 @@ message RpcIndexQuery {
   RpcQueryLimits queryLimits = 10;
   string queryStrategy = 11;      // Named query configuration
   RpcResultRankingParameters parameters = 12;
+
+  NSFW_FILTER_TIER nsfwFilterTier = 13;
+
+  enum NSFW_FILTER_TIER {
+    NONE = 0;
+    DANGER = 1;
+    PORN_AND_GAMBLING = 2;
+  };
 }
 
+
 /* A tagged union encoding some limit on a field */
 message RpcSpecLimit {
   int32 value = 1;

code/functions/search-query/build.gradle

@@ -19,6 +19,7 @@ dependencies {
     implementation project(':code:common:model')
     implementation project(':code:common:service')
 
+    implementation project(':code:functions:nsfw-domain-filter')
     implementation project(':code:functions:search-query:api')
 
     implementation project(':code:index:query')

code/functions/search-query/java/nu/marginalia/functions/searchquery/QueryGRPCService.java

@@ -11,6 +11,7 @@ import nu.marginalia.api.searchquery.model.query.QueryParams;
 import nu.marginalia.api.searchquery.model.results.DecoratedSearchResultItem;
 import nu.marginalia.api.searchquery.model.results.PrototypeRankingParameters;
 import nu.marginalia.index.api.IndexClient;
+import nu.marginalia.nsfw.NsfwDomainFilter;
 import nu.marginalia.service.server.DiscoverableService;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -34,13 +35,16 @@ public class QueryGRPCService
 
 
     private final QueryFactory queryFactory;
+    private final NsfwDomainFilter nsfwDomainFilter;
     private final IndexClient indexClient;
 
     @Inject
     public QueryGRPCService(QueryFactory queryFactory,
+                            NsfwDomainFilter nsfwDomainFilter,
                             IndexClient indexClient)
     {
         this.queryFactory = queryFactory;
+        this.nsfwDomainFilter = nsfwDomainFilter;
         this.indexClient = indexClient;
     }
 

code/functions/search-query/test/nu/marginalia/query/svc/QueryFactoryTest.java

@@ -3,6 +3,7 @@ package nu.marginalia.query.svc;
 import nu.marginalia.WmsaHome;
 import nu.marginalia.api.searchquery.RpcQueryLimits;
 import nu.marginalia.api.searchquery.RpcTemporalBias;
+import nu.marginalia.api.searchquery.model.query.NsfwFilterTier;
 import nu.marginalia.api.searchquery.model.query.QueryParams;
 import nu.marginalia.api.searchquery.model.query.SearchSpecification;
 import nu.marginalia.functions.searchquery.QueryFactory;
@@ -58,6 +59,7 @@ public class QueryFactoryTest {
                         "NONE",
                         QueryStrategy.AUTO,
                         RpcTemporalBias.Bias.NONE,
+                        NsfwFilterTier.OFF,
                         0), null).specs;
     }
 

code/index/api/build.gradle

@@ -17,6 +17,7 @@ dependencies {
     implementation project(':code:common:service')
     implementation project(':code:common:db')
     implementation project(':code:libraries:message-queue')
+    implementation project(':code:functions:nsfw-domain-filter')
     implementation project(':code:functions:search-query:api')
 
     implementation libs.bundles.slf4j

code/index/api/java/nu/marginalia/index/api/IndexClient.java

@@ -2,11 +2,13 @@ package nu.marginalia.index.api;
 
 import com.google.inject.Inject;
 import com.google.inject.Singleton;
+import io.prometheus.client.Counter;
 import nu.marginalia.api.searchquery.IndexApiGrpc;
 import nu.marginalia.api.searchquery.RpcDecoratedResultItem;
 import nu.marginalia.api.searchquery.RpcIndexQuery;
 import nu.marginalia.db.DomainBlacklistImpl;
 import nu.marginalia.model.id.UrlIdCodec;
+import nu.marginalia.nsfw.NsfwDomainFilter;
 import nu.marginalia.service.client.GrpcChannelPoolFactory;
 import nu.marginalia.service.client.GrpcMultiNodeChannelPool;
 import nu.marginalia.service.discovery.property.ServiceKey;
@@ -28,14 +30,26 @@ public class IndexClient {
     private static final Logger logger = LoggerFactory.getLogger(IndexClient.class);
     private final GrpcMultiNodeChannelPool<IndexApiGrpc.IndexApiBlockingStub> channelPool;
     private final DomainBlacklistImpl blacklist;
+    private final NsfwDomainFilter nsfwDomainFilter;
+
+    Counter wmsa_index_query_count = Counter.build()
+            .name("wmsa_nsfw_filter_result_count")
+            .labelNames("tier")
+            .help("Count of results filtered by NSFW tier")
+            .register();
+
     private static final ExecutorService executor = Executors.newCachedThreadPool();
 
     @Inject
-    public IndexClient(GrpcChannelPoolFactory channelPoolFactory, DomainBlacklistImpl blacklist) {
+    public IndexClient(GrpcChannelPoolFactory channelPoolFactory,
+                       DomainBlacklistImpl blacklist,
+                       NsfwDomainFilter nsfwDomainFilter
+                       ) {
         this.channelPool = channelPoolFactory.createMulti(
                 ServiceKey.forGrpcApi(IndexApiGrpc.class, ServicePartition.multi()),
                 IndexApiGrpc::newBlockingStub);
         this.blacklist = blacklist;
+        this.nsfwDomainFilter = nsfwDomainFilter;
     }
 
     private static final Comparator<RpcDecoratedResultItem> comparator =
@@ -52,7 +66,7 @@ public class IndexClient {
     public AggregateQueryResponse executeQueries(RpcIndexQuery indexRequest, Pagination pagination) {
 
         final int requestedMaxResults = indexRequest.getQueryLimits().getResultsTotal();
-
+        int filterTier = indexRequest.getNsfwFilterTierValue();
         AtomicInteger totalNumResults = new AtomicInteger(0);
 
         List<RpcDecoratedResultItem> results =
@@ -74,7 +88,7 @@ public class IndexClient {
                             }
                         })
                         .flatMap(List::stream)
-                        .filter(item -> !isBlacklisted(item))
+                        .filter(item -> !isBlacklisted(item, filterTier))
                         .sorted(comparator)
                         .skip(Math.max(0, (pagination.page - 1) * pagination.pageSize))
                         .limit(pagination.pageSize)
@@ -83,8 +97,23 @@ public class IndexClient {
         return new AggregateQueryResponse(results, pagination.page(), totalNumResults.get());
     }
 
-    private boolean isBlacklisted(RpcDecoratedResultItem item) {
+    static String[] tierNames = {
-        return blacklist.isBlacklisted(UrlIdCodec.getDomainId(item.getRawItem().getCombinedId()));
+            "OFF",
+            "DANGER",
+            "NSFW"
+    };
+
+    private boolean isBlacklisted(RpcDecoratedResultItem item, int filterTier) {
+        int domainId = UrlIdCodec.getDomainId(item.getRawItem().getCombinedId());
+
+        if (blacklist.isBlacklisted(domainId)) {
+            return true;
+        }
+        if (nsfwDomainFilter.isBlocked(domainId, filterTier)) {
+            wmsa_index_query_count.labels(tierNames[filterTier]).inc();
+            return true;
+        }
+        return false;
     }
 
 }

code/index/index-forward/java/nu/marginalia/index/forward/construction/ForwardIndexConverter.java

@@ -84,7 +84,7 @@ public class ForwardIndexConverter {
 
             LongArray docFileData = LongArrayFactory.mmapForWritingConfined(outputFileDocsData, ForwardIndexParameters.ENTRY_SIZE * docsFileId.size());
 
-            ByteBuffer workArea = ByteBuffer.allocate(65536);
+            ByteBuffer workArea = ByteBuffer.allocate(1024*1024*100);
             for (var instance : journal.pages()) {
                 try (var slopTable = new SlopTable(instance.baseDir(), instance.page()))
                 {

code/libraries/blocking-thread-pool/java/nu/marginalia/util/SimpleBlockingThreadPool.java

@@ -23,16 +23,33 @@ public class SimpleBlockingThreadPool {
     private final Logger logger = LoggerFactory.getLogger(SimpleBlockingThreadPool.class);
 
     public SimpleBlockingThreadPool(String name, int poolSize, int queueSize) {
+        this(name, poolSize, queueSize, ThreadType.PLATFORM);
+    }
+
+    public SimpleBlockingThreadPool(String name, int poolSize, int queueSize, ThreadType threadType) {
         tasks = new ArrayBlockingQueue<>(queueSize);
 
         for (int i = 0; i < poolSize; i++) {
-            Thread worker = new Thread(this::worker, name  + "[" + i + "]");
+
-            worker.setDaemon(true);
+            Thread.Builder threadBuilder = switch (threadType) {
-            worker.start();
+                case VIRTUAL -> Thread.ofVirtual();
+                case PLATFORM -> Thread.ofPlatform().daemon(true);
+            };
+
+            Thread worker = threadBuilder
+                    .name(name  + "[" + i + "]")
+                    .start(this::worker);
+
             workers.add(worker);
         }
 
     }
+
+    public enum ThreadType {
+        VIRTUAL,
+        PLATFORM
+    }
+
     public void submit(Task task) throws InterruptedException {
         tasks.put(task);
     }

code/libraries/geo-ip/build.gradle

@@ -15,6 +15,10 @@ dependencies {
 
     implementation libs.bundles.slf4j
     implementation libs.opencsv
+    implementation libs.guava
+    implementation dependencies.create(libs.guice.get()) {
+        exclude group: 'com.google.guava'
+    }
 
     testImplementation libs.bundles.slf4j.test
     testImplementation libs.bundles.junit

code/libraries/geo-ip/java/nu/marginalia/geoip/GeoIpDictionary.java

@@ -1,5 +1,6 @@
 package nu.marginalia.geoip;
 
+import com.google.inject.Singleton;
 import nu.marginalia.WmsaHome;
 import nu.marginalia.geoip.sources.AsnMapping;
 import nu.marginalia.geoip.sources.AsnTable;
@@ -10,6 +11,7 @@ import org.slf4j.LoggerFactory;
 import java.net.InetAddress;
 import java.util.Optional;
 
