docs: fix leftover port 5000 in README start instructions

Aligns the example curl-and-start snippet with the 8085 default that
run.py, the systemd unit and the architecture docs already use.

🤖 Generated with Codebuff
Co-Authored-By: Codebuff <noreply@codebuff.com>

.gitignore

@@ -174,3 +174,10 @@ cython_debug/
 # PyPI configuration file
 .pypirc
 
+
+# project
+venv/
+instance/
+certs/
+__pycache__/
+*.pyc

README.md

@@ -1,3 +1,70 @@
-# crushdb
+# crushdb – verschlüsselte Datensätze zwischen Personen
 
-a secure database for saving transaktions betwen persons
+Sichere Datenbank für Transaktionen zwischen zwei Personen
+(Person A, Person B, Timestamp-Start, Timestamp-Ende, Aktion).
+
+Alle Datensätze werden **im Browser** verschlüsselt
+(XChaCha20-Poly1305, Schlüsselableitung per Argon2id aus der Passphrase).
+Der Server sieht ausschließlich Chiffretext – auch bei vollständiger
+Serverkompromittierung bleiben die Inhalte ohne Passphrase vertraulich.
+
+## Starten (HTTP; TLS über Reverse Proxy)
+
+```bash
+python3 -m venv venv
+./venv/bin/pip install -r requirements.txt
+./venv/bin/python run.py
+# → http://127.0.0.1:8085
+```
+
+Die App lauscht auf dem lokalen HTTP-Port (default `127.0.0.1:8085`,
+überschreibbar via `HOST`/`PORT`). Ein vertrauenswürdiger Reverse Proxy
+(nginx, caddy, …) beendet HTTPS und reicht intern an die App weiter.
+Die App selbst muss von außen nicht erreichbar sein.
+
+### Per systemd laufen lassen
+
+Ein fertiges Unit-File liegt in `deploy/crushdb.service` (Reinheit:
+nur `instance/` schreibbar, keine Capabilities, Restart on failure).
+Nach dem Kopieren nach `/etc/systemd/system/` und `daemon-reload`
+startet `systemctl enable --now crushdb` die App dauerhaft.
+
+## Tests
+
+```bash
+./venv/bin/python -m pytest backend/test_integrity.py -v   # Integrität/Replay
+node .e2e.mjs  # optional: End-to-End (siehe unten)
+```
+
+## Architektur
+
+| Komponente |Datei | Aufgabe |
+|---|---|---|
+| Speicher | `backend/storage.py` | SQLite, speichert nur Chiffretext (uuid, ciphertext, nonce, salt, version) |
+| API | `backend/app.py` | Flask REST API, Validierung nur von Form/Größe, Versions-Konfliktprüfung |
+| Krypto (Client) | `backend/static/crypto.js` | Argon2id-KDF + XChaCha20-Poly1305, uuid+salt als AEAD-AAD gebunden |
+| UI | `backend/static/ui.js` + `index.html`/`style.css` | Tabelle, Erstellen, Bearbeiten, Löschen; Passphrase bleibt im RAM |
+| Server | `run.py` | HTTP intern hinter Reverse Proxy, DB-Init |
+
+## Sicherheitsmodell
+
+- Passphrase → Argon2id (t=2, m=16 MiB) → 256-Bit-Sitzungsschlüssel,
+  ausschließlich im Browser/im Arbeitsspeicher; niemals am Server.
+- Jeder Datensatz: eigene zufällige 16-Byte-Salt + 24-Byte-Nonce.
+  uuid + salt sind als AEAD-Associated-Data gebunden – Vertauschen oder
+  Um-Salten wird beim Entschlüsseln sicher erkannt.
+- Monotoner Versionszähler pro Datensatz: veraltete Updates/Deletes
+  (Replay alter Stände) → `409`, Duplikate → `409`.
+- Klartexte werden clientseitig nie persistiert; beim Sperren/Beenden
+  wird der Schlüssel verworfen (`wipeKey`).
+- HTTPS wird außen vom vertrauenswürdigen **Reverse Proxy** beendet
+  (nginx/caddy/…); die App lauscht nur intern auf HTTP.
+
+## Bekannte Grenzen
+
+- Datenverlust (Löschen/Überschreiben) durch Angreifer bleibt möglich –
+  Vertraulichkeit ist geschützt, Verfügbarkeit nicht.
+- Backups der SQLite-Datei enthalten ausschließlich Chiffretext.
+- Backups des Passphrasen-Verlusts = Datenverlust (by design).
+
+Siehe `spezification.txt` für die Anforderungen.

