@postalsys/email-content 0.0.1

package/dist/corpus/thread-index.js

@@ -0,0 +1,182 @@
+"use strict";
+/**
+ * Thread reconstruction over a scanned corpus.
+ *
+ * Resolves a message's In-Reply-To and References to files that are actually present, so that
+ * Stage 5 has something to match against. Measured over the working corpus, 90.9 percent of replies
+ * resolve their direct parent this way.
+ *
+ * Ancestors are returned nearest first. The direct parent carries most of the value: it is what the
+ * sender's client quoted from. Older ancestors matter for deeply nested threads, where the reply
+ * contains several generations of history at once.
+ */
+var __createBinding = (this && this.__createBinding) || (Object.create ? (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    var desc = Object.getOwnPropertyDescriptor(m, k);
+    if (!desc || ("get" in desc ? !m.__esModule : desc.writable || desc.configurable)) {
+      desc = { enumerable: true, get: function() { return m[k]; } };
+    }
+    Object.defineProperty(o, k2, desc);
+}) : (function(o, m, k, k2) {
+    if (k2 === undefined) k2 = k;
+    o[k2] = m[k];
+}));
+var __setModuleDefault = (this && this.__setModuleDefault) || (Object.create ? (function(o, v) {
+    Object.defineProperty(o, "default", { enumerable: true, value: v });
+}) : function(o, v) {
+    o["default"] = v;
+});
+var __importStar = (this && this.__importStar) || (function () {
+    var ownKeys = function(o) {
+        ownKeys = Object.getOwnPropertyNames || function (o) {
+            var ar = [];
+            for (var k in o) if (Object.prototype.hasOwnProperty.call(o, k)) ar[ar.length] = k;
+            return ar;
+        };
+        return ownKeys(o);
+    };
+    return function (mod) {
+        if (mod && mod.__esModule) return mod;
+        var result = {};
+        if (mod != null) for (var k = ownKeys(mod), i = 0; i < k.length; i++) if (k[i] !== "default") __createBinding(result, mod, k[i]);
+        __setModuleDefault(result, mod);
+        return result;
+    };
+})();
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.DEFAULT_MAX_ANCESTORS = void 0;
+exports.corpusThreadResolver = corpusThreadResolver;
+exports.folderThreadResolver = folderThreadResolver;
+const node_fs_1 = __importDefault(require("node:fs"));
+const node_path_1 = __importDefault(require("node:path"));
+const history_1 = require("../match/history");
+/** How many ancestors to load by default. Each one costs a file read and a parse. */
+exports.DEFAULT_MAX_ANCESTORS = 3;
+/**
+ * Build a resolver backed by a scanned corpus database.
+ *
+ * The lookup is a single indexed query per candidate Message-ID, so no in memory index of the whole
+ * corpus is needed. That matters: a Message-ID map over 320k messages is tens of megabytes that
+ * would otherwise be paid for on every process start.
+ */
+function corpusThreadResolver(store, root) {
+    const cache = new Map();
+    const load = async (messageId, isParent) => {
+        const cached = cache.get(messageId);
+        if (cached !== undefined) {
+            return cached ? { ...cached, isParent } : null;
+        }
+        const row = store.queryOne('SELECT path FROM messages WHERE message_id = ? LIMIT 1', messageId);
+        if (!row) {
+            cache.set(messageId, null);
+            return null;
+        }
+        try {
+            const raw = node_fs_1.default.readFileSync(node_path_1.default.join(root, row.path));
+            const document = await (0, history_1.buildHistoryDocument)(raw, row.path, isParent);
+            cache.set(messageId, document);
+            return document;
+        }
+        catch {
+            cache.set(messageId, null);
+            return null;
+        }
+    };
+    return {
+        async resolve(headers, maxAncestors = exports.DEFAULT_MAX_ANCESTORS) {
+            const documents = [];
+            const seen = new Set();
+            if (headers.inReplyTo) {
+                seen.add(headers.inReplyTo);
+                const parent = await load(headers.inReplyTo, true);
+                if (parent) {
+                    documents.push(parent);
+                }
+            }
+            // References run oldest first, so walk backwards for nearest ancestors.
+            for (let i = headers.references.length - 1; i >= 0 && documents.length < maxAncestors; i--) {
+                const reference = headers.references[i];
+                if (seen.has(reference)) {
+                    continue;
+                }
+                seen.add(reference);
+                const ancestor = await load(reference, false);
+                if (ancestor) {
+                    documents.push(ancestor);
+                }
+            }
+            return documents;
+        }
+    };
+}
+/** Build a resolver from a folder of .eml files, for callers with no corpus database. */
+function folderThreadResolver(folder) {
+    let index = null;
+    const buildIndex = async () => {
+        const map = new Map();
+        let entries;
+        try {
+            entries = node_fs_1.default.readdirSync(folder).filter(name => /\.eml$/i.test(name));
+        }
+        catch {
+            return map;
+        }
+        const { parseMessage } = await Promise.resolve().then(() => __importStar(require('../mime/parse')));
+        for (const name of entries) {
+            try {
+                // Headers live at the front, so a bounded read is enough to index a folder.
+                const handle = node_fs_1.default.openSync(node_path_1.default.join(folder, name), 'r');
+                const buffer = Buffer.alloc(32768);
+                const read = node_fs_1.default.readSync(handle, buffer, 0, buffer.length, 0);
+                node_fs_1.default.closeSync(handle);
+                const message = await parseMessage(buffer.subarray(0, read));
+                if (message.headers.messageId) {
+                    map.set(message.headers.messageId, name);
+                }
+            }
+            catch {
+                // A file that cannot be indexed simply is not available as history.
+            }
+        }
+        return map;
+    };
+    return {
+        async resolve(headers, maxAncestors = exports.DEFAULT_MAX_ANCESTORS) {
+            if (!index) {
+                index = await buildIndex();
+            }
+            const documents = [];
+            const wanted = [];
+            if (headers.inReplyTo) {
+                wanted.push({ id: headers.inReplyTo, isParent: true });
+            }
+            for (let i = headers.references.length - 1; i >= 0; i--) {
+                const reference = headers.references[i];
+                if (reference !== headers.inReplyTo) {
+                    wanted.push({ id: reference, isParent: false });
+                }
+            }
+            for (const item of wanted) {
+                if (documents.length >= maxAncestors) {
+                    break;
+                }
+                const name = index.get(item.id);
+                if (!name) {
+                    continue;
+                }
+                try {
+                    const raw = node_fs_1.default.readFileSync(node_path_1.default.join(folder, name));
+                    documents.push(await (0, history_1.buildHistoryDocument)(raw, name, item.isParent));
+                }
+                catch {
+                    // Unreadable ancestor, skip.
+                }
+            }
+            return documents;
+        }
+    };
+}
+//# sourceMappingURL=thread-index.js.map