+@Singleton
 public class GeoIpDictionary {
     private volatile IP2LocationMapping ip2locMapping = null;
     private volatile AsnTable asnTable = null;
@@ -76,7 +78,7 @@ public class GeoIpDictionary {
     }
 
     public Optional<AsnTable.AsnInfo> getAsnInfo(int ipAddress) {
-        if (null == asnTable) { // not loaded yet or failed to load
+        if (null == asnMapping || null == asnTable) { // not loaded yet or failed to load
             return Optional.empty();
         }
 

code/processes/converting-process/build.gradle

@@ -62,6 +62,7 @@ dependencies {
     implementation libs.jwarc
 
     implementation libs.jsoup
+    implementation libs.pdfbox
 
     implementation libs.guava
     implementation dependencies.create(libs.guice.get()) {
@@ -87,6 +88,8 @@ dependencies {
     implementation libs.commons.compress
     implementation libs.sqlite
 
+    implementation libs.bundles.httpcomponents
+
     testImplementation libs.bundles.slf4j.test
     testImplementation libs.bundles.junit
     testImplementation libs.mockito

code/processes/converting-process/java/nu/marginalia/converting/model/ProcessedDocumentDetails.java

@@ -1,8 +1,8 @@
 package nu.marginalia.converting.model;
 
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.HtmlFeature;
-import nu.marginalia.model.html.HtmlStandard;
 import nu.marginalia.model.idx.DocumentMetadata;
 
 import javax.annotation.Nullable;
@@ -21,7 +21,7 @@ public class ProcessedDocumentDetails {
     public long hashCode;
 
     public Set<HtmlFeature> features;
-    public HtmlStandard standard;
+    public DocumentFormat format;
 
     public List<EdgeUrl> linksInternal;
     public List<EdgeUrl> linksExternal;
@@ -30,6 +30,6 @@ public class ProcessedDocumentDetails {
     public GeneratorType generator;
 
     public String toString() {
-        return "ProcessedDocumentDetails(title=" + this.title + ", description=" + this.description + ", pubYear=" + this.pubYear + ", length=" + this.length + ", quality=" + this.quality + ", hashCode=" + this.hashCode + ", features=" + this.features + ", standard=" + this.standard + ", linksInternal=" + this.linksInternal + ", linksExternal=" + this.linksExternal + ", metadata=" + this.metadata + ", generator=" + this.generator + ")";
+        return "ProcessedDocumentDetails(title=" + this.title + ", description=" + this.description + ", pubYear=" + this.pubYear + ", length=" + this.length + ", quality=" + this.quality + ", hashCode=" + this.hashCode + ", features=" + this.features + ", standard=" + this.format + ", linksInternal=" + this.linksInternal + ", linksExternal=" + this.linksExternal + ", metadata=" + this.metadata + ", generator=" + this.generator + ")";
     }
 }

code/processes/converting-process/java/nu/marginalia/converting/processor/DocumentProcessor.java

@@ -7,6 +7,7 @@ import nu.marginalia.converting.model.DisqualifiedException;
 import nu.marginalia.converting.model.ProcessedDocument;
 import nu.marginalia.converting.processor.plugin.AbstractDocumentProcessorPlugin;
 import nu.marginalia.converting.processor.plugin.HtmlDocumentProcessorPlugin;
+import nu.marginalia.converting.processor.plugin.PdfDocumentProcessorPlugin;
 import nu.marginalia.converting.processor.plugin.PlainTextDocumentProcessorPlugin;
 import nu.marginalia.keyword.LinkTexts;
 import nu.marginalia.model.EdgeDomain;
@@ -33,7 +34,8 @@ public class DocumentProcessor {
     private static final Set<String> acceptedContentTypes = Set.of("application/xhtml+xml",
             "application/xhtml",
             "text/html",
-            "text/plain");
+            "text/plain",
+            "application/pdf");
 
 
     private final List<AbstractDocumentProcessorPlugin> processorPlugins = new ArrayList<>();
@@ -42,12 +44,14 @@ public class DocumentProcessor {
     @Inject
     public DocumentProcessor(HtmlDocumentProcessorPlugin htmlDocumentProcessorPlugin,
                              PlainTextDocumentProcessorPlugin plainTextDocumentProcessorPlugin,
+                             PdfDocumentProcessorPlugin pdfDocumentProcessorPlugin,
                              AnchorTextKeywords anchorTextKeywords)
     {
         this.anchorTextKeywords = anchorTextKeywords;
 
         processorPlugins.add(htmlDocumentProcessorPlugin);
         processorPlugins.add(plainTextDocumentProcessorPlugin);
+        processorPlugins.add(pdfDocumentProcessorPlugin);
     }
 
     public ProcessedDocument process(CrawledDocument crawledDocument,

code/processes/converting-process/java/nu/marginalia/converting/processor/logic/DocumentValuator.java

@@ -2,9 +2,9 @@ package nu.marginalia.converting.processor.logic;
 
 import crawlercommons.utils.Strings;
 import nu.marginalia.converting.model.DisqualifiedException;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.crawl.HtmlFeature;
 import nu.marginalia.model.crawldata.CrawledDocument;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jetbrains.annotations.NotNull;
 import org.jsoup.nodes.Document;
 import org.jsoup.nodes.Element;
@@ -17,7 +17,7 @@ import java.util.Set;
 public class DocumentValuator {
 
     public double getQuality(CrawledDocument crawledDocument,
-                             HtmlStandard htmlStandard,
+                             DocumentFormat htmlStandard,
                              Document parsedDocument,
                              int textLength) throws DisqualifiedException {
 

code/processes/converting-process/java/nu/marginalia/converting/processor/logic/HtmlStandardExtractor.java

@@ -1,7 +1,7 @@
 package nu.marginalia.converting.processor.logic;
 
 import com.google.common.base.Strings;
-import nu.marginalia.model.html.HtmlStandard;
+import nu.marginalia.model.DocumentFormat;
 import org.jsoup.nodes.Document;
 import org.jsoup.nodes.DocumentType;
 import org.slf4j.Logger;
@@ -12,54 +12,54 @@ public class HtmlStandardExtractor {
 
     private static final Logger logger = LoggerFactory.getLogger(HtmlStandardExtractor.class);
 
-    public static HtmlStandard parseDocType(DocumentType docType) {
+    public static DocumentFormat parseDocType(DocumentType docType) {
         if (null == docType) {
-            return HtmlStandard.UNKNOWN;
+            return DocumentFormat.UNKNOWN;
         }
 
         String publicId = docType.publicId();
         if (Strings.isNullOrEmpty(publicId))
-            return HtmlStandard.HTML5;
+            return DocumentFormat.HTML5;
 
         publicId = publicId.toUpperCase();
         if (publicId.startsWith("-//SOFTQUAD SOFTWARE//DTD") && publicId.contains("HTML 4")) {
-            return HtmlStandard.HTML4;
+            return DocumentFormat.HTML4;
         }
         if (publicId.startsWith("-//SOFTQUAD SOFTWARE//DTD") && publicId.contains("HTML 3")) {
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         }
         if (publicId.startsWith("-//INTERNET/RFC XXXX//EN"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//NETSCAPE COMM. CORP"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//SQ//DTD HTML 2"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//SOFTQUAD//DTD HTML 2"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//W3O//DTD W3 HTML 2"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//IETF//DTD HTML 2"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//IETF//DTD HTML//EN"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-/W3C//DTD HTML 3"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-/W3C/DTD HTML 3"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//IETF//DTD HTML 3"))
-            return HtmlStandard.HTML123;
+            return DocumentFormat.HTML123;
         if (publicId.startsWith("-//W3C//DTD XHTML"))
-            return HtmlStandard.XHTML;
+            return DocumentFormat.XHTML;
         if (publicId.startsWith("ISO/IEC 15445:2000//DTD"))
-            return HtmlStandard.XHTML;
+            return DocumentFormat.XHTML;
         if (publicId.startsWith("-//W3C//DTD HTML"))
-            return HtmlStandard.HTML4;
+            return DocumentFormat.HTML4;
 
         logger.debug("Unknown publicID standard {}", publicId);
-        return HtmlStandard.UNKNOWN;
+        return DocumentFormat.UNKNOWN;
     }
 
-    public static HtmlStandard sniffHtmlStandard(Document parsed) {
+    public static DocumentFormat sniffHtmlStandard(Document parsed) {
         int html4Attributes = 0;
         int html5Attributes = 0;
 
@@ -73,11 +73,11 @@ public class HtmlStandardExtractor {
             html4Attributes++;
         }
         if (html5Attributes > 0) {
-            return HtmlStandard.HTML5;
+            return DocumentFormat.HTML5;
         }
         if (html4Attributes > 0) {
-            return HtmlStandard.HTML4;
+            return DocumentFormat.HTML4;
         }
-        return HtmlStandard.HTML123;
+        return DocumentFormat.HTML123;
     }
 }

code/processes/converting-process/java/nu/marginalia/converting/processor/plugin/AbstractDocumentProcessorPlugin.java

@@ -7,11 +7,11 @@ import nu.marginalia.keyword.LinkTexts;
 import nu.marginalia.keyword.model.DocumentKeywordsBuilder;
 import nu.marginalia.language.filter.LanguageFilter;
 import nu.marginalia.language.model.DocumentLanguageData;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.HtmlFeature;
 import nu.marginalia.model.crawl.PubDate;
 import nu.marginalia.model.crawldata.CrawledDocument;
-import nu.marginalia.model.html.HtmlStandard;
 
 import javax.annotation.Nullable;
 import java.io.IOException;
@@ -73,7 +73,7 @@ public abstract class AbstractDocumentProcessorPlugin {
             return this;
         }
 
-        public MetaTagsBuilder addFormat(HtmlStandard standard) {
+        public MetaTagsBuilder addFormat(DocumentFormat standard) {
 
             add("format", standard);
 

code/processes/converting-process/java/nu/marginalia/converting/processor/plugin/HtmlDocumentProcessorPlugin.java