backend/app.py

@@ -0,0 +1,155 @@
+"""Flask application exposing the records API.
+
+The server is a dumb ciphertext store: it never sees plaintext or the
+key. Every record body is opaque base64. Requests are validated only
+for shape and size, never for content.
+
+Security properties:
+  - No plaintext is logged, stored, or echoed.
+  - Version counters are enforced server-side to detect replays.
+  - HTTPS is handled by ssl context in run.py (TLS for dev).
+"""
+
+import base64
+import binascii
+
+from flask import Flask, jsonify, request, send_from_directory
+
+from . import storage
+
+MAX_CIPHERTEXT_B64 = 32 * 1024  # plenty for one record, stops abuse
+MAX_SALT_NONCE_B64 = 512
+
+app = Flask(__name__, static_folder=None)
+
+
+# ---------- basic hardening ----------
+
+@app.after_request
+def _security_headers(resp):
+    resp.headers["X-Content-Type-Options"] = "nosniff"
+    resp.headers["X-Frame-Options"] = "DENY"
+    resp.headers["Referrer-Policy"] = "no-referrer"
+    resp.headers["Cache-Control"] = "no-store"
+    return resp
+
+
+def _b64_field(data, field, max_len, required=True):
+    """Validate and decode a base64 field from the JSON body. Returns bytes or None."""
+    if field not in data:
+        if required:
+            raise ValueError(f"missing field: {field}")
+        return None
+    raw = data[field]
+    if not isinstance(raw, str) or not raw or len(raw) > max_len:
+        raise ValueError(f"invalid field: {field}")
+    try:
+        return base64.b64decode(raw, validate=True)
+    except (binascii.Error, ValueError):
+        raise ValueError(f"invalid base64 in field: {field}")
+
+
+def _record_json(rec):
+    return {
+        "uuid": rec["uuid"],
+        "ciphertext": base64.b64encode(rec["ciphertext"]).decode(),
+        "nonce": base64.b64encode(rec["nonce"]).decode(),
+        "salt": base64.b64encode(rec["salt"]).decode(),
+        "version": rec["version"],
+        "created_at": rec["created_at"],
+        "updated_at": rec["updated_at"],
+    }
+
+
+# ---------- static UI ----------
+
+@app.route("/")
+def index():
+    return send_from_directory("static", "index.html")
+
+
+@app.route("/static/<path:filename>")
+def static_files(filename):
+    return send_from_directory("static", filename)
+
+
+# ---------- API ----------
+
+@app.get("/api/records")
+def api_list():
+    return jsonify([_record_json(r) for r in storage.list_records()])
+
+
+@app.get("/api/records/<uuid>")
+def api_get(uuid):
+    rec = storage.get_record(uuid)
+    if rec is None:
+        return jsonify({"error": "not found"}), 404
+    return jsonify(_record_json(rec))
+
+
+@app.post("/api/records")
+def api_create():
+    data = request.get_json(silent=True)
+    if not isinstance(data, dict):
+        return jsonify({"error": "invalid JSON body"}), 400
+    try:
+        uuid = data.get("uuid")
+        if not isinstance(uuid, str) or not uuid or len(uuid) > 64:
+            raise ValueError("invalid field: uuid")
+        ct = _b64_field(data, "ciphertext", MAX_CIPHERTEXT_B64)
+        nonce = _b64_field(data, "nonce", MAX_SALT_NONCE_B64)
+        salt = _b64_field(data, "salt", MAX_SALT_NONCE_B64)
+    except ValueError as e:
+        return jsonify({"error": str(e)}), 400
+
+    if storage.get_record(uuid) is not None:
+        return jsonify({"error": "record already exists"}), 409
+
+    rec = storage.create_record(uuid, ct, nonce, salt)
+    return jsonify(_record_json(rec)), 201
+
+
+@app.put("/api/records/<uuid>")
+def api_update(uuid):
+    data = request.get_json(silent=True)
+    if not isinstance(data, dict):
+        return jsonify({"error": "invalid JSON body"}), 400
+    try:
+        version = data.get("version")
+        if not isinstance(version, int) or version < 1:
+            raise ValueError("invalid field: version")
+        ct = _b64_field(data, "ciphertext", MAX_CIPHERTEXT_B64)
+        nonce = _b64_field(data, "nonce", MAX_SALT_NONCE_B64)
+        salt = _b64_field(data, "salt", MAX_SALT_NONCE_B64)
+    except ValueError as e:
+        return jsonify({"error": str(e)}), 400
+
+    rec = storage.update_record(uuid, ct, nonce, salt, expected_version=version)
+    if rec is None:
+        # 409: version mismatch (concurrent edit or replay) — do not
+        # distinguish from "not found" to avoid leaking existence.
+        return jsonify({"error": "version conflict or not found"}), 409
+    return jsonify(_record_json(rec))
+
+
+@app.delete("/api/records/<uuid>")
+def api_delete(uuid):
+    data = request.get_json(silent=True) or {}
+    version = data.get("version")
+    if not isinstance(version, int) or version < 1:
+        return jsonify({"error": "invalid field: version"}), 400
+    if not storage.delete_record(uuid, expected_version=version):
+        return jsonify({"error": "version conflict or not found"}), 409
+    return "", 204
+
+
+@app.errorhandler(404)
+def not_found(_e):
+    return jsonify({"error": "not found"}), 404
+
+
+@app.errorhandler(500)
+def server_error(_e):
+    # Never leak stack traces or data.
+    return jsonify({"error": "internal server error"}), 500