package/dist/corpus/thread-index.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"thread-index.js","sourceRoot":"","sources":["../../src/corpus/thread-index.ts"],"names":[],"mappings":";AAAA;;;;;;;;;;GAUG;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAuBH,oDAqDC;AAGD,oDAiEC;AA9ID,sDAAyB;AACzB,0DAA6B;AAC7B,8CAA8E;AAS9E,qFAAqF;AACxE,QAAA,qBAAqB,GAAG,CAAC,CAAC;AAEvC;;;;;;GAMG;AACH,SAAgB,oBAAoB,CAAC,KAAkB,EAAE,IAAY;IACjE,MAAM,KAAK,GAAG,IAAI,GAAG,EAAkC,CAAC;IAExD,MAAM,IAAI,GAAG,KAAK,EAAE,SAAiB,EAAE,QAAiB,EAAmC,EAAE;QACzF,MAAM,MAAM,GAAG,KAAK,CAAC,GAAG,CAAC,SAAS,CAAC,CAAC;QACpC,IAAI,MAAM,KAAK,SAAS,EAAE,CAAC;YACvB,OAAO,MAAM,CAAC,CAAC,CAAC,EAAE,GAAG,MAAM,EAAE,QAAQ,EAAE,CAAC,CAAC,CAAC,IAAI,CAAC;QACnD,CAAC;QACD,MAAM,GAAG,GAAG,KAAK,CAAC,QAAQ,CAAmB,wDAAwD,EAAE,SAAS,CAAC,CAAC;QAClH,IAAI,CAAC,GAAG,EAAE,CAAC;YACP,KAAK,CAAC,GAAG,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;YAC3B,OAAO,IAAI,CAAC;QAChB,CAAC;QACD,IAAI,CAAC;YACD,MAAM,GAAG,GAAG,iBAAE,CAAC,YAAY,CAAC,mBAAI,CAAC,IAAI,CAAC,IAAI,EAAE,GAAG,CAAC,IAAI,CAAC,CAAC,CAAC;YACvD,MAAM,QAAQ,GAAG,MAAM,IAAA,8BAAoB,EAAC,GAAG,EAAE,GAAG,CAAC,IAAI,EAAE,QAAQ,CAAC,CAAC;YACrE,KAAK,CAAC,GAAG,CAAC,SAAS,EAAE,QAAQ,CAAC,CAAC;YAC/B,OAAO,QAAQ,CAAC;QACpB,CAAC;QAAC,MAAM,CAAC;YACL,KAAK,CAAC,GAAG,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;YAC3B,OAAO,IAAI,CAAC;QAChB,CAAC;IACL,CAAC,CAAC;IAEF,OAAO;QACH,KAAK,CAAC,OAAO,CAAC,OAAuB,EAAE,YAAY,GAAG,6BAAqB;YACvE,MAAM,SAAS,GAAsB,EAAE,CAAC;YACxC,MAAM,IAAI,GAAG,IAAI,GAAG,EAAU,CAAC;YAE/B,IAAI,OAAO,CAAC,SAAS,EAAE,CAAC;gBACpB,IAAI,CAAC,GAAG,CAAC,OAAO,CAAC,SAAS,CAAC,CAAC;gBAC5B,MAAM,MAAM,GAAG,MAAM,IAAI,CAAC,OAAO,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;gBACnD,IAAI,MAAM,EAAE,CAAC;oBACT,SAAS,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;gBAC3B,CAAC;YACL,CAAC;YAED,wEAAwE;YACxE,KAAK,IAAI,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,IAAI,SAAS,CAAC,MAAM,GAAG,YAAY,EAAE,CAAC,EAAE,EAAE,CAAC;gBACzF,MAAM,SAAS,GAAG,OAAO,CAAC,UAAU,CAAC,CAAC,CAAW,CAAC;gBAClD,IAAI,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,EAAE,CAAC;oBACtB,SAAS;gBACb,CAAC;gBACD,IAAI,CAAC,GAAG,CAAC,SAAS,CAAC,CAAC;gBACpB,MAAM,QAAQ,GAAG,MAAM,IAAI,CAAC,SAAS,EAAE,KAAK,CAAC,CAAC;gBAC9C,IAAI,QAAQ,EAAE,CAAC;oBACX,SAAS,CAAC,IAAI,CAAC,QAAQ,CAAC,CAAC;gBAC7B,CAAC;YACL,CAAC;YAED,OAAO,SAAS,CAAC;QACrB,CAAC;KACJ,CAAC;AACN,CAAC;AAED,yFAAyF;AACzF,SAAgB,oBAAoB,CAAC,MAAc;IAC/C,IAAI,KAAK,GAA+B,IAAI,CAAC;IAE7C,MAAM,UAAU,GAAG,KAAK,IAAkC,EAAE;QACxD,MAAM,GAAG,GAAG,IAAI,GAAG,EAAkB,CAAC;QACtC,IAAI,OAAiB,CAAC;QACtB,IAAI,CAAC;YACD,OAAO,GAAG,iBAAE,CAAC,WAAW,CAAC,MAAM,CAAC,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,CAAC,SAAS,CAAC,IAAI,CAAC,IAAI,CAAC,CAAC,CAAC;QAC1E,CAAC;QAAC,MAAM,CAAC;YACL,OAAO,GAAG,CAAC;QACf,CAAC;QACD,MAAM,EAAE,YAAY,EAAE,GAAG,wDAAa,eAAe,GAAC,CAAC;QACvD,KAAK,MAAM,IAAI,IAAI,OAAO,EAAE,CAAC;YACzB,IAAI,CAAC;gBACD,4EAA4E;gBAC5E,MAAM,MAAM,GAAG,iBAAE,CAAC,QAAQ,CAAC,mBAAI,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,EAAE,GAAG,CAAC,CAAC;gBACzD,MAAM,MAAM,GAAG,MAAM,CAAC,KAAK,CAAC,KAAK,CAAC,CAAC;gBACnC,MAAM,IAAI,GAAG,iBAAE,CAAC,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,CAAC,CAAC,CAAC;gBAC9D,iBAAE,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC;gBACrB,MAAM,OAAO,GAAG,MAAM,YAAY,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,EAAE,IAAI,CAAC,CAAC,CAAC;gBAC7D,IAAI,OAAO,CAAC,OAAO,CAAC,SAAS,EAAE,CAAC;oBAC5B,GAAG,CAAC,GAAG,CAAC,OAAO,CAAC,OAAO,CAAC,SAAS,EAAE,IAAI,CAAC,CAAC;gBAC7C,CAAC;YACL,CAAC;YAAC,MAAM,CAAC;gBACL,oEAAoE;YACxE,CAAC;QACL,CAAC;QACD,OAAO,GAAG,CAAC;IACf,CAAC,CAAC;IAEF,OAAO;QACH,KAAK,CAAC,OAAO,CAAC,OAAuB,EAAE,YAAY,GAAG,6BAAqB;YACvE,IAAI,CAAC,KAAK,EAAE,CAAC;gBACT,KAAK,GAAG,MAAM,UAAU,EAAE,CAAC;YAC/B,CAAC;YACD,MAAM,SAAS,GAAsB,EAAE,CAAC;YACxC,MAAM,MAAM,GAAwC,EAAE,CAAC;YACvD,IAAI,OAAO,CAAC,SAAS,EAAE,CAAC;gBACpB,MAAM,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,OAAO,CAAC,SAAS,EAAE,QAAQ,EAAE,IAAI,EAAE,CAAC,CAAC;YAC3D,CAAC;YACD,KAAK,IAAI,CAAC,GAAG,OAAO,CAAC,UAAU,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,EAAE,CAAC,EAAE,EAAE,CAAC;gBACtD,MAAM,SAAS,GAAG,OAAO,CAAC,UAAU,CAAC,CAAC,CAAW,CAAC;gBAClD,IAAI,SAAS,KAAK,OAAO,CAAC,SAAS,EAAE,CAAC;oBAClC,MAAM,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,SAAS,EAAE,QAAQ,EAAE,KAAK,EAAE,CAAC,CAAC;gBACpD,CAAC;YACL,CAAC;YAED,KAAK,MAAM,IAAI,IAAI,MAAM,EAAE,CAAC;gBACxB,IAAI,SAAS,CAAC,MAAM,IAAI,YAAY,EAAE,CAAC;oBACnC,MAAM;gBACV,CAAC;gBACD,MAAM,IAAI,GAAG,KAAK,CAAC,GAAG,CAAC,IAAI,CAAC,EAAE,CAAC,CAAC;gBAChC,IAAI,CAAC,IAAI,EAAE,CAAC;oBACR,SAAS;gBACb,CAAC;gBACD,IAAI,CAAC;oBACD,MAAM,GAAG,GAAG,iBAAE,CAAC,YAAY,CAAC,mBAAI,CAAC,IAAI,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC,CAAC;oBACrD,SAAS,CAAC,IAAI,CAAC,MAAM,IAAA,8BAAoB,EAAC,GAAG,EAAE,IAAI,EAAE,IAAI,CAAC,QAAQ,CAAC,CAAC,CAAC;gBACzE,CAAC;gBAAC,MAAM,CAAC;oBACL,6BAA6B;gBACjC,CAAC;YACL,CAAC;YACD,OAAO,SAAS,CAAC;QACrB,CAAC;KACJ,CAAC;AACN,CAAC"}