@@ -25,12 +25,12 @@ import nu.marginalia.language.model.DocumentLanguageData;
 import nu.marginalia.language.sentence.ThreadLocalSentenceExtractorProvider;
 import nu.marginalia.link_parser.FeedExtractor;
 import nu.marginalia.link_parser.LinkParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.HtmlFeature;
 import nu.marginalia.model.crawl.PubDate;
 import nu.marginalia.model.crawldata.CrawledDocument;
-import nu.marginalia.model.html.HtmlStandard;
 import nu.marginalia.model.idx.DocumentFlags;
 import nu.marginalia.model.idx.DocumentMetadata;
 import org.jsoup.nodes.Document;
@@ -137,8 +137,8 @@ public class HtmlDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin
 
 
         final int length = getLength(doc);
-        final HtmlStandard standard = getHtmlStandard(doc);
+        final DocumentFormat format = getDocumentFormat(doc);
-        final double quality = documentValuator.getQuality(crawledDocument, standard, doc, length);
+        final double quality = documentValuator.getQuality(crawledDocument, format, doc, length);
 
         if (isDisqualified(documentClass, url, quality, doc.title())) {
             throw new DisqualifiedException(DisqualificationReason.QUALITY);
@@ -152,7 +152,7 @@ public class HtmlDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin
         var ret = new ProcessedDocumentDetails();
 
         ret.length = length;
-        ret.standard = standard;
+        ret.format = format;
         ret.title = specialization.getTitle(doc, dld, crawledDocument.url);
 
         final Set<HtmlFeature> features = featureExtractor.getFeatures(url, doc, documentHeaders, dld);
@@ -161,7 +161,7 @@ public class HtmlDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin
         ret.quality = documentValuator.adjustQuality(quality, features);
         ret.hashCode = dld.localitySensitiveHashCode();
 
-        PubDate pubDate = pubDateSniffer.getPubDate(documentHeaders, url, doc, standard, true);
+        PubDate pubDate = pubDateSniffer.getPubDate(documentHeaders, url, doc, format, true);
 
         EnumSet<DocumentFlags> documentFlags = documentFlags(features, generatorParts.type());
 
@@ -180,7 +180,7 @@ public class HtmlDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin
                 .addPubDate(pubDate)
                 .addUrl(url)
                 .addFeatures(features)
-                .addFormat(standard)
+                .addFormat(format)
                 .addGenerator(generatorParts.keywords())
                 .build();
 
@@ -316,12 +316,12 @@ public class HtmlDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin
         return linkTerms;
     }
 
-    private HtmlStandard getHtmlStandard(Document doc) {
+    private DocumentFormat getDocumentFormat(Document doc) {
-        HtmlStandard htmlStandard = HtmlStandardExtractor.parseDocType(doc.documentType());
+        DocumentFormat format = HtmlStandardExtractor.parseDocType(doc.documentType());
-        if (HtmlStandard.UNKNOWN.equals(htmlStandard)) {
+        if (DocumentFormat.UNKNOWN.equals(format)) {
             return HtmlStandardExtractor.sniffHtmlStandard(doc);
         }
-        return htmlStandard;
+        return format;
     }
 
     private int getLength(Document doc) {

code/processes/converting-process/java/nu/marginalia/converting/processor/plugin/PdfDocumentProcessorPlugin.java

@@ -0,0 +1,286 @@
+package nu.marginalia.converting.processor.plugin;
+
+import com.google.inject.Inject;
+import com.google.inject.name.Named;
+import nu.marginalia.converting.model.DisqualifiedException;
+import nu.marginalia.converting.model.ProcessedDocumentDetails;
+import nu.marginalia.converting.processor.DocumentClass;
+import nu.marginalia.converting.processor.logic.DocumentLengthLogic;
+import nu.marginalia.converting.processor.plugin.specialization.DefaultSpecialization;
+import nu.marginalia.keyword.DocumentKeywordExtractor;
+import nu.marginalia.keyword.LinkTexts;
+import nu.marginalia.keyword.model.DocumentKeywordsBuilder;
+import nu.marginalia.language.filter.LanguageFilter;
+import nu.marginalia.language.model.DocumentLanguageData;
+import nu.marginalia.language.sentence.ThreadLocalSentenceExtractorProvider;
+import nu.marginalia.model.DocumentFormat;
+import nu.marginalia.model.EdgeUrl;
+import nu.marginalia.model.crawl.HtmlFeature;
+import nu.marginalia.model.crawl.PubDate;
+import nu.marginalia.model.crawldata.CrawledDocument;
+import nu.marginalia.model.idx.DocumentFlags;
+import nu.marginalia.model.idx.DocumentMetadata;
+import org.apache.commons.lang3.StringUtils;
+import org.apache.pdfbox.Loader;
+import org.apache.pdfbox.text.HeadingAwarePDFTextStripper;
+import org.jsoup.Jsoup;
+import org.jsoup.nodes.Document;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+import java.io.IOException;
+import java.net.URISyntaxException;
+import java.time.LocalDate;
+import java.util.*;
+
+
+public class PdfDocumentProcessorPlugin extends AbstractDocumentProcessorPlugin {
+
+    private final int maxTitleLength;
+    private final DocumentKeywordExtractor keywordExtractor;
+    private final ThreadLocalSentenceExtractorProvider sentenceExtractorProvider;
+    private final DocumentLengthLogic documentLengthLogic;
+    private final DefaultSpecialization defaultSpecialization;
+
+    private static final Logger logger = LoggerFactory.getLogger(PdfDocumentProcessorPlugin.class);
+
+    @Inject
+    public PdfDocumentProcessorPlugin(@Named("max-title-length") Integer maxTitleLength,
+                                      LanguageFilter languageFilter,
+                                      ThreadLocalSentenceExtractorProvider sentenceExtractorProvider,
+                                      DocumentKeywordExtractor keywordExtractor,
+                                      DocumentLengthLogic documentLengthLogic,
+                                      DefaultSpecialization defaultSpecialization)
+
+    {
+        super(languageFilter);
+        this.sentenceExtractorProvider = sentenceExtractorProvider;
+        this.documentLengthLogic = documentLengthLogic;
+        this.maxTitleLength = maxTitleLength;
+        this.keywordExtractor = keywordExtractor;
+        this.defaultSpecialization = defaultSpecialization;
+    }
+
+    @Override
+    public boolean isApplicable(CrawledDocument doc) {
+        String contentType = doc.contentType.toLowerCase();
+
+        if (contentType.equals("application/pdf"))
+            return true;
+        if (contentType.startsWith("application/pdf;")) // charset=blabla
+            return true;
+
+        return false;
+    }
+
+    @Override
+    public DetailsWithWords createDetails(CrawledDocument crawledDocument,
+                                          LinkTexts linkTexts,
+                                          DocumentClass documentClass)
+            throws DisqualifiedException, URISyntaxException, IOException {
+
+        String documentBody = crawledDocument.documentBody();
+
+        if (languageFilter.isBlockedUnicodeRange(documentBody)) {
+            throw new DisqualifiedException(DisqualifiedException.DisqualificationReason.LANGUAGE);
+        }
+
+        final EdgeUrl url = new EdgeUrl(crawledDocument.url);
+
+
+        Document doc;
+        try {
+            doc = convertPdfToHtml(crawledDocument.documentBodyBytes);
+        } catch (IOException e) {
+            logger.error("Failed to convert PDF file {} - {}", url, e.getMessage());
+            throw new DisqualifiedException(DisqualifiedException.DisqualificationReason.ERROR);
+        }
+
+        DocumentLanguageData dld = sentenceExtractorProvider.get().extractSentences(doc);
+
+        checkDocumentLanguage(dld);
+
+        documentLengthLogic.validateLength(dld, 1.0);
+
+        var ret = new ProcessedDocumentDetails();
+
+        ret.length = documentBody.length();
+
+        ret.format = DocumentFormat.PDF;
+        ret.title = StringUtils.truncate(defaultSpecialization.getTitle(doc, dld, url.toString()), maxTitleLength);
+
+        ret.quality = -5;
+
+        ret.features = Set.of(HtmlFeature.PDF);
+        ret.description = getDescription(doc);
+        ret.hashCode = dld.localitySensitiveHashCode();
+
+        final PubDate pubDate = new PubDate(LocalDate.ofYearDay(1993, 1));
+
+        EnumSet<DocumentFlags> documentFlags = EnumSet.of(DocumentFlags.PdfFile);
+
+        ret.metadata = new DocumentMetadata(
+                documentLengthLogic.getEncodedAverageLength(dld),
+                pubDate.yearByte(),
+                (int) -ret.quality,
+                documentFlags);
+
+        DocumentKeywordsBuilder words = keywordExtractor.extractKeywords(dld, linkTexts, url);
+
+        var tagWords = new MetaTagsBuilder()
+                .addPubDate(pubDate)
+                .addUrl(url)
+                .addFeatures(ret.features)
+                .addFormat(ret.format)
+                .build();
+
+        words.addAllSyntheticTerms(tagWords);
+
+        if (pubDate.hasYear()) {
+            ret.pubYear = pubDate.year();
+        }
+
+        /* These are assumed to be populated */
+        ret.linksInternal = new ArrayList<>();
+        ret.linksExternal = new ArrayList<>();
+
+        return new DetailsWithWords(ret, words);
+    }
+
+    private String getDescription(Document doc) {
+        int cnt = 0;
+        boolean useNext = false;
+        for (var ptag : doc.getElementsByTag("p")) {
+            String text = ptag.text();
+
+            // Many academic documents have an abstract at the start of the document,
+            // which makes a nice summary.  Though they tend to bleed into the text,
+            // so we check for the word "Abstract" at the start of the paragraph.
+
+            if (text.startsWith("Abstract ")) {
+                return StringUtils.abbreviate(text.substring("Abstract ".length()), "...", 255);
+            }
+            else if (text.equals("Abstract")) {
+                useNext = true;
+            }
+            else if (useNext) {
+                return StringUtils.abbreviate(text, "...", 255);
+            }
+
+            if (++cnt > 15) { // Don't scan the entire document
+                break;
+            }
+        }
+
+        // Fall back to the default specialization
+        return defaultSpecialization.getSummary(doc, Set.of());
+
+    }
+
+    /** Convert the provided PDF bytes into a HTML rendering that can be fed
+     * to the HTML processor.
+     */
+    Document convertPdfToHtml(byte[] pdfBytes) throws IOException {
+        try (var doc = Loader.loadPDF(pdfBytes)) {
+            String docMetaTitle = Objects.requireNonNullElse(doc.getDocumentInformation().getTitle(), "");
+
+            var stripper = new HeadingAwarePDFTextStripper();
+            stripper.setStartPage(1);
+            stripper.setSortByPosition(true);
+            stripper.setWordSeparator(" ");
+
+            // Increase the tolerance for line spacing to deal better with paragraphs.
+            stripper.setDropThreshold(5f);
+
+            stripper.setPageStart("<div>");
+            stripper.setParagraphStart("<p>");
+            stripper.setParagraphEnd("</p>\n");
+            stripper.setPageEnd("</div>\n");
+            stripper.setHeadingStart("<h1>");
+            stripper.setHeadingEnd("</h1>\n");
+            stripper.setLineSeparator("\n");
+
+            String text = stripper.getText(doc);
+
+            StringBuilder htmlBuilder = new StringBuilder(text.length() + 1024);
+            htmlBuilder.append("<html><body>")
+                    .append(text)
+                    .append("</body></html>");
+
+            var parsed = Jsoup.parse(htmlBuilder.toString());
+
+            repairDOM(parsed);
+
+            for (var heading : parsed.getElementsByTag("h1")) {
+                String headingText = heading.text();
+                if (headingText.length() > 2) {
+                    parsed.title(headingText);
+                    break;
+                }
+            }
+
+
+            if (parsed.title().isEmpty()) {
+                // Prefer setting the title to the first paragraph in the
+                // document, as this is almost always correct.  Otherwise,
+                // we fall back on the metadata title, which is almost always
+                // useless
+
+                var firstP = parsed.getElementsByTag("p").first();
+                if (firstP != null) parsed.title(firstP.text());
+                else parsed.title(docMetaTitle);
+            }
+            return parsed;
+        }
+
+
+    }
+
+    /** Repair the DOM to remove some common issues with PDF conversion,
+     * including empty paragraphs, and multiline headers that are split into multiple
+     * conescutive h1 tags.
+     */
+    private void repairDOM(Document parsed) {
+
+        // <p><h1>...</h1></p> -> <h1>...</h1>
+        parsed.getElementsByTag("h1").forEach(h1 -> {
+            var parent = h1.parent();
+            if (parent == null || !"p".equals(parent.tagName())) {
+                return;
+            }
+
+            if (parent.childrenSize() == 1) {
+                parent.replaceWith(h1);
+            }
+        });
+
+        // Remove empty <p> tags
+        parsed.getElementsByTag("p").forEach(p -> {
+            if (p.childrenSize() == 0 && !p.hasText()) {
+                p.remove();
+            }
+        });
+
+        // <h1>...</h1><h1>...</h1> -> <h1>...</h1>
+        parsed.getElementsByTag("h1").forEach(h1 -> {
+            var nextSibling = h1.nextElementSibling();
+            if (nextSibling == null || !"h1".equals(nextSibling.tagName())) {
+                return; // Short-circuit to avoid unnecessary work
+            }
+
+            StringJoiner joiner = new StringJoiner(" ");
+            joiner.add(h1.text());
+
+            for (var sibling : h1.nextElementSiblings()) {
+                if (!"h1".equals(sibling.tagName()))
+                    break;
+                joiner.add(sibling.text());
+                sibling.remove();
+            }
+
+            h1.text(joiner.toString());
+        });
+
+    }
+
+}