backend/static/crypto.js

@@ -0,0 +1,155 @@
+/**
+ * crypto.js – client-side encryption module (vanilla ES module).
+ *
+ * All cryptography happens in the browser. The passphrase and derived
+ * key never leave this module's memory and are never sent anywhere.
+ *
+ * Design:
+ *  - Key derivation: Argon2id (via @noble/hashes), one key per session,
+ *    derived once from the passphrase the user enters on opening the
+ *    site. The salt used for derivation is *not* the per-record salt;
+ *    it is a random domain salt stored alongside each record.
+ *  - Cipher: XChaCha20-Poly1305 (AEAD) with a fresh random 24-byte nonce
+ *    per record (via @noble/ciphers).
+ *  - Each record carries its own random salt. That salt is bound into
+ *    the AEAD header (as an extra attribute) so re-deriving the key for
+ *    a record with a modified salt reliably fails. TLS-like: header is
+ *    encrypted alongside the payload body.
+ *  - AAD (associated data) binds the record's uuid into the AEAD header,
+ *    so a ciphertext cannot be swapped between records undetected.
+ *
+ * Everything is in-memory only: this file has no persistence, and the
+ * session key must be wiped explicitly by the caller (wipeKey).
+ */
+
+import { xchacha20poly1305 } from "./vendor/noble-ciphers/chacha.js";
+import { randomBytes } from "./vendor/noble-ciphers/utils.js";
+import { argon2id } from "./vendor/noble-hashes/argon2.js";
+
+const ARGON2_PARAMS = {
+  // Moderate defaults so unlocking stays usable on typical hardware in-JS.
+  t: 2, // passes
+  m: 16384, // memory, KiB (16 MiB)
+  p: 1, // parallelism
+  dkLen: 32, // 256-bit key for XChaCha20
+};
+
+const REC_SALT_LEN = 16; // per-record salt is independent from KDF salt? No:
+                          // the KDF salt IS the per-record salt (see below).
+
+// ---------- base64 / utf8 helpers ----------
+
+export function b64encode(bytes) {
+  let bin = "";
+  for (const b of bytes) bin += String.fromCharCode(b);
+  return btoa(bin);
+}
+
+export function b64decode(str) {
+  const bin = atob(str);
+  const out = new Uint8Array(bin.length);
+  for (let i = 0; i < bin.length; i++) out[i] = bin.charCodeAt(i);
+  return out;
+}
+
+export function utf8encode(str) {
+  return new TextEncoder().encode(str);
+}
+
+export function utf8decode(bytes) {
+  return new TextDecoder().decode(bytes);
+}
+
+export function randomUuid() {
+  return crypto.randomUUID();
+}
+
+// ---------- key derivation ----------
+
+/**
+ * Derive a 32-byte key from a passphrase + salt with Argon2id.
+ * Runs entirely in the browser; the passphrase never leaves it.
+ */
+export function deriveKey(passphrase, saltBytes) {
+  if (typeof passphrase !== "string" || passphrase.length === 0)
+    throw new Error("invalid passphrase");
+  if (!(saltBytes instanceof Uint8Array) || saltBytes.length < 8)
+    throw new Error("invalid salt");
+  const pwBytes = utf8encode(passphrase);
+  try {
+    return argon2id(pwBytes, saltBytes, ARGON2_PARAMS);
+  } finally {
+    pwBytes.fill(0); // best-effort scratch wipe
+  }
+}
+
+// ---------- record encryption ----------
+
+/**
+ * Encrypt a record payload.
+ *
+ * For a NEW record pass salt = randomSalt(). For an UPDATE pass the
+ * salt already stored on the record — the payload must be encrypted
+ * under a key derived from the SAME salt, otherwise decryption on the
+ * other side fails loudly. This binds the salt to the ciphertext:
+ * a modified salt (= malicious re-key) cannot go unnoticed.
+ *
+ * Payload is any JSON-serializable value. Returns the record body
+ * ready for the API ({uuid, ciphertext, nonce, salt}) with the AEAD
+ * computed over uuid+salt as associated data.
+ */
+export function encryptRecord(key, uuid, payload, saltBytes) {
+  if (!(saltBytes instanceof Uint8Array) || saltBytes.length !== REC_SALT_LEN)
+    throw new Error("invalid record salt");
+  const nonce = randomBytes(24);
+  // AAD binds uuid AND salt: swapping ciphertext between records or
+  // silently changing the salt breaks the tag.
+  const aad = utf8encode(
+    "crushdb:record:" + uuid + ":" + b64encode(saltBytes)
+  );
+  const plaintext = utf8encode(JSON.stringify(payload));
+  const ciphertext = xchacha20poly1305(key, nonce, aad).encrypt(plaintext);
+  plaintext.fill(0); // no plaintext residue in heap
+  return {
+    uuid,
+    ciphertext: b64encode(ciphertext),
+    nonce: b64encode(nonce),
+    salt: b64encode(saltBytes),
+  };
+}
+
+export function randomSalt() {
+  return randomBytes(REC_SALT_LEN);
+}
+
+// ---------- record decryption ----------
+
+/**
+ * Decrypt a record fetched from the API.
+ * Throws on wrong key, wrong salt, or any tampering (AEAD tag).
+ */
+export function decryptRecord(key, rec) {
+  if (!rec || typeof rec.uuid !== "string") throw new Error("invalid record");
+  const saltBytes = b64decode(rec.salt);
+  const aad = utf8encode(
+    "crushdb:record:" + rec.uuid + ":" + b64encode(saltBytes)
+  );
+  const ct = b64decode(rec.ciphertext);
+  const nonce = b64decode(rec.nonce);
+  const plaintext = xchacha20poly1305(key, nonce, aad).decrypt(ct);
+  try {
+    return JSON.parse(utf8decode(plaintext));
+  } finally {
+    plaintext.fill(0);
+  }
+}
+
+// ---------- hygiene ----------
+
+/**
+ * Best-effort wipe of a derived key (32 bytes). JS cannot guarantee
+ * zeroing, but this removes the obvious copy.
+ */
+export function wipeKey(key) {
+  if (key instanceof Uint8Array) key.fill(0);
+}