package/dist/corpus/worker.d.ts

@@ -0,0 +1,21 @@
+/**
+ * Corpus scan worker.
+ *
+ * One message per task. Every failure mode is caught and reported as a result rather than as a
+ * thrown error, so that a malformed file degrades to a row with status 'failed' instead of killing
+ * the worker. The only thing that can still take a worker down is running out of memory or an
+ * infinite loop, and the pool's task timeout covers both.
+ */
+import { type MessageRecord } from './analyze';
+import { type FileRecord } from './store';
+export interface ScanTask {
+    /** Path relative to the corpus root. */
+    relPath: string;
+    size: number;
+    mtime: number;
+}
+export interface ScanResult {
+    file: FileRecord;
+    message: MessageRecord | null;
+}
+//# sourceMappingURL=worker.d.ts.map

package/dist/corpus/worker.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"worker.d.ts","sourceRoot":"","sources":["../../src/corpus/worker.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAOH,OAAO,EAAkB,KAAK,aAAa,EAAE,MAAM,WAAW,CAAC;AAC/D,OAAO,EAAE,KAAK,UAAU,EAAmB,MAAM,SAAS,CAAC;AAE3D,MAAM,WAAW,QAAQ;IACrB,wCAAwC;IACxC,OAAO,EAAE,MAAM,CAAC;IAChB,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,EAAE,MAAM,CAAC;CACjB;AAED,MAAM,WAAW,UAAU;IACvB,IAAI,EAAE,UAAU,CAAC;IACjB,OAAO,EAAE,aAAa,GAAG,IAAI,CAAC;CACjC"}

package/dist/corpus/worker.js

@@ -0,0 +1,74 @@
+"use strict";
+/**
+ * Corpus scan worker.
+ *
+ * One message per task. Every failure mode is caught and reported as a result rather than as a
+ * thrown error, so that a malformed file degrades to a row with status 'failed' instead of killing
+ * the worker. The only thing that can still take a worker down is running out of memory or an
+ * infinite loop, and the pool's task timeout covers both.
+ */
+var __importDefault = (this && this.__importDefault) || function (mod) {
+    return (mod && mod.__esModule) ? mod : { "default": mod };
+};
+Object.defineProperty(exports, "__esModule", { value: true });
+const node_fs_1 = __importDefault(require("node:fs"));
+const node_path_1 = __importDefault(require("node:path"));
+const node_worker_threads_1 = require("node:worker_threads");
+const parse_1 = require("../mime/parse");
+const hash_1 = require("../util/hash");
+const analyze_1 = require("./analyze");
+const config = node_worker_threads_1.workerData;
+function errorDetail(err) {
+    if (err instanceof Error) {
+        return `${err.name}: ${err.message}`.slice(0, 300);
+    }
+    return String(err).slice(0, 300);
+}
+async function handle(task) {
+    const absolute = node_path_1.default.join(config.root, task.relPath);
+    const started = Date.now();
+    const base = {
+        path: task.relPath,
+        size: task.size,
+        mtime: task.mtime,
+        sha256: null,
+        status: 'ok',
+        errorCode: null,
+        errorDetail: null,
+        durationMs: null
+    };
+    if (task.size > config.maxFileBytes) {
+        // Oversize messages are recorded, not silently dropped, so the corpus report can show how
+        // much of the tail was skipped and why.
+        return { file: { ...base, status: 'oversize', errorCode: 'oversize', durationMs: Date.now() - started }, message: null };
+    }
+    let raw;
+    try {
+        raw = node_fs_1.default.readFileSync(absolute);
+    }
+    catch (err) {
+        return { file: { ...base, status: 'failed', errorCode: 'read_error', errorDetail: errorDetail(err), durationMs: Date.now() - started }, message: null };
+    }
+    const hash = (0, hash_1.sha256)(raw);
+    try {
+        const parsed = await (0, parse_1.parseMessage)(raw, { maxTextPartBytes: config.maxTextPartBytes });
+        const record = (0, analyze_1.analyzeMessage)(parsed, config.salt);
+        return {
+            file: { ...base, sha256: hash, status: 'ok', durationMs: Date.now() - started },
+            message: record
+        };
+    }
+    catch (err) {
+        return {
+            file: { ...base, sha256: hash, status: 'failed', errorCode: 'parse_error', errorDetail: errorDetail(err), durationMs: Date.now() - started },
+            message: null
+        };
+    }
+}
+if (node_worker_threads_1.parentPort) {
+    const port = node_worker_threads_1.parentPort;
+    port.on('message', (envelope) => {
+        handle(envelope.task).then(result => port.postMessage({ id: envelope.id, ok: true, result }), err => port.postMessage({ id: envelope.id, ok: false, error: errorDetail(err) }));
+    });
+}
+//# sourceMappingURL=worker.js.map