code/processes/converting-process/java/nu/marginalia/converting/processor/plugin/PlainTextDocumentProcessorPlugin.java

@@ -13,10 +13,10 @@ import nu.marginalia.keyword.LinkTexts;
 import nu.marginalia.keyword.model.DocumentKeywordsBuilder;
 import nu.marginalia.language.filter.LanguageFilter;
 import nu.marginalia.language.sentence.ThreadLocalSentenceExtractorProvider;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
 import nu.marginalia.model.crawldata.CrawledDocument;
-import nu.marginalia.model.html.HtmlStandard;
 import nu.marginalia.model.idx.DocumentFlags;
 import nu.marginalia.model.idx.DocumentMetadata;
 import org.apache.commons.lang3.StringUtils;
@@ -91,7 +91,7 @@ public class PlainTextDocumentProcessorPlugin extends AbstractDocumentProcessorP
 
         ret.length = documentBody.length();
 
-        ret.standard = HtmlStandard.PLAIN;
+        ret.format = DocumentFormat.PLAIN;
         ret.title = StringUtils.truncate(plainTextLogic.getTitle(url, firstFewLines), maxTitleLength);
 
         ret.quality = -1;
@@ -113,7 +113,7 @@ public class PlainTextDocumentProcessorPlugin extends AbstractDocumentProcessorP
                 .addPubDate(pubDate)
                 .addUrl(url)
                 .addFeatures(ret.features)
-                .addFormat(ret.standard)
+                .addFormat(ret.format)
                 .build();
 
         words.addAllSyntheticTerms(tagWords);

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/PubDateFromHtmlStandard.java

@@ -1,12 +1,13 @@
 package nu.marginalia.converting.processor.pubdate;
 
-import nu.marginalia.model.html.HtmlStandard;
+import nu.marginalia.model.DocumentFormat;
 
 public class PubDateFromHtmlStandard {
     /** Used to bias pub date heuristics */
-    public static int blindGuess(HtmlStandard standard) {
+    public static int blindGuess(DocumentFormat format) {
-        return switch (standard) {
+        return switch (format) {
             case PLAIN -> 1993;
+            case PDF -> 2010;
             case HTML123 -> 1997;
             case HTML4, XHTML -> 2006;
             case HTML5 -> 2018;
@@ -21,8 +22,8 @@ public class PubDateFromHtmlStandard {
      * Discovering publication year involves a lot of guesswork, this helps
      * keep the guesses relatively sane.
      */
-    public static boolean isGuessPlausible(HtmlStandard standard, int year) {
+    public static boolean isGuessPlausible(DocumentFormat format, int year) {
-        switch (standard) {
+        switch (format) {
             case HTML123:
                 return year <= 2000;
             case XHTML:

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/PubDateHeuristic.java

@@ -1,14 +1,14 @@
 package nu.marginalia.converting.processor.pubdate;
 
 import nu.marginalia.converting.model.DocumentHeaders;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
 
 public interface PubDateHeuristic {
 
-    Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard);
+    Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard);
 }

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/PubDateParser.java

@@ -1,7 +1,7 @@
 package nu.marginalia.converting.processor.pubdate;
 
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 
 import java.time.DateTimeException;
 import java.time.LocalDate;
@@ -26,7 +26,7 @@ public class PubDateParser {
                 .filter(PubDateParser::validateDate);
     }
 
-    public static Optional<PubDate> attemptParseDate(String date, HtmlStandard standard) {
+    public static Optional<PubDate> attemptParseDate(String date, DocumentFormat standard) {
         return Optional.ofNullable(date)
                 .filter(str -> str.length() >= 4 && str.length() < 32)
                 .flatMap(str ->
@@ -81,7 +81,7 @@ public class PubDateParser {
     }
 
 
-    public static Optional<PubDate> dateFromHighestYearLookingSubstringWithGuess(String maybe, HtmlStandard standard) {
+    public static Optional<PubDate> dateFromHighestYearLookingSubstringWithGuess(String maybe, DocumentFormat standard) {
         int guess = PubDateFromHtmlStandard.blindGuess(standard);
 
         var matcher = yearPattern.matcher(maybe);
@@ -135,7 +135,7 @@ public class PubDateParser {
         return (max + min) / 2;
     }
 
-    public static int guessYear(HtmlStandard standard) {
+    public static int guessYear(DocumentFormat standard) {
         // Create some jitter to avoid having documents piling up in the same four years
         // as this would make searching in those years disproportionately useless
 

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/PubDateSniffer.java

@@ -2,9 +2,9 @@ package nu.marginalia.converting.processor.pubdate;
 
 import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.heuristic.*;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.ArrayList;
@@ -38,7 +38,7 @@ public class PubDateSniffer {
         heuristics.add(new PubDateHeuristicGuessFromHtmlStandard());
     }
 
-    public PubDate getPubDate(DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard, boolean runExpensive) {
+    public PubDate getPubDate(DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard, boolean runExpensive) {
         final PubDateEffortLevel effortLevel = runExpensive ? PubDateEffortLevel.HIGH : PubDateEffortLevel.LOW;
 
         for (var heuristic : heuristics) {

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicDOMParsingPass1.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jetbrains.annotations.NotNull;
 import org.jsoup.nodes.Document;
 import org.jsoup.nodes.Element;
@@ -19,7 +19,7 @@ import java.util.Optional;
 public class PubDateHeuristicDOMParsingPass1 implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         if (effortLevel == PubDateEffortLevel.LOW)
             return Optional.empty();
 
@@ -33,9 +33,9 @@ public class PubDateHeuristicDOMParsingPass1 implements PubDateHeuristic {
 
     private static class DateExtractingNodeVisitorPass implements NodeFilter {
         public PubDate pubDate;
-        private final HtmlStandard htmlStandard;
+        private final DocumentFormat htmlStandard;
 
-        private DateExtractingNodeVisitorPass(HtmlStandard htmlStandard) {
+        private DateExtractingNodeVisitorPass(DocumentFormat htmlStandard) {
             this.htmlStandard = htmlStandard;
         }
 
@@ -135,7 +135,7 @@ public class PubDateHeuristicDOMParsingPass1 implements PubDateHeuristic {
         }
 
         private void parse(String text) {
-            if (htmlStandard == HtmlStandard.UNKNOWN) {
+            if (htmlStandard == DocumentFormat.UNKNOWN) {
                 PubDateParser
                         .dateFromHighestYearLookingSubstring(text)
                         .ifPresent(this::setPubDate);

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicDOMParsingPass2.java

@@ -5,9 +5,9 @@ import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateFromHtmlStandard;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jetbrains.annotations.NotNull;
 import org.jsoup.nodes.Document;
 import org.jsoup.nodes.Node;
@@ -19,7 +19,7 @@ import java.util.Optional;
 public class PubDateHeuristicDOMParsingPass2 implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         if (effortLevel == PubDateEffortLevel.LOW)
             return Optional.empty();
 
@@ -33,9 +33,9 @@ public class PubDateHeuristicDOMParsingPass2 implements PubDateHeuristic {
 
     private static class DateExtractingNodeVisitor implements NodeFilter {
         public PubDate pubDate;
-        private final HtmlStandard htmlStandard;
+        private final DocumentFormat htmlStandard;
 