backend/static/index.html

@@ -0,0 +1,84 @@
+<!DOCTYPE html>
+<html lang="de">
+<head>
+  <meta charset="utf-8" />
+  <meta name="viewport" content="width=device-width, initial-scale=1" />
+  <title>crushdb</title>
+  <link rel="stylesheet" href="/static/style.css" />
+</head>
+<body>
+  <h1>crushdb</h1>
+  <p class="tagline">Verschlüsselte Datensätze zwischen zwei Personen</p>
+
+  <section id="flash" class="flash" aria-live="polite"></section>
+
+  <!-- unlock screen: passphrase stays in memory only -->
+  <section id="unlock">
+    <form id="unlock-form" autocomplete="off">
+      <label for="passphrase">Passphrase</label>
+      <input
+        type="password"
+        id="passphrase"
+        autocomplete="off"
+        required
+        placeholder="Passphrase eingeben …"
+      />
+      <button type="submit">Entsperren</button>
+      <p class="hint">
+        Die Passphrase verlässt niemals deinen Browser. Sie wird per
+        Argon2id lokal abgeleitet und nur im Arbeitsspeicher gehalten.
+      </p>
+    </form>
+  </section>
+
+  <!-- main app, hidden until unlocked -->
+  <section id="app" hidden>
+    <div class="toolbar">
+      <button id="new-btn" type="button">+ Neuer Datensatz</button>
+      <button id="lock-btn" type="button">Sperren</button>
+    </div>
+
+    <table id="records">
+      <thead>
+        <tr>
+          <th>Person A</th>
+          <th>Person B</th>
+          <th>Start</th>
+          <th>Ende</th>
+          <th>Aktion</th>
+          <th></th>
+        </tr>
+      </thead>
+      <tbody></tbody>
+    </table>
+  </section>
+
+  <!-- create/edit modal -->
+  <div id="editor-backdrop" hidden>
+    <form id="editor" autocomplete="off">
+      <h2 id="editor-title">Neuer Datensatz</h2>
+      <label>Person A
+        <input id="f-persona" type="text" required maxlength="120" />
+      </label>
+      <label>Person B
+        <input id="f-personb" type="text" required maxlength="120" />
+      </label>
+      <label>Start (optional)
+        <input id="f-tsstart" type="datetime-local" />
+      </label>
+      <label>Ende (optional)
+        <input id="f-tsend" type="datetime-local" />
+      </label>
+      <label>Aktion
+        <input id="f-action" type="text" required maxlength="500" />
+      </label>
+      <div class="actions">
+        <button type="submit">Speichern</button>
+        <button type="button" id="editor-cancel">Abbrechen</button>
+      </div>
+    </form>
+  </div>
+
+  <script type="module" src="/static/ui.js"></script>
+</body>
+</html>