package/dist/corpus/worker.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"worker.js","sourceRoot":"","sources":["../../src/corpus/worker.ts"],"names":[],"mappings":";AAAA;;;;;;;GAOG;;;;;AAEH,sDAAyB;AACzB,0DAA6B;AAC7B,6DAA6D;AAC7D,yCAA6C;AAC7C,uCAAsC;AACtC,uCAA+D;AAsB/D,MAAM,MAAM,GAAG,gCAA0B,CAAC;AAE1C,SAAS,WAAW,CAAC,GAAY;IAC7B,IAAI,GAAG,YAAY,KAAK,EAAE,CAAC;QACvB,OAAO,GAAG,GAAG,CAAC,IAAI,KAAK,GAAG,CAAC,OAAO,EAAE,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;IACvD,CAAC;IACD,OAAO,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,CAAC;AACrC,CAAC;AAED,KAAK,UAAU,MAAM,CAAC,IAAc;IAChC,MAAM,QAAQ,GAAG,mBAAI,CAAC,IAAI,CAAC,MAAM,CAAC,IAAI,EAAE,IAAI,CAAC,OAAO,CAAC,CAAC;IACtD,MAAM,OAAO,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IAE3B,MAAM,IAAI,GAAe;QACrB,IAAI,EAAE,IAAI,CAAC,OAAO;QAClB,IAAI,EAAE,IAAI,CAAC,IAAI;QACf,KAAK,EAAE,IAAI,CAAC,KAAK;QACjB,MAAM,EAAE,IAAI;QACZ,MAAM,EAAE,IAAkB;QAC1B,SAAS,EAAE,IAAI;QACf,WAAW,EAAE,IAAI;QACjB,UAAU,EAAE,IAAI;KACnB,CAAC;IAEF,IAAI,IAAI,CAAC,IAAI,GAAG,MAAM,CAAC,YAAY,EAAE,CAAC;QAClC,0FAA0F;QAC1F,wCAAwC;QACxC,OAAO,EAAE,IAAI,EAAE,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,UAAU,EAAE,SAAS,EAAE,UAAU,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC;IAC7H,CAAC;IAED,IAAI,GAAW,CAAC;IAChB,IAAI,CAAC;QACD,GAAG,GAAG,iBAAE,CAAC,YAAY,CAAC,QAAQ,CAAC,CAAC;IACpC,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACX,OAAO,EAAE,IAAI,EAAE,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,YAAY,EAAE,WAAW,EAAE,WAAW,CAAC,GAAG,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,EAAE,EAAE,OAAO,EAAE,IAAI,EAAE,CAAC;IAC5J,CAAC;IAED,MAAM,IAAI,GAAG,IAAA,aAAM,EAAC,GAAG,CAAC,CAAC;IAEzB,IAAI,CAAC;QACD,MAAM,MAAM,GAAG,MAAM,IAAA,oBAAY,EAAC,GAAG,EAAE,EAAE,gBAAgB,EAAE,MAAM,CAAC,gBAAgB,EAAE,CAAC,CAAC;QACtF,MAAM,MAAM,GAAG,IAAA,wBAAc,EAAC,MAAM,EAAE,MAAM,CAAC,IAAI,CAAC,CAAC;QACnD,OAAO;YACH,IAAI,EAAE,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,EAAE;YAC/E,OAAO,EAAE,MAAM;SAClB,CAAC;IACN,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACX,OAAO;YACH,IAAI,EAAE,EAAE,GAAG,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,SAAS,EAAE,aAAa,EAAE,WAAW,EAAE,WAAW,CAAC,GAAG,CAAC,EAAE,UAAU,EAAE,IAAI,CAAC,GAAG,EAAE,GAAG,OAAO,EAAE;YAC5I,OAAO,EAAE,IAAI;SAChB,CAAC;IACN,CAAC;AACL,CAAC;AAED,IAAI,gCAAU,EAAE,CAAC;IACb,MAAM,IAAI,GAAG,gCAAU,CAAC;IACxB,IAAI,CAAC,EAAE,CAAC,SAAS,EAAE,CAAC,QAAwC,EAAE,EAAE;QAC5D,MAAM,CAAC,QAAQ,CAAC,IAAI,CAAC,CAAC,IAAI,CACtB,MAAM,CAAC,EAAE,CAAC,IAAI,CAAC,WAAW,CAAC,EAAE,EAAE,EAAE,QAAQ,CAAC,EAAE,EAAE,EAAE,EAAE,IAAI,EAAE,MAAM,EAAE,CAAC,EACjE,GAAG,CAAC,EAAE,CAAC,IAAI,CAAC,WAAW,CAAC,EAAE,EAAE,EAAE,QAAQ,CAAC,EAAE,EAAE,EAAE,EAAE,KAAK,EAAE,KAAK,EAAE,WAAW,CAAC,GAAG,CAAC,EAAE,CAAC,CACnF,CAAC;IACN,CAAC,CAAC,CAAC;AACP,CAAC"}

package/dist/detect/client.d.ts

@@ -0,0 +1,28 @@
+/**
+ * Client fingerprinting.
+ *
+ * Knowing which client composed a message is evidence, not truth. It is used to prioritise
+ * structural detectors and to stratify sampling, never to decide a label on its own. A message
+ * whose client cannot be identified must still extract correctly, which is why "unknown" is a
+ * first class outcome rather than a failure.
+ *
+ * Signals, in decreasing reliability: Message-ID domain (hard to spoof accidentally, set by the
+ * submitting server), X-Mailer and User-Agent, MIME boundary shape, then HTML structure hints.
+ */
+import { type MessageHeaders } from '../mime/model';
+export type ClientId = 'gmail' | 'gmail-mobile' | 'outlook-web' | 'outlook-desktop' | 'outlook-mobile' | 'apple-mail' | 'apple-mail-mobile' | 'thunderbird' | 'proton' | 'yahoo' | 'zimbra' | 'roundcube' | 'superhuman' | 'hey' | 'front' | 'zendesk' | 'hubspot' | 'github' | 'atlassian' | 'mailchimp' | 'sendgrid' | 'amazon-ses' | 'nodemailer' | 'phpmailer' | 'mutt' | 'webmail-generic' | 'library-generic' | 'unknown';
+export interface ClientFingerprint {
+    client: ClientId;
+    /** Which signal decided it. */
+    source: 'message-id' | 'x-mailer' | 'user-agent' | 'boundary' | 'html' | 'header' | 'none';
+    confidence: number;
+}
+/**
+ * Identify the composing client.
+ *
+ * @param headers decoded message headers
+ * @param html first text/html part, or null. Only the head of the document is inspected.
+ * @param boundary Content-Type boundary parameter of the root part, if any
+ */
+export declare function fingerprintClient(headers: MessageHeaders, html: string | null, boundary: string | null): ClientFingerprint;
+//# sourceMappingURL=client.d.ts.map