-        private DateExtractingNodeVisitor(HtmlStandard htmlStandard) {
+        private DateExtractingNodeVisitor(DocumentFormat htmlStandard) {
             this.htmlStandard = htmlStandard;
         }
 
@@ -73,7 +73,7 @@ public class PubDateHeuristicDOMParsingPass2 implements PubDateHeuristic {
         }
 
         private void parse(String text) {
-            if (htmlStandard == HtmlStandard.UNKNOWN) {
+            if (htmlStandard == DocumentFormat.UNKNOWN) {
                 PubDateParser
                         .dateFromHighestYearLookingSubstring(text)
                         .ifPresent(this::setPubDate);

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicGuessFromHtmlStandard.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,8 +14,8 @@ import java.util.Optional;
 public class PubDateHeuristicGuessFromHtmlStandard implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
-        if (htmlStandard == HtmlStandard.UNKNOWN)
+        if (htmlStandard == DocumentFormat.UNKNOWN)
             return Optional.empty();
 
         return Optional.of(new PubDate(null, PubDateParser.guessYear(htmlStandard)));

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicHtml5AnyTimeTag.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicHtml5AnyTimeTag implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         // HTML5, alternative approach
         for (var tag : document.select("time")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("datetime"));

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicHtml5ArticleDateTag.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicHtml5ArticleDateTag implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         // HTML5
         for (var tag : document.select("time[pubdate=\"pubdate\"]")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("datetime"));

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicHtml5ItempropDateTag.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicHtml5ItempropDateTag implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         for (var tag : document.select("time[itemprop=\"datePublished\"]")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("content"));
             if (maybeDate.isPresent()) {

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicJSONLD.java

@@ -8,9 +8,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Collections;
@@ -21,7 +21,7 @@ import java.util.Optional;
 public class PubDateHeuristicJSONLD implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         for (var tag : document.select("script[type=\"application/ld+json\"]")) {
             var maybeDate = parseLdJson(tag.data())
                     .flatMap(PubDateParser::attemptParseDate);

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicLastModified.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.List;
@@ -15,7 +15,7 @@ import java.util.Optional;
 public class PubDateHeuristicLastModified implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         List<String> lastModified = headers.get("last-modified");
         if (lastModified.isEmpty())
             return Optional.empty();

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicMicrodata.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicMicrodata implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
 
         for (var tag : document.select("meta[itemprop=\"datePublished\"]")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("content"));

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicOpenGraph.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicOpenGraph implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         // OG
         for (var tag : document.select("meta[property=\"article:published_time\"]")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("content"));

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicRDFaTag.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -14,7 +14,7 @@ import java.util.Optional;
 public class PubDateHeuristicRDFaTag implements PubDateHeuristic {
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         for (var tag : document.select("meta[property=\"datePublished\"]")) {
             var maybeDate = PubDateParser.attemptParseDate(tag.attr("content"));
             if (maybeDate.isPresent()) {

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicUrlPatternPass1.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -21,7 +21,7 @@ public class PubDateHeuristicUrlPatternPass1 implements PubDateHeuristic {
     private static final int MIN_URL_PATTERN_YEAR = 2000;
 
     @Override
-    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, HtmlStandard htmlStandard) {
+    public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url, Document document, DocumentFormat htmlStandard) {
         final String urlString = url.path;
 
         var matcher = yearUrlPattern.matcher(urlString);

code/processes/converting-process/java/nu/marginalia/converting/processor/pubdate/heuristic/PubDateHeuristicUrlPatternPass2.java

@@ -4,9 +4,9 @@ import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.PubDateEffortLevel;
 import nu.marginalia.converting.processor.pubdate.PubDateHeuristic;
 import nu.marginalia.converting.processor.pubdate.PubDateParser;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.PubDate;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.nodes.Document;
 
 import java.util.Optional;
@@ -19,7 +19,7 @@ public class PubDateHeuristicUrlPatternPass2 implements PubDateHeuristic {
 
     @Override
     public Optional<PubDate> apply(PubDateEffortLevel effortLevel, DocumentHeaders headers, EdgeUrl url,
-                                   Document document, HtmlStandard htmlStandard) {
+                                   Document document, DocumentFormat htmlStandard) {
         final String urlString = url.path;
 
         var matcher = yearUrlPattern.matcher(urlString);

code/processes/converting-process/java/nu/marginalia/converting/sideload/SideloaderProcessing.java

@@ -8,12 +8,12 @@ import nu.marginalia.converting.model.ProcessedDocument;
 import nu.marginalia.converting.processor.DocumentClass;
 import nu.marginalia.converting.processor.plugin.HtmlDocumentProcessorPlugin;
 import nu.marginalia.keyword.LinkTexts;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.HtmlFeature;
 import nu.marginalia.model.crawl.PubDate;
 import nu.marginalia.model.crawl.UrlIndexingState;
 import nu.marginalia.model.crawldata.CrawledDocument;
-import nu.marginalia.model.html.HtmlStandard;
 import nu.marginalia.model.idx.DocumentFlags;
 import nu.marginalia.model.idx.DocumentMetadata;
 import nu.marginalia.model.idx.WordFlags;
@@ -53,6 +53,7 @@ public class SideloaderProcessing {
                 "",
                 body.getBytes(StandardCharsets.UTF_8),
                 false,
+                -1,
                 null,
                 null
         );
@@ -83,7 +84,7 @@ public class SideloaderProcessing {
             // that we can't get from the sideloaded data since it's
             // so stripped down
 
-            ret.details.standard = HtmlStandard.HTML5;
+            ret.details.format = DocumentFormat.HTML5;
             ret.details.pubYear = pubYear;
             ret.details.features.add(HtmlFeature.JS);
             ret.details.features.add(HtmlFeature.TRACKING);

code/processes/converting-process/java/nu/marginalia/converting/sideload/stackexchange/StackexchangeSideloader.java

@@ -9,13 +9,13 @@ import nu.marginalia.integration.stackexchange.sqlite.StackExchangePostsDb;
 import nu.marginalia.keyword.DocumentKeywordExtractor;
 import nu.marginalia.keyword.LinkTexts;
 import nu.marginalia.language.sentence.ThreadLocalSentenceExtractorProvider;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.EdgeUrl;
 import nu.marginalia.model.crawl.DomainIndexingState;
 import nu.marginalia.model.crawl.HtmlFeature;
 import nu.marginalia.model.crawl.PubDate;
 import nu.marginalia.model.crawl.UrlIndexingState;
-import nu.marginalia.model.html.HtmlStandard;
 import nu.marginalia.model.idx.DocumentFlags;
 import nu.marginalia.model.idx.DocumentMetadata;
 import nu.marginalia.model.idx.WordFlags;
@@ -165,7 +165,7 @@ public class StackexchangeSideloader implements SideloadSource {
             ret.details.description = StringUtils.truncate(doc.body().text(), 255);
             ret.details.length = 128;
 
-            ret.details.standard = HtmlStandard.HTML5;
+            ret.details.format = DocumentFormat.HTML5;
             ret.details.linksExternal = List.of();
             ret.details.linksInternal = List.of();
             ret.state = UrlIndexingState.OK;

code/processes/converting-process/java/nu/marginalia/converting/writer/ConverterBatchWriter.java

@@ -124,7 +124,7 @@ public class ConverterBatchWriter implements AutoCloseable, ConverterBatchWriter
                     document.details.title,
                     document.details.description,
                     HtmlFeature.encode(document.details.features),
-                    document.details.standard.name(),
+                    document.details.format.name(),
                     document.details.length,
                     document.details.hashCode,
                     (float) document.details.quality,

code/processes/converting-process/test/nu/marginalia/converting/ConvertingIntegrationTest.java

@@ -6,6 +6,7 @@ import com.google.inject.Injector;
 import nu.marginalia.converting.model.ProcessedDocument;
 import nu.marginalia.converting.processor.DomainProcessor;
 import nu.marginalia.io.SerializableCrawlDataStream;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeDomain;
 import nu.marginalia.model.crawl.DomainIndexingState;
 import nu.marginalia.model.crawl.PubDate;
@@ -13,7 +14,6 @@ import nu.marginalia.model.crawl.UrlIndexingState;
 import nu.marginalia.model.crawldata.CrawledDocument;
 import nu.marginalia.model.crawldata.CrawledDomain;
 import nu.marginalia.model.crawldata.SerializableCrawlData;
-import nu.marginalia.model.html.HtmlStandard;
 import org.junit.jupiter.api.BeforeEach;
 import org.junit.jupiter.api.Tag;
 import org.junit.jupiter.api.Test;
@@ -91,7 +91,7 @@ public class ConvertingIntegrationTest {
 
             assertTrue(details.title.length() > 4);
             assertTrue(details.description.length() > 4);
-            assertEquals(HtmlStandard.HTML5, details.standard);
+            assertEquals(DocumentFormat.HTML5, details.format);
 
         }
     }
@@ -125,7 +125,7 @@ public class ConvertingIntegrationTest {
             assertTrue(details.metadata.size() > 0);
             assertTrue(details.title.length() > 4);
             assertTrue(details.description.length() > 4);
-            assertEquals(HtmlStandard.HTML5, details.standard);
+            assertEquals(DocumentFormat.HTML5, details.format);
         }
     }
 
@@ -148,6 +148,7 @@ public class ConvertingIntegrationTest {
                     "",
                     readClassPathFile(p.toString()).getBytes(),
                     false,
+                    -1,
                     null,
                     null
                     );

code/processes/converting-process/test/nu/marginalia/converting/CrawlingThenConvertingIntegrationTest.java

@@ -8,7 +8,6 @@ import nu.marginalia.converting.model.ProcessedDomain;
 import nu.marginalia.converting.processor.DomainProcessor;
 import nu.marginalia.crawl.CrawlerMain;
 import nu.marginalia.crawl.DomainStateDb;
-import nu.marginalia.crawl.fetcher.Cookies;
 import nu.marginalia.crawl.fetcher.HttpFetcher;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 import nu.marginalia.crawl.fetcher.warc.WarcRecorder;
@@ -247,7 +246,7 @@ public class CrawlingThenConvertingIntegrationTest {
     private CrawledDomain crawl(CrawlerMain.CrawlSpecRecord specs, Predicate<EdgeDomain> domainBlacklist) throws Exception {
         List<SerializableCrawlData> data = new ArrayList<>();
 