backend/static/style.css

@@ -0,0 +1,129 @@
+/* crushdb – simple, clean dark UI */
+
+:root {
+  --bg: #101418;
+  --panel: #1a2028;
+  --panel-2: #202830;
+  --border: #2c3642;
+  --text: #e6edf3;
+  --muted: #8b98a5;
+  --accent: #4da3ff;
+  --danger: #ff5d5d;
+  --ok: #47d68a;
+}
+
+* { box-sizing: border-box; }
+
+body {
+  margin: 0 auto;
+  max-width: 960px;
+  padding: 2rem 1rem;
+  font-family: system-ui, -apple-system, "Segoe UI", sans-serif;
+  background: var(--bg);
+  color: var(--text);
+  line-height: 1.5;
+}
+
+h1 { margin: 0 0 0.25rem; font-size: 1.6rem; }
+.tagline { margin: 0 0 1.5rem; color: var(--muted); }
+
+#flash {
+  min-height: 1.4rem;
+  margin-bottom: 1rem;
+  color: var(--ok);
+}
+#flash.error { color: var(--danger); }
+
+#unlock {
+  max-width: 420px;
+  margin: 3rem auto;
+  background: var(--panel);
+  border: 1px solid var(--border);
+  border-radius: 10px;
+  padding: 1.5rem;
+}
+
+#unlock label { display: block; margin-bottom: 0.5rem; }
+#unlock input {
+  width: 100%;
+  margin-bottom: 1rem;
+}
+.hint { color: var(--muted); font-size: 0.85rem; margin-top: 1rem; }
+
+input, button {
+  font: inherit;
+  color: var(--text);
+}
+
+input {
+  background: var(--panel-2);
+  border: 1px solid var(--border);
+  border-radius: 6px;
+  padding: 0.5rem 0.7rem;
+}
+input:focus { outline: 1px solid var(--accent); }
+
+button {
+  background: var(--accent);
+  color: #0b1016;
+  border: none;
+  border-radius: 6px;
+  padding: 0.5rem 1rem;
+  cursor: pointer;
+  font-weight: 600;
+}
+button:hover { filter: brightness(1.1); }
+
+#app .toolbar {
+  display: flex;
+  gap: 0.75rem;
+  margin-bottom: 1rem;
+}
+#lock-btn { background: var(--panel-2); color: var(--text); border: 1px solid var(--border); }
+
+table {
+  width: 100%;
+  border-collapse: collapse;
+  background: var(--panel);
+  border: 1px solid var(--border);
+  border-radius: 10px;
+  overflow: hidden;
+}
+
+th, td {
+  padding: 0.6rem 0.75rem;
+  text-align: left;
+  border-bottom: 1px solid var(--border);
+}
+thead th { background: var(--panel-2); color: var(--muted); font-weight: 600; }
+tbody tr:last-child td { border-bottom: none; }
+tbody tr:hover { background: var(--panel-2); }
+
+.row-error td { color: var(--danger); font-size: 0.9rem; }
+
+td:last-child, td button { white-space: nowrap; }
+td button + button { margin-left: 0.4rem; }
+
+/* modal */
+#editor-backdrop {
+  position: fixed;
+  inset: 0;
+  background: rgba(0, 0, 0, 0.6);
+  display: flex;
+  align-items: center;
+  justify-content: center;
+}
+
+#editor {
+  background: var(--panel);
+  border: 1px solid var(--border);
+  border-radius: 10px;
+  padding: 1.5rem;
+  width: min(440px, 92vw);
+}
+
+#editor h2 { margin-top: 0; font-size: 1.15rem; }
+#editor label { display: block; margin-bottom: 0.75rem; color: var(--muted); }
+#editor input { width: 100%; margin-top: 0.25rem; }
+#editor .actions { display: flex; gap: 0.75rem; margin-top: 1rem; }
+#editor-cancel { background: var(--panel-2); color: var(--text); border: 1px solid var(--border); }