package/dist/detect/client.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.d.ts","sourceRoot":"","sources":["../../src/detect/client.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;GAUG;AAEH,OAAO,EAAE,KAAK,cAAc,EAAE,MAAM,eAAe,CAAC;AAEpD,MAAM,MAAM,QAAQ,GACd,OAAO,GACP,cAAc,GACd,aAAa,GACb,iBAAiB,GACjB,gBAAgB,GAChB,YAAY,GACZ,mBAAmB,GACnB,aAAa,GACb,QAAQ,GACR,OAAO,GACP,QAAQ,GACR,WAAW,GACX,YAAY,GACZ,KAAK,GACL,OAAO,GACP,SAAS,GACT,SAAS,GACT,QAAQ,GACR,WAAW,GACX,WAAW,GACX,UAAU,GACV,YAAY,GACZ,YAAY,GACZ,WAAW,GACX,MAAM,GACN,iBAAiB,GACjB,iBAAiB,GACjB,SAAS,CAAC;AAEhB,MAAM,WAAW,iBAAiB;IAC9B,MAAM,EAAE,QAAQ,CAAC;IACjB,+BAA+B;IAC/B,MAAM,EAAE,YAAY,GAAG,UAAU,GAAG,YAAY,GAAG,UAAU,GAAG,MAAM,GAAG,QAAQ,GAAG,MAAM,CAAC;IAC3F,UAAU,EAAE,MAAM,CAAC;CACtB;AA8ED;;;;;;GAMG;AACH,wBAAgB,iBAAiB,CAAC,OAAO,EAAE,cAAc,EAAE,IAAI,EAAE,MAAM,GAAG,IAAI,EAAE,QAAQ,EAAE,MAAM,GAAG,IAAI,GAAG,iBAAiB,CAsC1H"}

package/dist/detect/client.js

@@ -0,0 +1,121 @@
+"use strict";
+/**
+ * Client fingerprinting.
+ *
+ * Knowing which client composed a message is evidence, not truth. It is used to prioritise
+ * structural detectors and to stratify sampling, never to decide a label on its own. A message
+ * whose client cannot be identified must still extract correctly, which is why "unknown" is a
+ * first class outcome rather than a failure.
+ *
+ * Signals, in decreasing reliability: Message-ID domain (hard to spoof accidentally, set by the
+ * submitting server), X-Mailer and User-Agent, MIME boundary shape, then HTML structure hints.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.fingerprintClient = fingerprintClient;
+/** Message-ID domain patterns. Set by the submitting infrastructure, so quite reliable. */
+const MESSAGE_ID_RULES = [
+    { client: 'gmail', regex: /@mail\.gmail\.com>?$/i, confidence: 0.95 },
+    { client: 'outlook-web', regex: /@[a-z0-9.-]*prod\.outlook\.com>?$/i, confidence: 0.9 },
+    { client: 'outlook-web', regex: /@[a-z0-9.-]*namprd[0-9]+\.prod\.outlook\.com>?$/i, confidence: 0.92 },
+    { client: 'yahoo', regex: /@[a-z0-9.-]*(yahoo|ymail)\.com>?$/i, confidence: 0.8 },
+    { client: 'proton', regex: /@(protonmail|proton\.me)/i, confidence: 0.9 },
+    { client: 'github', regex: /@github\.com>?$/i, confidence: 0.95 },
+    { client: 'atlassian', regex: /@(atlassian\.net|jira)/i, confidence: 0.85 },
+    { client: 'zendesk', regex: /@[a-z0-9.-]*zendesk\.com>?$/i, confidence: 0.9 },
+    { client: 'superhuman', regex: /@superhuman\.com>?$/i, confidence: 0.9 },
+    { client: 'hey', regex: /@hey\.com>?$/i, confidence: 0.85 },
+    { client: 'front', regex: /@(frontapp|front)\.com>?$/i, confidence: 0.85 },
+    { client: 'mailchimp', regex: /@mail[0-9]*\.(mcsv|mailchimpapp)\.net>?$/i, confidence: 0.9 },
+    { client: 'sendgrid', regex: /@(sendgrid|sendgrid\.net)/i, confidence: 0.85 },
+    { client: 'amazon-ses', regex: /@(email\.amazonses\.com|[a-z0-9.-]*amazonses\.com)>?$/i, confidence: 0.9 }
+];
+/** X-Mailer and User-Agent values. Self declared, so slightly weaker. */
+const AGENT_RULES = [
+    { client: 'apple-mail', regex: /\bApple\s?Mail\b/i, confidence: 0.9 },
+    { client: 'apple-mail-mobile', regex: /\biPhone Mail\b|\biPad Mail\b/i, confidence: 0.9 },
+    { client: 'outlook-desktop', regex: /Microsoft (Outlook|Office Outlook)(?! Express)/i, confidence: 0.9 },
+    { client: 'outlook-desktop', regex: /Microsoft Outlook Express/i, confidence: 0.9 },
+    { client: 'outlook-mobile', regex: /Outlook for (iOS|Android)|ms-outlook-mobile/i, confidence: 0.9 },
+    { client: 'thunderbird', regex: /Thunderbird|Icedove/i, confidence: 0.95 },
+    { client: 'mutt', regex: /\bMutt\b|\bNeoMutt\b/i, confidence: 0.95 },
+    { client: 'roundcube', regex: /Roundcube/i, confidence: 0.95 },
+    { client: 'zimbra', regex: /Zimbra/i, confidence: 0.95 },
+    { client: 'proton', regex: /ProtonMail|Proton Mail/i, confidence: 0.9 },
+    { client: 'nodemailer', regex: /nodemailer/i, confidence: 0.95 },
+    { client: 'phpmailer', regex: /PHPMailer|phpmailer/i, confidence: 0.95 },