-        try (var recorder = new WarcRecorder(fileName, new Cookies());
+        try (var recorder = new WarcRecorder(fileName);
              var db = new DomainStateDb(dbTempFile))
         {
             new CrawlerRetreiver(httpFetcher, new DomainProber(domainBlacklist), specs, db, recorder).crawlDomain();

code/processes/converting-process/test/nu/marginalia/converting/processor/plugin/PdfDocumentProcessorPluginTest.java

@@ -0,0 +1,95 @@
+package nu.marginalia.converting.processor.plugin;
+
+import nu.marginalia.WmsaHome;
+import nu.marginalia.converting.processor.DocumentClass;
+import nu.marginalia.converting.processor.logic.DocumentLengthLogic;
+import nu.marginalia.converting.processor.logic.TitleExtractor;
+import nu.marginalia.converting.processor.plugin.specialization.DefaultSpecialization;
+import nu.marginalia.converting.processor.summary.SummaryExtractor;
+import nu.marginalia.converting.processor.summary.heuristic.*;
+import nu.marginalia.keyword.DocumentKeywordExtractor;
+import nu.marginalia.keyword.LinkTexts;
+import nu.marginalia.language.filter.LanguageFilter;
+import nu.marginalia.language.sentence.ThreadLocalSentenceExtractorProvider;
+import nu.marginalia.model.crawldata.CrawledDocument;
+import nu.marginalia.term_frequency_dict.TermFrequencyDict;
+import org.junit.jupiter.api.BeforeAll;
+import org.junit.jupiter.api.Disabled;
+import org.junit.jupiter.api.Tag;
+import org.junit.jupiter.api.Test;
+
+import java.io.IOException;
+import java.net.HttpURLConnection;
+import java.net.URI;
+import java.net.URISyntaxException;
+import java.nio.file.Files;
+import java.nio.file.Path;
+import java.time.Instant;
+
+@Tag("flaky")
+class PdfDocumentProcessorPluginTest {
+    static PdfDocumentProcessorPlugin plugin;
+
+    @BeforeAll
+    static void setUpBeforeClass() throws Exception {
+        var lm = WmsaHome.getLanguageModels();
+        plugin = new PdfDocumentProcessorPlugin(255,
+                new LanguageFilter(lm),
+                new ThreadLocalSentenceExtractorProvider(lm),
+                new DocumentKeywordExtractor(new TermFrequencyDict(lm)),
+                new DocumentLengthLogic(100),
+                new DefaultSpecialization(new SummaryExtractor(
+                        255,
+                        new DomFilterHeuristic(255),
+                        new TagDensityHeuristic(255),
+                        new OpenGraphDescriptionHeuristic(),
+                        new MetaDescriptionHeuristic(),
+                        new FallbackHeuristic()
+                ),
+                        new TitleExtractor(255)
+                        ));
+    }
+    public AbstractDocumentProcessorPlugin.DetailsWithWords testPdfFile(byte[] pdfBytes) throws Exception {
+        var doc = new CrawledDocument("test", "https://www.example.com/sample.pdf", "application/pdf", Instant.now().toString(), 200, "OK", "OK", "", pdfBytes, false, -1, null, null);
+        return plugin.createDetails(doc, new LinkTexts(), DocumentClass.NORMAL);
+    }
+
+    public AbstractDocumentProcessorPlugin.DetailsWithWords testPdfFile(Path file) throws Exception {
+        return testPdfFile(Files.readAllBytes(file));
+    }
+
+    private byte[] downloadPDF(String url) throws IOException, URISyntaxException {
+        HttpURLConnection conn = (HttpURLConnection) new URI(url).toURL().openConnection();
+        try {
+            return conn.getInputStream().readAllBytes();
+        } catch (IOException e) {
+            throw new RuntimeException(e);
+        } finally {
+            conn.disconnect();
+        }
+    }
+
+
+    @Disabled
+    @Test
+    void testingTool() throws Exception {
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample.pdf")).details().title);
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample2.pdf")).details().title);
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample3.pdf")).details().title);
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample4.pdf")).details().title);
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample5.pdf")).details().title);
+        System.out.println(testPdfFile(Path.of("/home/st_work/Work/sample6.pdf")).details().title);
+    }
+
+    @Disabled
+    @Test
+    void testingTool2() throws Exception {
+        System.out.println(plugin.convertPdfToHtml(Files.readAllBytes(Path.of("/home/st_work/Work/sample6.pdf"))));
+    }
+
+    @Test
+    void testMarginaliaSample() throws Exception {
+        var doc = plugin.convertPdfToHtml(downloadPDF("https://www.marginalia.nu/junk/test.pdf"));
+        System.out.println(doc.html());
+    }
+}

code/processes/converting-process/test/nu/marginalia/converting/processor/pubdate/PubDateSnifferTest.java

@@ -3,8 +3,8 @@ package nu.marginalia.converting.processor.pubdate;
 import nu.marginalia.WmsaHome;
 import nu.marginalia.converting.model.DocumentHeaders;
 import nu.marginalia.converting.processor.pubdate.heuristic.PubDateHeuristicDOMParsingPass2;
+import nu.marginalia.model.DocumentFormat;
 import nu.marginalia.model.EdgeUrl;
-import nu.marginalia.model.html.HtmlStandard;
 import org.jsoup.Jsoup;
 import org.junit.jupiter.api.Test;
 
@@ -74,7 +74,7 @@ class PubDateSnifferTest {
                         <time pubdate="pubdate" datetime="2022-08-24">time</time>
                         Wow, sure lor 'em boss
                         </article>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2022-08-24", ret.dateIso8601());
@@ -90,7 +90,7 @@ class PubDateSnifferTest {
                         <time>2022-08-24</time>
                         Wow, sure lor 'em boss
                         </article>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2022-08-24", ret.dateIso8601());
@@ -106,7 +106,7 @@ class PubDateSnifferTest {
                         <time class="published" datetime="July 13, 2006">July 13, 2006</time>
                         Wow, sure lor 'em boss
                         </article>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals(2006, ret.year());
@@ -116,14 +116,14 @@ class PubDateSnifferTest {
     public void testProblemCases() throws IOException, URISyntaxException {
         var ret = dateSniffer.getPubDate(new DocumentHeaders(""),
                 new EdgeUrl("https://www.example.com/"),
-                Jsoup.parse(Files.readString(WmsaHome.getHomePath().resolve("test-data/The Switch to Linux Begins .html"))), HtmlStandard.HTML5, true);
+                Jsoup.parse(Files.readString(WmsaHome.getHomePath().resolve("test-data/The Switch to Linux Begins .html"))), DocumentFormat.HTML5, true);
 
         assertFalse(ret.isEmpty());
         assertEquals(2006, ret.year());
 
         ret = dateSniffer.getPubDate(new DocumentHeaders(""),
                 new EdgeUrl("https://www.example.com/"),
-                Jsoup.parse(Files.readString(WmsaHome.getHomePath().resolve("test-data/Black Hat USA 2010 Understanding and Deploying DNSSEC by Paul Wouters and Patrick Nauber.html"))), HtmlStandard.XHTML, true);
+                Jsoup.parse(Files.readString(WmsaHome.getHomePath().resolve("test-data/Black Hat USA 2010 Understanding and Deploying DNSSEC by Paul Wouters and Patrick Nauber.html"))), DocumentFormat.XHTML, true);
 
         assertFalse(ret.isEmpty());
         assertEquals(2010, ret.year());
@@ -146,7 +146,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <meta itemprop="datePublished" content="2022-08-24" />
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2022-08-24", ret.dateIso8601());
@@ -160,7 +160,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <meta property="datePublished" content="2022-08-24" />
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2022-08-24", ret.dateIso8601());
@@ -174,7 +174,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <script type="application/ld+json">{"@context":"https:\\/\\/schema.org","@type":"Article","name":"In the Year 2525","url":"https:\\/\\/en.wikipedia.org\\/wiki\\/In_the_Year_2525","sameAs":"http:\\/\\/www.wikidata.org\\/entity\\/Q145269","mainEntity":"http:\\/\\/www.wikidata.org\\/entity\\/Q145269","author":{"@type":"Organization","name":"Contributors to Wikimedia projects"},"publisher":{"@type":"Organization","name":"Wikimedia Foundation, Inc.","logo":{"@type":"ImageObject","url":"https:\\/\\/www.wikimedia.org\\/static\\/images\\/wmf-hor-googpub.png"}},"datePublished":"2004-08-24T14:39:14Z","dateModified":"2022-10-20T11:54:37Z","image":"https:\\/\\/upload.wikimedia.org\\/wikipedia\\/commons\\/4\\/4a\\/In_the_Year_2525_by_Zager_and_Evans_US_vinyl_Side-A_RCA_release.png","headline":"song written and compsoed by Rick Evans, originally recorded by Zager and Evans and released in 1969"}</script><script type="application/ld+json">{"@context":"https:\\/\\/schema.org","@type":"Article","name":"In the Year 2525","url":"https:\\/\\/en.wikipedia.org\\/wiki\\/In_the_Year_2525","sameAs":"http:\\/\\/www.wikidata.org\\/entity\\/Q145269","mainEntity":"http:\\/\\/www.wikidata.org\\/entity\\/Q145269","author":{"@type":"Organization","name":"Contributors to Wikimedia projects"},"publisher":{"@type":"Organization","name":"Wikimedia Foundation, Inc.","logo":{"@type":"ImageObject","url":"https:\\/\\/www.wikimedia.org\\/static\\/images\\/wmf-hor-googpub.png"}},"datePublished":"2004-08-24T14:39:14Z","dateModified":"2022-10-20T11:54:37Z","image":"https:\\/\\/upload.wikimedia.org\\/wikipedia\\/commons\\/4\\/4a\\/In_the_Year_2525_by_Zager_and_Evans_US_vinyl_Side-A_RCA_release.png","headline":"song written and compsoed by Rick Evans, originally recorded by Zager and Evans and released in 1969"}</script>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2004-08-24", ret.dateIso8601());
@@ -188,7 +188,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <script type="application/ld+json" class="aioseop-schema">{"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://socialnomics.net/#organization","url":"https://socialnomics.net/","name":"Socialnomics","sameAs":[]},{"@type":"WebSite","@id":"https://socialnomics.net/#website","url":"https://socialnomics.net/","name":"Socialnomics","publisher":{"@id":"https://socialnomics.net/#organization"}},{"@type":"WebPage","@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#webpage","url":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/","inLanguage":"en-US","name":"3 Reasons Why You Should Adopt Java-based Technology For Your Business","isPartOf":{"@id":"https://socialnomics.net/#website"},"breadcrumb":{"@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#breadcrumblist"},"datePublished":"2016-12-27T21:01:36-06:00","dateModified":"2016-12-22T21:02:32-06:00"},{"@type":"Article","@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#article","isPartOf":{"@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#webpage"},"author":{"@id":"https://socialnomics.net/author/rahis-saifi/#author"},"headline":"3 Reasons Why You Should Adopt Java-based Technology For Your Business","datePublished":"2016-12-27T21:01:36-06:00","dateModified":"2016-12-22T21:02:32-06:00","commentCount":0,"mainEntityOfPage":{"@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#webpage"},"publisher":{"@id":"https://socialnomics.net/#organization"},"articleSection":"Business, business, java, Java Developers, programming languages"},{"@type":"Person","@id":"https://socialnomics.net/author/rahis-saifi/#author","name":"Rahis Saifi","sameAs":["https://www.facebook.com/RahisSaifiOfficial","https://www.twitter.com/57rahis"],"image":{"@type":"ImageObject","@id":"https://socialnomics.net/#personlogo","url":"https://secure.gravatar.com/avatar/e67f630f0b8bc87e59e111d5e955961d?s=96&d=mm&r=g","width":96,"height":96,"caption":"Rahis Saifi"}},{"@type":"BreadcrumbList","@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/#breadcrumblist","itemListElement":[{"@type":"ListItem","position":1,"item":{"@type":"WebPage","@id":"https://socialnomics.net/","url":"https://socialnomics.net/","name":"Socialnomics Blog"}},{"@type":"ListItem","position":2,"item":{"@type":"WebPage","@id":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/","url":"https://socialnomics.net/2016/12/27/3-reasons-why-you-should-adopt-java-based-technology-for-your-business/","name":"3 Reasons Why You Should Adopt Java-based Technology For Your Business"}}]}]}</script>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2016-12-27", ret.dateIso8601());
@@ -202,7 +202,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <title>No date in the HTML</title>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertNull(ret.dateIso8601());
@@ -217,7 +217,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <title>No date in the HTML</title>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertEquals("2022-02-03", ret.dateIso8601());
@@ -232,7 +232,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <p>Published 2003, updated 2022</p>
-                        """), HtmlStandard.HTML5, true);
+                        """), DocumentFormat.HTML5, true);
 