backend/static/ui.js

@@ -0,0 +1,275 @@
+/**
+ * ui.js – browser UI logic for crushdb.
+ *
+ * Flow:
+ *  1. User enters a passphrase → Argon2id derives the session key in
+ *     memory (never sent anywhere, never persisted).
+ *  2. Fetch ciphertext records from /api/records, decrypt each in
+ *     memory; decrypt failures are surfaced as "wrong passphrase or
+ *     tampered record" without leaking details.
+ *  3. Create/edit records: form values are encrypted client-side
+ *     before POST/PUT; the server only ever sees ciphertext.
+ *  4. Delete requires the record's current version (optimistic lock).
+ *  5. "Lock" wipes the passphrase field and releases decrypted rows so
+ *     nothing sensitive lingers in the DOM.
+ */
+
+import {
+  deriveKey,
+  encryptRecord,
+  decryptRecord,
+  randomSalt,
+  randomUuid,
+  wipeKey,
+} from "./crypto.js";
+
+// ---------- tiny DOM helpers ----------
+
+const $ = (sel) => document.querySelector(sel);
+
+function el(tag, attrs = {}, children = []) {
+  const node = document.createElement(tag);
+  for (const [k, v] of Object.entries(attrs)) {
+    if (k === "class") node.className = v;
+    else if (k === "text") node.textContent = v;
+    else node.setAttribute(k, v);
+  }
+  for (const child of children) node.appendChild(child);
+  return node;
+}
+
+function flash(msg, isError = false) {
+  const box = $("#flash");
+  box.textContent = msg;
+  box.className = isError ? "flash error" : "flash";
+  setTimeout(() => { box.textContent = ""; }, 4000);
+}
+
+// ---------- session state (memory only) ----------
+
+let sessionKey = null; // Uint8Array(32) or null
+
+function resetSession() {
+  wipeKey(sessionKey);
+  sessionKey = null;
+  $("#records").replaceChildren();
+  $("#app").hidden = true;
+  $("#unlock").hidden = false;
+  $("#passphrase").value = "";
+  $("#passphrase").focus();
+}
+
+// ---------- API helpers ----------
+
+async function api(path, opts = {}) {
+  const res = await fetch(path, {
+    headers: opts.body ? { "Content-Type": "application/json" } : {},
+    ...opts,
+  });
+  if (!res.ok) {
+    let detail = "";
+    try { detail = (await res.json()).error ?? ""; } catch { /* noop */ }
+    throw new Error(`HTTP ${res.status}${detail ? ": " + detail : ""}`);
+  }
+  if (res.status === 204) return null;
+  return res.json();
+}
+
+// ---------- rendering ----------
+
+function iso(ts) {
+  if (ts === null || ts === "") return "–";
+  const d = new Date(ts);
+  return isNaN(d) ? String(ts) : d.toLocaleString();
+}
+
+function renderRow(record, payload) {
+  const tr = el("tr");
+  const cells = [
+    payload.personA,
+    payload.personB,
+    iso(payload.tsStart),
+    iso(payload.tsEnd),
+    payload.action,
+  ];
+  for (const text of cells) {
+    const td = el("td");
+    td.textContent = text ?? ""; // textContent = no HTML injection
+    tr.appendChild(td);
+  }
+
+  const ops = el("td");
+  ops.appendChild(
+    el("button", {
+      text: "Bearbeiten",
+      type: "button",
+      onclick: () => openEditor(record, payload),
+    })
+  );
+  ops.appendChild(
+    el("button", {
+      text: "Löschen",
+      type: "button",
+      onclick: () => removeRecord(record),
+    })
+  );
+  tr.appendChild(ops);
+  return tr;
+}
+
+async function reload() {
+  const tbody = $("#records tbody");
+  tbody.replaceChildren();
+  const raw = await api("/api/records");
+  let failures = 0;
+  for (const record of raw) {
+    try {
+      const payload = decryptRecord(sessionKey, record);
+      payload.__version = record.version; // used for optimistic locking
+      tbody.appendChild(renderRow(record, payload));
+    } catch {
+      failures++;
+      tbody.appendChild(
+        el("tr", { class: "row-error" }, [
+          el("td", { colspan: "6", text: "Datensatz nicht entschlüsselbar (falsche Passphrase oder manipuliert)." }),
+        ])
+      );
+    }
+  }
+  if (failures > 0) flash(`${failures} Datensatz/Datensätze konnten nicht entschlüsselt werden.`, true);
+}
+
+function openEditor(record, payload) {
+  const form = $("#editor");
+  form.dataset.uuid = record.uuid;
+  form.dataset.version = payload.__version;
+  form.dataset.salt = record.salt;
+  $("#editor-title").textContent = "Datensatz bearbeiten";
+  $("#f-persona").value = payload.personA ?? "";
+  $("#f-personb").value = payload.personB ?? "";