+    { client: 'webmail-generic', regex: /SquirrelMail|Horde|Webmail|MessagingEngine/i, confidence: 0.7 },
+    { client: 'library-generic', regex: /JavaMail|Mailer|SwiftMailer|Sendmail|python|Zend_Mail|CakePHP/i, confidence: 0.6 }
+];
+/** MIME boundary shapes that some clients use consistently. */
+const BOUNDARY_RULES = [
+    { client: 'apple-mail', regex: /Apple-Mail[=-]/i, confidence: 0.85 },
+    { client: 'library-generic', regex: /----=_Part_[0-9]+_[0-9]+/, confidence: 0.6 },
+    { client: 'outlook-desktop', regex: /----_=_NextPart_/i, confidence: 0.7 }
+];
+/** HTML structure hints. Weakest signal, used only when nothing else matched. */
+const HTML_RULES = [
+    { client: 'gmail', regex: /class="?gmail_(quote|attr|signature)/i, confidence: 0.8 },
+    { client: 'outlook-web', regex: /id="?(divRplyFwdMsg|appendonsend)/i, confidence: 0.8 },
+    { client: 'outlook-web', regex: /class="?[^"]*elementToProof/i, confidence: 0.7 },
+    { client: 'outlook-desktop', regex: /class="?MsoNormal|WordSection1/i, confidence: 0.6 },
+    { client: 'thunderbird', regex: /moz-cite-prefix|moz-signature/i, confidence: 0.85 },
+    { client: 'apple-mail', regex: /Apple-interchange-newline|AppleMailSignature/i, confidence: 0.75 },
+    { client: 'proton', regex: /protonmail_quote/i, confidence: 0.85 },
+    { client: 'yahoo', regex: /yahoo_quoted/i, confidence: 0.85 },
+    { client: 'zimbra', regex: /id="?zwchr/i, confidence: 0.8 },
+    { client: 'hubspot', regex: /hubspot/i, confidence: 0.6 },
+    { client: 'zendesk', regex: /zendesk/i, confidence: 0.6 }
+];
+function match(rules, value) {
+    if (!value) {
+        return null;
+    }
+    for (const rule of rules) {
+        if (rule.regex.test(value)) {
+            return rule;
+        }
+    }
+    return null;
+}
+/**
+ * Identify the composing client.
+ *
+ * @param headers decoded message headers
+ * @param html first text/html part, or null. Only the head of the document is inspected.
+ * @param boundary Content-Type boundary parameter of the root part, if any
+ */
+function fingerprintClient(headers, html, boundary) {
+    const byMessageId = match(MESSAGE_ID_RULES, headers.messageId);
+    if (byMessageId) {
+        // A mobile Outlook or Gmail message still carries the provider Message-ID, so refine with
+        // the self declared agent when it disagrees on the same provider family.
+        const agent = `${headers.xMailer || ''} ${headers.userAgent || ''}`;
+        if (byMessageId.client === 'outlook-web' && /Outlook for (iOS|Android)/i.test(agent)) {
+            return { client: 'outlook-mobile', source: 'user-agent', confidence: 0.9 };
+        }
+        if (byMessageId.client === 'gmail' && /Android|iPhone|iPad/i.test(agent)) {
+            return { client: 'gmail-mobile', source: 'user-agent', confidence: 0.8 };
+        }
+        return { client: byMessageId.client, source: 'message-id', confidence: byMessageId.confidence };
+    }
+    const byMailer = match(AGENT_RULES, headers.xMailer);
+    if (byMailer) {
+        return { client: byMailer.client, source: 'x-mailer', confidence: byMailer.confidence };
+    }
+    const byAgent = match(AGENT_RULES, headers.userAgent);
+    if (byAgent) {
+        return { client: byAgent.client, source: 'user-agent', confidence: byAgent.confidence };
+    }
+    const byBoundary = match(BOUNDARY_RULES, boundary);
+    if (byBoundary) {
+        return { client: byBoundary.client, source: 'boundary', confidence: byBoundary.confidence };
+    }
+    // Only the head of the document is scanned. Client markers appear early, and a full scan of a
+    // multi megabyte newsletter would dominate the cost of the whole pipeline.
+    const byHtml = match(HTML_RULES, html ? html.slice(0, 65536) : null);
+    if (byHtml) {
+        return { client: byHtml.client, source: 'html', confidence: byHtml.confidence };
+    }
+    return { client: 'unknown', source: 'none', confidence: 0 };
+}
+//# sourceMappingURL=client.js.map