         assertFalse(ret.isEmpty());
         assertNull(ret.dateIso8601());
@@ -258,7 +258,7 @@ class PubDateSnifferTest {
                         <!doctype html>
                         <html>
                         <div style="float: left;">&nbsp;<b>Post subject:</b> Keyboards.</div><div style="float: right;"><span class="postdetails"><b><img src="./styles/subsilver2/imageset/icon_post_target.gif" width="12" height="9" alt="Post" title="Post" /> <a  href="./viewtopic.php?p=34580&amp;sid=cf0c13dedebb4fea1f03fa73e510cd9f#p34580">#1</a></b></span>&nbsp;<b>Posted:</b> Sun Oct 03, 2010 5:37 pm&nbsp;</div>
-                        """), HtmlStandard.UNKNOWN, true);
+                        """), DocumentFormat.UNKNOWN, true);
 
         assertFalse(ret.isEmpty());
         assertNull(ret.dateIso8601());

code/processes/crawling-process/build.gradle

@@ -58,12 +58,15 @@ dependencies {
     implementation libs.jsoup
     implementation libs.opencsv
     implementation libs.fastutil
+    implementation libs.bundles.curator
 
     implementation libs.bundles.mariadb
+    implementation libs.bundles.httpcomponents
 
     testImplementation libs.bundles.slf4j.test
     testImplementation libs.bundles.junit
     testImplementation libs.mockito
+    testImplementation libs.wiremock
 
     testImplementation project(':code:processes:test-data')
 }

code/processes/crawling-process/java/nu/marginalia/crawl/CrawlerMain.java

@@ -25,9 +25,12 @@ import nu.marginalia.mq.MessageQueueFactory;
 import nu.marginalia.process.ProcessConfiguration;
 import nu.marginalia.process.ProcessConfigurationModule;
 import nu.marginalia.process.ProcessMainClass;
+import nu.marginalia.process.control.ProcessEventLog;
 import nu.marginalia.process.control.ProcessHeartbeatImpl;
 import nu.marginalia.process.log.WorkLog;
+import nu.marginalia.service.discovery.ServiceRegistryIf;
 import nu.marginalia.service.module.DatabaseModule;
+import nu.marginalia.service.module.ServiceDiscoveryModule;
 import nu.marginalia.slop.SlopCrawlDataRecord;
 import nu.marginalia.storage.FileStorageService;
 import nu.marginalia.storage.model.FileStorageId;
@@ -41,11 +44,9 @@ import java.nio.file.Files;
 import java.nio.file.Path;
 import java.nio.file.StandardCopyOption;
 import java.security.Security;
-import java.util.ArrayList;
+import java.util.*;
-import java.util.Collections;
-import java.util.List;
-import java.util.Map;
 import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.LinkedBlockingQueue;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicInteger;
 
@@ -56,6 +57,7 @@ public class CrawlerMain extends ProcessMainClass {
 
     private final UserAgent userAgent;
     private final ProcessHeartbeatImpl heartbeat;
+    private final ProcessEventLog eventLog;
     private final DomainProber domainProber;
     private final FileStorageService fileStorageService;
     private final AnchorTagsSourceFactory anchorTagsSourceFactory;
@@ -63,12 +65,15 @@ public class CrawlerMain extends ProcessMainClass {
     private final HikariDataSource dataSource;
     private final DomainBlacklist blacklist;
     private final int node;
+    private final ServiceRegistryIf serviceRegistry;
     private final SimpleBlockingThreadPool pool;
 
     private final DomainLocks domainLocks = new DomainLocks();
 
     private final Map<String, CrawlTask> pendingCrawlTasks = new ConcurrentHashMap<>();
 
+    private final LinkedBlockingQueue<CrawlTask> retryQueue = new LinkedBlockingQueue<>();
+
     private final AtomicInteger tasksDone = new AtomicInteger(0);
     private final HttpFetcherImpl fetcher;
 
@@ -84,6 +89,7 @@ public class CrawlerMain extends ProcessMainClass {
     public CrawlerMain(UserAgent userAgent,
                        HttpFetcherImpl httpFetcher,
                        ProcessHeartbeatImpl heartbeat,
+                       ProcessEventLog eventLog,
                        MessageQueueFactory messageQueueFactory, DomainProber domainProber,
                        FileStorageService fileStorageService,
                        ProcessConfiguration processConfiguration,
@@ -91,6 +97,7 @@ public class CrawlerMain extends ProcessMainClass {
                        WarcArchiverFactory warcArchiverFactory,
                        HikariDataSource dataSource,
                        DomainBlacklist blacklist,
+                       ServiceRegistryIf serviceRegistry,
                        Gson gson) throws InterruptedException {
 
         super(messageQueueFactory, processConfiguration, gson, CRAWLER_INBOX);
@@ -98,6 +105,7 @@ public class CrawlerMain extends ProcessMainClass {
         this.userAgent = userAgent;
         this.fetcher = httpFetcher;
         this.heartbeat = heartbeat;
+        this.eventLog = eventLog;
         this.domainProber = domainProber;
         this.fileStorageService = fileStorageService;
         this.anchorTagsSourceFactory = anchorTagsSourceFactory;
@@ -105,10 +113,20 @@ public class CrawlerMain extends ProcessMainClass {
         this.dataSource = dataSource;
         this.blacklist = blacklist;
         this.node = processConfiguration.node();
+        this.serviceRegistry = serviceRegistry;
+
+        SimpleBlockingThreadPool.ThreadType threadType;
+        if (Boolean.getBoolean("crawler.useVirtualThreads")) {
+            threadType = SimpleBlockingThreadPool.ThreadType.VIRTUAL;
+        }
+        else {
+            threadType = SimpleBlockingThreadPool.ThreadType.PLATFORM;
+        }
 
         pool = new SimpleBlockingThreadPool("CrawlerPool",
                 Integer.getInteger("crawler.poolSize", 256),
-                1);
+                1,
+                threadType);
 
 
         // Wait for the blacklist to be loaded before starting the crawl
@@ -138,12 +156,17 @@ public class CrawlerMain extends ProcessMainClass {
             Injector injector = Guice.createInjector(
                     new CrawlerModule(),
                     new ProcessConfigurationModule("crawler"),
+                    new ServiceDiscoveryModule(),
                     new DatabaseModule(false)
             );
             var crawler = injector.getInstance(CrawlerMain.class);
 
             var instructions = crawler.fetchInstructions(nu.marginalia.mqapi.crawling.CrawlRequest.class);
+
+            crawler.serviceRegistry.registerProcess("crawler", crawler.node);
+
             try {
+                crawler.eventLog.logEvent("CRAWLER-INFO", "Crawling started");
                 var req = instructions.value();
                 if (req.targetDomainName != null) {
                     crawler.runForSingleDomain(req.targetDomainName, req.crawlStorage);
@@ -151,11 +174,15 @@ public class CrawlerMain extends ProcessMainClass {
                 else {
                     crawler.runForDatabaseDomains(req.crawlStorage);
                 }
+                crawler.eventLog.logEvent("CRAWLER-INFO", "Crawl completed successfully");
                 instructions.ok();
             } catch (Exception ex) {
                 logger.error("Crawler failed", ex);
                 instructions.err();
             }
+            finally {
+                crawler.serviceRegistry.deregisterProcess("crawler", crawler.node);
+            }
 
             TimeUnit.SECONDS.sleep(5);
         }
@@ -224,10 +251,7 @@ public class CrawlerMain extends ProcessMainClass {
 
         logger.info("Loaded {} domains", crawlSpecRecords.size());
 
-        // Shuffle the domains to ensure we get a good mix of domains in each crawl,
+        crawlSpecRecords.sort(crawlSpecArrangement(crawlSpecRecords));
-        // so that e.g. the big domains don't get all crawled at once, or we end up
-        // crawling the same server in parallel from different subdomains...
-        Collections.shuffle(crawlSpecRecords);
 