+  $("#f-tsstart").value = payload.tsStart ? new Date(payload.tsStart).toISOString().slice(0, 16) : "";
+  $("#f-tsend").value = payload.tsEnd ? new Date(payload.tsEnd).toISOString().slice(0, 16) : "";
+  $("#f-action").value = payload.action ?? "";
+  $("#editor-backdrop").hidden = false;
+}
+
+function newRecord() {
+  const form = $("#editor");
+  form.reset();
+  form.dataset.uuid = randomUuid();
+  form.dataset.version = "";
+  form.dataset.salt = "";
+  $("#editor-title").textContent = "Neuer Datensatz";
+  $("#editor-backdrop").hidden = false;
+}
+
+function closeEditor() {
+  $("#editor-backdrop").hidden = true;
+  $("#editor").reset();
+}
+
+function tsOrNull(inputValue) {
+  if (!inputValue) return null;
+  const d = new Date(inputValue);
+  return isNaN(d) ? null : d.toISOString();
+}
+
+// ---------- record operations ----------
+
+async function saveRecord(ev) {
+  ev.preventDefault();
+  if (!sessionKey) { flash("Bitte zuerst entsperren.", true); return; }
+
+  const form = $("#editor");
+  const payload = {
+    personA: $("#f-persona").value.trim(),
+    personB: $("#f-personb").value.trim(),
+    tsStart: tsOrNull($("#f-tsstart").value),
+    tsEnd: tsOrNull($("#f-tsend").value),
+    action: $("#f-action").value.trim(),
+  };
+  if (!payload.personA || !payload.personB || !payload.action) {
+    flash("Person A, Person B und Aktion sind erforderlich.", true);
+    return;
+  }
+
+  const isEdit = form.dataset.version !== "";
+  const salt = isEdit ? new Uint8Array(atob(form.dataset.salt).split("").map((c) => c.charCodeAt(0))) : randomSalt();
+  const enc = encryptRecord(sessionKey, form.dataset.uuid, payload, salt);
+
+  try {
+    if (isEdit) {
+      enc.version = Number(form.dataset.version);
+      await api(`/api/records/${encodeURIComponent(enc.uuid)}`, { method: "PUT", body: JSON.stringify(enc) });
+    } else {
+      await api("/api/records", { method: "POST", body: JSON.stringify(enc) });
+    }
+    closeEditor();
+    await reload();
+    flash(isEdit ? "Datensatz aktualisiert." : "Datensatz erstellt.");
+  } catch (e) {
+    closeEditor();
+    if (String(e).includes("409")) {
+      flash("Konflikt: Der Datensatz wurde zwischenzeitlich geändert. Liste wird neu geladen.", true);
+    } else {
+      flash("Speichern fehlgeschlagen: " + e.message, true);
+    }
+    await reload();
+  }
+}
+
+async function removeRecord(record) {
+  if (!sessionKey) { flash("Bitte zuerst entsperren.", true); return; }
+  if (!confirm("Diesen Datensatz wirklich löschen?")) return;
+  try {
+    const current = await api(`/api/records/${encodeURIComponent(record.uuid)}`);
+    await api(`/api/records/${encodeURIComponent(record.uuid)}`, {
+      method: "DELETE",
+      body: JSON.stringify({ version: current.version }),
+    });
+    await reload();
+    flash("Datensatz gelöscht.");
+  } catch (e) {
+    if (String(e).includes("409")) flash("Konflikt beim Löschen (Version geändert). Bitte erneut versuchen.", true);
+    else flash("Löschen fehlgeschlagen: " + e.message, true);
+    await reload();
+  }
+}
+
+// ---------- unlock / lock ----------
+
+async function unlock(ev) {
+  ev.preventDefault();
+  const passphrase = $("#passphrase").value;
+  if (!passphrase) { flash("Bitte Passphrase eingeben.", true); return; }
+  // Salt for key derivation: per-record salts are used for the AEAD;
+  // for the session key we use a fixed domain-salt so all records can
+  // be opened with one passphrase. It contains no secret.
+  const domainSalt = new TextEncoder().encode("crushdb://session-key-v1");
+  try {
+    sessionKey = deriveKey(passphrase, domainSalt);
+    $("#passphrase").value = "";
+    $("#unlock").hidden = true;
+    $("#app").hidden = false;
+    await reload();
+  } catch (e) {
+    flash("Entsperren fehlgeschlagen: " + e.message, true);
+  }
+}
+
+function lock() {
+  if (!confirm("Sperren? Alle entschlüsselten Daten werden verworfen.")) return;
+  resetSession();
+}
+
+// ---------- wire-up ----------
+
+document.addEventListener("DOMContentLoaded", () => {
+  $("#unlock-form").addEventListener("submit", unlock);
+  $("#lock-btn").addEventListener("click", lock);
+  $("#new-btn").addEventListener("click", newRecord);
+  $("#editor").addEventListener("submit", saveRecord);
+  $("#editor-cancel").addEventListener("click", closeEditor);
+  window.addEventListener("beforeunload", () => wipeKey(sessionKey));
+  $("#passphrase").focus();
+});