package/dist/detect/client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"client.js","sourceRoot":"","sources":["../../src/detect/client.ts"],"names":[],"mappings":";AAAA;;;;;;;;;;GAUG;;AA4HH,8CAsCC;AAnHD,2FAA2F;AAC3F,MAAM,gBAAgB,GAAW;IAC7B,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,uBAAuB,EAAE,UAAU,EAAE,IAAI,EAAE;IACrE,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,oCAAoC,EAAE,UAAU,EAAE,GAAG,EAAE;IACvF,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,kDAAkD,EAAE,UAAU,EAAE,IAAI,EAAE;IACtG,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,oCAAoC,EAAE,UAAU,EAAE,GAAG,EAAE;IACjF,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,2BAA2B,EAAE,UAAU,EAAE,GAAG,EAAE;IACzE,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,kBAAkB,EAAE,UAAU,EAAE,IAAI,EAAE;IACjE,EAAE,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,yBAAyB,EAAE,UAAU,EAAE,IAAI,EAAE;IAC3E,EAAE,MAAM,EAAE,SAAS,EAAE,KAAK,EAAE,8BAA8B,EAAE,UAAU,EAAE,GAAG,EAAE;IAC7E,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,sBAAsB,EAAE,UAAU,EAAE,GAAG,EAAE;IACxE,EAAE,MAAM,EAAE,KAAK,EAAE,KAAK,EAAE,eAAe,EAAE,UAAU,EAAE,IAAI,EAAE;IAC3D,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,4BAA4B,EAAE,UAAU,EAAE,IAAI,EAAE;IAC1E,EAAE,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,2CAA2C,EAAE,UAAU,EAAE,GAAG,EAAE;IAC5F,EAAE,MAAM,EAAE,UAAU,EAAE,KAAK,EAAE,4BAA4B,EAAE,UAAU,EAAE,IAAI,EAAE;IAC7E,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,wDAAwD,EAAE,UAAU,EAAE,GAAG,EAAE;CAC7G,CAAC;AAEF,yEAAyE;AACzE,MAAM,WAAW,GAAW;IACxB,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,mBAAmB,EAAE,UAAU,EAAE,GAAG,EAAE;IACrE,EAAE,MAAM,EAAE,mBAAmB,EAAE,KAAK,EAAE,gCAAgC,EAAE,UAAU,EAAE,GAAG,EAAE;IACzF,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,iDAAiD,EAAE,UAAU,EAAE,GAAG,EAAE;IACxG,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,4BAA4B,EAAE,UAAU,EAAE,GAAG,EAAE;IACnF,EAAE,MAAM,EAAE,gBAAgB,EAAE,KAAK,EAAE,8CAA8C,EAAE,UAAU,EAAE,GAAG,EAAE;IACpG,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,sBAAsB,EAAE,UAAU,EAAE,IAAI,EAAE;IAC1E,EAAE,MAAM,EAAE,MAAM,EAAE,KAAK,EAAE,uBAAuB,EAAE,UAAU,EAAE,IAAI,EAAE;IACpE,EAAE,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,YAAY,EAAE,UAAU,EAAE,IAAI,EAAE;IAC9D,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,SAAS,EAAE,UAAU,EAAE,IAAI,EAAE;IACxD,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,yBAAyB,EAAE,UAAU,EAAE,GAAG,EAAE;IACvE,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,aAAa,EAAE,UAAU,EAAE,IAAI,EAAE;IAChE,EAAE,MAAM,EAAE,WAAW,EAAE,KAAK,EAAE,sBAAsB,EAAE,UAAU,EAAE,IAAI,EAAE;IACxE,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,6CAA6C,EAAE,UAAU,EAAE,GAAG,EAAE;IACpG,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,gEAAgE,EAAE,UAAU,EAAE,GAAG,EAAE;CAC1H,CAAC;AAEF,+DAA+D;AAC/D,MAAM,cAAc,GAAW;IAC3B,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,iBAAiB,EAAE,UAAU,EAAE,IAAI,EAAE;IACpE,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,0BAA0B,EAAE,UAAU,EAAE,GAAG,EAAE;IACjF,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,mBAAmB,EAAE,UAAU,EAAE,GAAG,EAAE;CAC7E,CAAC;AAEF,iFAAiF;AACjF,MAAM,UAAU,GAAW;IACvB,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,uCAAuC,EAAE,UAAU,EAAE,GAAG,EAAE;IACpF,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,oCAAoC,EAAE,UAAU,EAAE,GAAG,EAAE;IACvF,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,8BAA8B,EAAE,UAAU,EAAE,GAAG,EAAE;IACjF,EAAE,MAAM,EAAE,iBAAiB,EAAE,KAAK,EAAE,iCAAiC,EAAE,UAAU,EAAE,GAAG,EAAE;IACxF,EAAE,MAAM,EAAE,aAAa,EAAE,KAAK,EAAE,gCAAgC,EAAE,UAAU,EAAE,IAAI,EAAE;IACpF,EAAE,MAAM,EAAE,YAAY,EAAE,KAAK,EAAE,+CAA+C,EAAE,UAAU,EAAE,IAAI,EAAE;IAClG,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,mBAAmB,EAAE,UAAU,EAAE,IAAI,EAAE;IAClE,EAAE,MAAM,EAAE,OAAO,EAAE,KAAK,EAAE,eAAe,EAAE,UAAU,EAAE,IAAI,EAAE;IAC7D,EAAE,MAAM,EAAE,QAAQ,EAAE,KAAK,EAAE,aAAa,EAAE,UAAU,EAAE,GAAG,EAAE;IAC3D,EAAE,MAAM,EAAE,SAAS,EAAE,KAAK,EAAE,UAAU,EAAE,UAAU,EAAE,GAAG,EAAE;IACzD,EAAE,MAAM,EAAE,SAAS,EAAE,KAAK,EAAE,UAAU,EAAE,UAAU,EAAE,GAAG,EAAE;CAC5D,CAAC;AAEF,SAAS,KAAK,CAAC,KAAsB,EAAE,KAAgC;IACnE,IAAI,CAAC,KAAK,EAAE,CAAC;QACT,OAAO,IAAI,CAAC;IAChB,CAAC;IACD,KAAK,MAAM,IAAI,IAAI,KAAK,EAAE,CAAC;QACvB,IAAI,IAAI,CAAC,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YACzB,OAAO,IAAI,CAAC;QAChB,CAAC;IACL,CAAC;IACD,OAAO,IAAI,CAAC;AAChB,CAAC;AAED;;;;;;GAMG;AACH,SAAgB,iBAAiB,CAAC,OAAuB,EAAE,IAAmB,EAAE,QAAuB;IACnG,MAAM,WAAW,GAAG,KAAK,CAAC,gBAAgB,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC;IAC/D,IAAI,WAAW,EAAE,CAAC;QACd,0FAA0F;QAC1F,yEAAyE;QACzE,MAAM,KAAK,GAAG,GAAG,OAAO,CAAC,OAAO,IAAI,EAAE,IAAI,OAAO,CAAC,SAAS,IAAI,EAAE,EAAE,CAAC;QACpE,IAAI,WAAW,CAAC,MAAM,KAAK,aAAa,IAAI,4BAA4B,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YACnF,OAAO,EAAE,MAAM,EAAE,gBAAgB,EAAE,MAAM,EAAE,YAAY,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC;QAC/E,CAAC;QACD,IAAI,WAAW,CAAC,MAAM,KAAK,OAAO,IAAI,sBAAsB,CAAC,IAAI,CAAC,KAAK,CAAC,EAAE,CAAC;YACvE,OAAO,EAAE,MAAM,EAAE,cAAc,EAAE,MAAM,EAAE,YAAY,EAAE,UAAU,EAAE,GAAG,EAAE,CAAC;QAC7E,CAAC;QACD,OAAO,EAAE,MAAM,EAAE,WAAW,CAAC,MAAM,EAAE,MAAM,EAAE,YAAY,EAAE,UAAU,EAAE,WAAW,CAAC,UAAU,EAAE,CAAC;IACpG,CAAC;IAED,MAAM,QAAQ,GAAG,KAAK,CAAC,WAAW,EAAE,OAAO,CAAC,OAAO,CAAC,CAAC;IACrD,IAAI,QAAQ,EAAE,CAAC;QACX,OAAO,EAAE,MAAM,EAAE,QAAQ,CAAC,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,UAAU,EAAE,QAAQ,CAAC,UAAU,EAAE,CAAC;IAC5F,CAAC;IAED,MAAM,OAAO,GAAG,KAAK,CAAC,WAAW,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC;IACtD,IAAI,OAAO,EAAE,CAAC;QACV,OAAO,EAAE,MAAM,EAAE,OAAO,CAAC,MAAM,EAAE,MAAM,EAAE,YAAY,EAAE,UAAU,EAAE,OAAO,CAAC,UAAU,EAAE,CAAC;IAC5F,CAAC;IAED,MAAM,UAAU,GAAG,KAAK,CAAC,cAAc,EAAE,QAAQ,CAAC,CAAC;IACnD,IAAI,UAAU,EAAE,CAAC;QACb,OAAO,EAAE,MAAM,EAAE,UAAU,CAAC,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,UAAU,EAAE,UAAU,CAAC,UAAU,EAAE,CAAC;IAChG,CAAC;IAED,8FAA8F;IAC9F,2EAA2E;IAC3E,MAAM,MAAM,GAAG,KAAK,CAAC,UAAU,EAAE,IAAI,CAAC,CAAC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,EAAE,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;IACrE,IAAI,MAAM,EAAE,CAAC;QACT,OAAO,EAAE,MAAM,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,MAAM,CAAC,UAAU,EAAE,CAAC;IACpF,CAAC;IAED,OAAO,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,EAAE,MAAM,EAAE,UAAU,EAAE,CAAC,EAAE,CAAC;AAChE,CAAC"}