         // First a validation run to ensure the file is all good to parse
         if (crawlSpecRecords.isEmpty()) {
@@ -258,28 +282,44 @@ public class CrawlerMain extends ProcessMainClass {
                 if (workLog.isJobFinished(crawlSpec.domain))
                     continue;
 
-                var task = new CrawlTask(
+                var task = new CrawlTask(crawlSpec, anchorTagsSource, outputDir, warcArchiver, domainStateDb, workLog);
-                        crawlSpec,
-                        anchorTagsSource,
-                        outputDir,
-                        warcArchiver,
-                        domainStateDb,
-                        workLog);
 
                 // Try to run immediately, to avoid unnecessarily keeping the entire work set in RAM
                 if (!trySubmitDeferredTask(task)) {
-                    // Otherwise add to the taskList for deferred execution
+
+                    // Drain the retry queue to the taskList, and try to submit any tasks that are in the retry queue
+                    retryQueue.drainTo(taskList);
+                    taskList.removeIf(this::trySubmitDeferredTask);
+
+                    // Then add this new task to the retry queue
                     taskList.add(task);
                 }
             }
 
              // Schedule viable tasks for execution until list is empty
-            while (!taskList.isEmpty()) {
+            for (int emptyRuns = 0;emptyRuns < 300;) {
-                taskList.removeIf(this::trySubmitDeferredTask);
+                boolean hasTasks = !taskList.isEmpty();
 
-                // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                // The order of these checks  very important to avoid a race condition
-                // we might have no new viable tasks to run for hours on end
+                // where we miss a task that is put into the retry queue
-                TimeUnit.MILLISECONDS.sleep(50);
+                boolean hasRunningTasks = pool.getActiveCount() > 0;
+                boolean hasRetryTasks = !retryQueue.isEmpty();
+
+                if (hasTasks || hasRetryTasks || hasRunningTasks) {
+                    retryQueue.drainTo(taskList);
+
+                    // Try to submit any tasks that are in the retry queue (this will block if the pool is full)
+                    taskList.removeIf(this::trySubmitDeferredTask);
+
+                    // Add a small pause here to avoid busy looping toward the end of the execution cycle when
+                    // we might have no new viable tasks to run for hours on end
+                    TimeUnit.MILLISECONDS.sleep(5);
+                } else {
+                    // We have no tasks to run, and no tasks in the retry queue
+                    // but we wait a bit to see if any new tasks come in via the retry queue
+                    emptyRuns++;
+                    TimeUnit.SECONDS.sleep(1);
+                }
             }
 
             logger.info("Shutting down the pool, waiting for tasks to complete...");
@@ -306,6 +346,30 @@ public class CrawlerMain extends ProcessMainClass {
         }
     }
 
+    /** Create a comparator that sorts the crawl specs in a way that is beneficial for the crawl,
+     * we want to enqueue domains that have common top domains first, but otherwise have a random
+     * order.
+     * <p></p>
+     * Note, we can't use hash codes for randomization as it is not desirable to have the same order
+     * every time the process is restarted (and CrawlSpecRecord is a record, which defines equals and
+     * hashcode based on the fields).
+     * */
+    private Comparator<CrawlSpecRecord> crawlSpecArrangement(List<CrawlSpecRecord> records) {
+        Random r = new Random();
+        Map<String, Integer> topDomainCounts = new HashMap<>(4 + (int) Math.sqrt(records.size()));
+        Map<String, Integer> randomOrder = new HashMap<>(records.size());
+
+        for (var spec : records) {
+            topDomainCounts.merge(EdgeDomain.getTopDomain(spec.domain), 1, Integer::sum);
+            randomOrder.put(spec.domain, r.nextInt());
+        }
+
+        return Comparator.comparing((CrawlSpecRecord spec) -> topDomainCounts.getOrDefault(EdgeDomain.getTopDomain(spec.domain), 0) >= 8)
+                .reversed()
+                .thenComparing(spec -> randomOrder.get(spec.domain))
+                .thenComparing(Record::hashCode); // non-deterministic tie-breaker to
+    }
+
     /** Submit a task for execution if it can be run, returns true if it was submitted
      * or if it can be discarded */
     private boolean trySubmitDeferredTask(CrawlTask task) {
@@ -387,7 +451,7 @@ public class CrawlerMain extends ProcessMainClass {
         /** Best effort indicator whether we could start this now without getting stuck in
          * DomainLocks purgatory */
         public boolean canRun() {
-            return domainLocks.canLock(new EdgeDomain(domain));
+            return domainLocks.isLockableHint(new EdgeDomain(domain));
         }
 
         @Override
@@ -398,66 +462,76 @@ public class CrawlerMain extends ProcessMainClass {
                 return;
             }
 
-            Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
+            Optional<DomainLocks.DomainLock> lock = domainLocks.tryLockDomain(new EdgeDomain(domain));
-            Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
+            // We don't have a lock, so we can't run this task
-            Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
+            // we return to avoid blocking the pool for too long
-
+            if (lock.isEmpty()) {
-            // Move the WARC file to a temp file if it exists, so we can resume the crawl using the old data
-            // while writing to the same file name as before
-            if (Files.exists(newWarcFile)) {
-                Files.move(newWarcFile, tempFile, StandardCopyOption.REPLACE_EXISTING);
-            }
-            else {
-                Files.deleteIfExists(tempFile);
-            }
-
-            try (var warcRecorder = new WarcRecorder(newWarcFile, fetcher); // write to a temp file for now
-                 var retriever = new CrawlerRetreiver(fetcher, domainProber, specification, domainStateDb, warcRecorder);
-                 CrawlDataReference reference = getReference()
-            )
-            {
-                // Resume the crawl if it was aborted
-                if (Files.exists(tempFile)) {
-                    retriever.syncAbortedRun(tempFile);
-                    Files.delete(tempFile);
-                }
-
-                DomainLinks domainLinks = anchorTagsSource.getAnchorTags(domain);
-
-                int size;
-                try (var lock = domainLocks.lockDomain(new EdgeDomain(domain))) {
-                    size = retriever.crawlDomain(domainLinks, reference);
-                }
-
-                // Delete the reference crawl data if it's not the same as the new one
-                // (mostly a case when migrating from legacy->warc)
-                reference.delete();
-
-                // Convert the WARC file to Parquet
-                SlopCrawlDataRecord
-                        .convertWarc(domain, userAgent, newWarcFile, slopFile);
-
-                // Optionally archive the WARC file if full retention is enabled,
-                // otherwise delete it:
-                warcArchiver.consumeWarc(newWarcFile, domain);
-
-                // Mark the domain as finished in the work log
-                workLog.setJobToFinished(domain, slopFile.toString(), size);
-
-                // Update the progress bar
-                heartbeat.setProgress(tasksDone.incrementAndGet() / (double) totalTasks);
-
-                logger.info("Fetched {}", domain);
-            } catch (Exception e) {
-                logger.error("Error fetching domain " + domain, e);
-            }
-            finally {
-                // We don't need to double-count these; it's also kept in the workLog
                 pendingCrawlTasks.remove(domain);
-                Thread.currentThread().setName("[idle]");
+                retryQueue.put(this);
+                return;
+            }
+            DomainLocks.DomainLock domainLock = lock.get();
 
-                Files.deleteIfExists(newWarcFile);
+            try (domainLock) {
-                Files.deleteIfExists(tempFile);
+                Thread.currentThread().setName("crawling:" + domain);
+
+                Path newWarcFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.LIVE);
+                Path tempFile = CrawlerOutputFile.createWarcPath(outputDir, id, domain, CrawlerOutputFile.WarcFileVersion.TEMP);
+                Path slopFile = CrawlerOutputFile.createSlopPath(outputDir, id, domain);
+
+                // Move the WARC file to a temp file if it exists, so we can resume the crawl using the old data
+                // while writing to the same file name as before
+                if (Files.exists(newWarcFile)) {
+                    Files.move(newWarcFile, tempFile, StandardCopyOption.REPLACE_EXISTING);
+                }
+                else {
+                    Files.deleteIfExists(tempFile);
+                }
+
+                try (var warcRecorder = new WarcRecorder(newWarcFile); // write to a temp file for now
+                     var retriever = new CrawlerRetreiver(fetcher, domainProber, specification, domainStateDb, warcRecorder);
+                     CrawlDataReference reference = getReference())
+                {
+                    // Resume the crawl if it was aborted
+                    if (Files.exists(tempFile)) {
+                        retriever.syncAbortedRun(tempFile);
+                        Files.delete(tempFile);
+                    }
+
+                    DomainLinks domainLinks = anchorTagsSource.getAnchorTags(domain);
+
+                    int size = retriever.crawlDomain(domainLinks, reference);
+
+                    // Delete the reference crawl data if it's not the same as the new one
+                    // (mostly a case when migrating from legacy->warc)
+                    reference.delete();
+
+                    // Convert the WARC file to Slop
+                    SlopCrawlDataRecord
+                            .convertWarc(domain, userAgent, newWarcFile, slopFile);
+
+                    // Optionally archive the WARC file if full retention is enabled,
+                    // otherwise delete it:
+                    warcArchiver.consumeWarc(newWarcFile, domain);
+
+                    // Mark the domain as finished in the work log
+                    workLog.setJobToFinished(domain, slopFile.toString(), size);
+
+                    // Update the progress bar
+                    heartbeat.setProgress(tasksDone.incrementAndGet() / (double) totalTasks);
+
+                    logger.info("Fetched {}", domain);
+                } catch (Exception e) {
+                    logger.error("Error fetching domain " + domain, e);
+                }
+                finally {
+                    // We don't need to double-count these; it's also kept in the workLog
+                    pendingCrawlTasks.remove(domain);
+                    Thread.currentThread().setName("[idle]");
+
+                    Files.deleteIfExists(newWarcFile);
+                    Files.deleteIfExists(tempFile);
+                }
             }
         }
 
@@ -474,7 +548,7 @@ public class CrawlerMain extends ProcessMainClass {
                     return new CrawlDataReference(slopPath);
                 }
 
-            } catch (IOException e) {
+            } catch (Exception e) {
                 logger.debug("Failed to read previous crawl data for {}", specification.domain());
             }