package/dist/detect/disclaimer.d.ts

@@ -0,0 +1,12 @@
+/**
+ * Legal and organisational footer detection.
+ *
+ * A disclaimer is appended by a gateway or a policy engine rather than typed by the sender, so it
+ * is collapsed by default. Detection requires both a phrase match and a trailing position: a
+ * discussion that happens to mention confidentiality in the middle of a thread is ordinary content
+ * and must stay visible.
+ */
+import { type Evidence } from '../segment/types';
+import { type BlockFeatures } from '../segment/types';
+export declare function detectDisclaimer(text: string, features: BlockFeatures): Evidence[];
+//# sourceMappingURL=disclaimer.d.ts.map

package/dist/detect/disclaimer.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"disclaimer.d.ts","sourceRoot":"","sources":["../../src/detect/disclaimer.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAGH,OAAO,EAAE,KAAK,QAAQ,EAAE,MAAM,kBAAkB,CAAC;AACjD,OAAO,EAAE,KAAK,aAAa,EAAE,MAAM,kBAAkB,CAAC;AAKtD,wBAAgB,gBAAgB,CAAC,IAAI,EAAE,MAAM,EAAE,QAAQ,EAAE,aAAa,GAAG,QAAQ,EAAE,CAYlF"}

package/dist/detect/disclaimer.js

@@ -0,0 +1,28 @@
+"use strict";
+/**
+ * Legal and organisational footer detection.
+ *
+ * A disclaimer is appended by a gateway or a policy engine rather than typed by the sender, so it
+ * is collapsed by default. Detection requires both a phrase match and a trailing position: a
+ * discussion that happens to mention confidentiality in the middle of a thread is ordinary content
+ * and must stay visible.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.detectDisclaimer = detectDisclaimer;
+const patterns_1 = require("./patterns");
+/** Disclaimers live at the end. Anything earlier is discussion about a disclaimer. */
+const MIN_POSITION = 0.4;
+function detectDisclaimer(text, features) {
+    if (features.position < MIN_POSITION || features.isAllBlank) {
+        return [];
+    }
+    const patterns = (0, patterns_1.loadPatterns)();
+    const found = [];
+    for (const pattern of patterns.disclaimerPhrases) {
+        if (pattern.regex.test(text)) {
+            found.push({ id: `disclaimer:${pattern.id}`, kind: 'disclaimer', strength: pattern.strength });
+        }
+    }
+    return found;
+}
+//# sourceMappingURL=disclaimer.js.map

package/dist/detect/disclaimer.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"disclaimer.js","sourceRoot":"","sources":["../../src/detect/disclaimer.ts"],"names":[],"mappings":";AAAA;;;;;;;GAOG;;AASH,4CAYC;AAnBD,yCAA0C;AAI1C,sFAAsF;AACtF,MAAM,YAAY,GAAG,GAAG,CAAC;AAEzB,SAAgB,gBAAgB,CAAC,IAAY,EAAE,QAAuB;IAClE,IAAI,QAAQ,CAAC,QAAQ,GAAG,YAAY,IAAI,QAAQ,CAAC,UAAU,EAAE,CAAC;QAC1D,OAAO,EAAE,CAAC;IACd,CAAC;IACD,MAAM,QAAQ,GAAG,IAAA,uBAAY,GAAE,CAAC;IAChC,MAAM,KAAK,GAAe,EAAE,CAAC;IAC7B,KAAK,MAAM,OAAO,IAAI,QAAQ,CAAC,iBAAiB,EAAE,CAAC;QAC/C,IAAI,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CAAC;YAC3B,KAAK,CAAC,IAAI,CAAC,EAAE,EAAE,EAAE,cAAc,OAAO,CAAC,EAAE,EAAE,EAAE,IAAI,EAAE,YAAY,EAAE,QAAQ,EAAE,OAAO,CAAC,QAAQ,EAAE,CAAC,CAAC;QACnG,CAAC;IACL,CAAC;IACD,OAAO,KAAK,CAAC;AACjB,CAAC"}

package/dist/detect/forward-header.d.ts

@@ -0,0 +1,10 @@
+/**
+ * Forward marker detection.
+ *
+ * Forwarded content was included on purpose, so a forward anchor opens a region that is collapsed
+ * and kept, never hidden. That difference in consequence is why forward markers are detected
+ * separately from reply markers rather than being folded into one "quote anchor" concept.
+ */
+import { type Evidence } from '../segment/types';
+export declare function detectForwardMarker(line: string): Evidence | null;
+//# sourceMappingURL=forward-header.d.ts.map

package/dist/detect/forward-header.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"forward-header.d.ts","sourceRoot":"","sources":["../../src/detect/forward-header.ts"],"names":[],"mappings":"AAAA;;;;;;GAMG;AAGH,OAAO,EAAE,KAAK,QAAQ,EAAE,MAAM,kBAAkB,CAAC;AAEjD,wBAAgB,mBAAmB,CAAC,IAAI,EAAE,MAAM,GAAG,QAAQ,GAAG,IAAI,CAsBjE"}

package/dist/detect/forward-header.js

@@ -0,0 +1,35 @@
+"use strict";
+/**
+ * Forward marker detection.
+ *
+ * Forwarded content was included on purpose, so a forward anchor opens a region that is collapsed
+ * and kept, never hidden. That difference in consequence is why forward markers are detected
+ * separately from reply markers rather than being folded into one "quote anchor" concept.
+ */
+Object.defineProperty(exports, "__esModule", { value: true });
+exports.detectForwardMarker = detectForwardMarker;
+const patterns_1 = require("./patterns");
+function detectForwardMarker(line) {
+    const trimmed = line.trim();
+    if (!trimmed || trimmed.length > 300) {
+        return null;
+    }
+    const patterns = (0, patterns_1.loadPatterns)();
+    let best = null;
+    for (const pattern of patterns.forwardHeaders) {
+        // The bare "From:" pattern is only useful as part of a header block, which
+        // detectQuotedHeaderBlock handles. Treating it as a forward anchor on its own would fire on
+        // ordinary prose.
+        if (pattern.id === 'fyi_forward_header_block') {
+            continue;
+        }
+        if (!pattern.regex.test(trimmed)) {
+            continue;
+        }
+        if (!best || pattern.strength > best.strength) {
+            best = { id: `forward:${pattern.id}`, kind: 'forward-anchor', strength: pattern.strength };
+        }
+    }
+    return best;
+}
+//# sourceMappingURL=forward-header.js.map

package/dist/detect/forward-header.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"forward-header.js","sourceRoot":"","sources":["../../src/detect/forward-header.ts"],"names":[],"mappings":";AAAA;;;;;;GAMG;;AAKH,kDAsBC;AAzBD,yCAA0C;AAG1C,SAAgB,mBAAmB,CAAC,IAAY;IAC5C,MAAM,OAAO,GAAG,IAAI,CAAC,IAAI,EAAE,CAAC;IAC5B,IAAI,CAAC,OAAO,IAAI,OAAO,CAAC,MAAM,GAAG,GAAG,EAAE,CAAC;QACnC,OAAO,IAAI,CAAC;IAChB,CAAC;IACD,MAAM,QAAQ,GAAG,IAAA,uBAAY,GAAE,CAAC;IAChC,IAAI,IAAI,GAAoB,IAAI,CAAC;IACjC,KAAK,MAAM,OAAO,IAAI,QAAQ,CAAC,cAAc,EAAE,CAAC;QAC5C,2EAA2E;QAC3E,4FAA4F;QAC5F,kBAAkB;QAClB,IAAI,OAAO,CAAC,EAAE,KAAK,0BAA0B,EAAE,CAAC;YAC5C,SAAS;QACb,CAAC;QACD,IAAI,CAAC,OAAO,CAAC,KAAK,CAAC,IAAI,CAAC,OAAO,CAAC,EAAE,CAAC;YAC/B,SAAS;QACb,CAAC;QACD,IAAI,CAAC,IAAI,IAAI,OAAO,CAAC,QAAQ,GAAG,IAAI,CAAC,QAAQ,EAAE,CAAC;YAC5C,IAAI,GAAG,EAAE,EAAE,EAAE,WAAW,OAAO,CAAC,EAAE,EAAE,EAAE,IAAI,EAAE,gBAAgB,EAAE,QAAQ,EAAE,OAAO,CAAC,QAAQ,EAAE,CAAC;QAC/F,CAAC;IACL,CAAC;IACD,OAAO,IAAI,CAAC;AAChB,CAAC"}