backend/static/vendor/noble-ciphers/_arx.js

@@ -0,0 +1,379 @@
+/**
+ * Basic utils for ARX (add-rotate-xor) salsa and chacha ciphers.
+
+RFC8439 requires multi-step cipher stream, where
+authKey starts with counter: 0, actual msg with counter: 1.
+
+For this, we need a way to re-use nonce / counter:
+
+    const counter = new Uint8Array(4);
+    chacha(..., counter, ...); // counter is now 1
+    chacha(..., counter, ...); // counter is now 2
+
+This is complicated:
+
+- 32-bit counters are enough, no need for 64-bit: max ArrayBuffer size in JS is 4GB
+- Original papers don't allow mutating counters
+- Counter overflow is undefined [^1]
+- Idea A: allow providing (nonce | counter) instead of just nonce, re-use it
+- Caveat: Cannot be re-used through all cases:
+- * chacha has (counter | nonce)
+- * xchacha has (nonce16 | counter | nonce16)
+- Idea B: separate nonce / counter and provide separate API for counter re-use
+- Caveat: there are different counter sizes depending on an algorithm.
+- salsa & chacha also differ in structures of key & sigma:
+  salsa20:      s[0] | k(4) | s[1] | nonce(2) | cnt(2) | s[2] | k(4) | s[3]
+  chacha:       s(4) | k(8) | cnt(1) | nonce(3)
+  chacha20orig: s(4) | k(8) | cnt(2) | nonce(2)
+- Idea C: helper method such as `setSalsaState(key, nonce, sigma, data)`
+- Caveat: we can't re-use counter array
+
+xchacha uses the subkey and remaining 8 byte nonce with ChaCha20 as normal
+(prefixed by 4 NUL bytes, since RFC8439 specifies a 12-byte nonce).
+Counter overflow is undefined; see {@link https://mailarchive.ietf.org/arch/msg/cfrg/gsOnTJzcbgG6OqD8Sc0GO5aR_tU/ | the CFRG thread}.
+Current noble policy is strict non-wrap for the shared 32-bit counter path:
+exported ARX ciphers reject initial `0xffffffff` and stop before any implicit
+wrap back to zero.
+See {@link https://datatracker.ietf.org/doc/html/draft-irtf-cfrg-xchacha#appendix-A.2 | the XChaCha appendix} for the extended-nonce construction.
+
+ * @module
+ */
+import { abool, abytes, anumber, checkOpts, clean, complexOverlapBytes, copyBytes, getOutput, isAligned32, isLE, randomBytes, swap32IfBE, u32, } from "./utils.js";
+// Replaces `TextEncoder` for ASCII literals, which is enough for sigma constants.
+// Non-ASCII input would not match UTF-8 `TextEncoder` output.
+const encodeStr = (str) => Uint8Array.from(str.split(''), (c) => c.charCodeAt(0));
+// Raw `createCipher(...)` exports consume these native-endian `u32(...)` views directly.
+// Public `wrapCipher(...)` APIs reject non-little-endian platforms before reaching this path.
+// RFC 8439 §2.3 / RFC 7539 §2.3 only define the 256-bit-key constants; this 16-byte sigma is
+// kept for legacy allowShortKeys Salsa/ChaCha variants.
+const sigma16_32 = /* @__PURE__ */ (() => swap32IfBE(u32(encodeStr('expand 16-byte k'))))();
+// RFC 8439 §2.3 / RFC 7539 §2.3 define words 0-3 as
+// `0x61707865 0x3320646e 0x79622d32 0x6b206574`, i.e. `expand 32-byte k`.
+const sigma32_32 = /* @__PURE__ */ (() => swap32IfBE(u32(encodeStr('expand 32-byte k'))))();
+/**
+ * Rotates a 32-bit word left.
+ * @param a - Input word.
+ * @param b - Rotation count in bits.
+ * @returns Rotated 32-bit word.
+ * @example
+ * Moves the top byte of `0x12345678` into the low byte position.
+ * ```ts
+ * rotl(0x12345678, 8);
+ * ```
+ */
+export function rotl(a, b) {
+    return (a << b) | (a >>> (32 - b));
+}
+// Salsa and Chacha block length is always 512-bit
+const BLOCK_LEN = 64;
+// RFC 8439 §2.2 / RFC 7539 §2.2: the ChaCha state has 16 32-bit words.
+const BLOCK_LEN32 = 16;
+// Counter policy for the shared public `counter` argument:
+// - RFC/IETF ChaCha20 uses a 32-bit counter.
+// - OpenSSL/Node `chacha20` instead treat the full 16-byte IV as a 128-bit
+//   counter state and carry into the next word.
+// - Raw `chacha20orig`, `salsa20`, `xsalsa20`, and `xchacha20` use 64-bit counters in libsodium
+//   and libtomcrypt, while some libs (for example libtomcrypt's RFC/IETF path) reject the max
+//   boundary instead of carrying.
+// - AEAD wrappers diverge too: libsodium `xchacha20poly1305` uses the IETF payload counter from
+//   block 1, while `secretstream_xchacha20poly1305` is a different protocol with rekey/reset.
+// Noble intentionally throws instead of silently picking one wrap model for users. In the default
+// path, even a 32-bit boundary would take 2^32 blocks * 64 bytes = 256 GiB, which is practically
+// unreachable for normal JS callers; advanced users who pass `counter` explicitly can implement
+// whatever wider carry / wrap policy they need on top.
+const MAX_COUNTER = /* @__PURE__ */ (() => 2 ** 32 - 1)();
+const U32_EMPTY = /* @__PURE__ */ Uint32Array.of();
+function runCipher(core, sigma, key, nonce, data, output, counter, rounds) {
+    const len = data.length;
+    const block = new Uint8Array(BLOCK_LEN);
+    const b32 = u32(block);
+    // Make sure that buffers aligned to 4 bytes
+    const isAligned = isLE && isAligned32(data) && isAligned32(output);
+    const d32 = isAligned ? u32(data) : U32_EMPTY;
+    const o32 = isAligned ? u32(output) : U32_EMPTY;
+    // RFC 8439 §2.4.1 / RFC 7539 §2.4.1 allow XORing one keystream block at a time and
+    // truncating the final partial block instead of materializing the whole keystream.
+    if (!isLE) {
+        for (let pos = 0; pos < len; counter++) {
+            core(sigma, key, nonce, b32, counter, rounds);
+            // RFC 8439 §2.4 / RFC 7539 §2.4 serialize keystream words in little-endian order.
+            swap32IfBE(b32);
+            if (counter >= MAX_COUNTER)
+                throw new Error('arx: counter overflow');
+            const take = Math.min(BLOCK_LEN, len - pos);
+            for (let j = 0, posj; j < take; j++) {
+                posj = pos + j;
+                output[posj] = data[posj] ^ block[j];
+            }
+            pos += take;
+        }
+        return;
+    }
+    for (let pos = 0; pos < len; counter++) {
+        core(sigma, key, nonce, b32, counter, rounds);
+        // See MAX_COUNTER policy note above: never silently wrap the shared public counter.
+        if (counter >= MAX_COUNTER)
+            throw new Error('arx: counter overflow');
+        const take = Math.min(BLOCK_LEN, len - pos);
+        // aligned to 4 bytes
+        if (isAligned && take === BLOCK_LEN) {
+            const pos32 = pos / 4;
+            if (pos % 4 !== 0)
+                throw new Error('arx: invalid block position');
+            for (let j = 0, posj; j < BLOCK_LEN32; j++) {
+                posj = pos32 + j;
+                o32[posj] = d32[posj] ^ b32[j];
+            }
+            pos += BLOCK_LEN;
+            continue;
+        }
+        for (let j = 0, posj; j < take; j++) {
+            posj = pos + j;
+            output[posj] = data[posj] ^ block[j];
+        }
+        pos += take;
+    }
+}
+/**
+ * Creates an ARX stream cipher from a 32-bit core permutation.
+ * Used internally to build the exported Salsa and ChaCha stream ciphers.
+ * @param core - Core function that fills one keystream block.
+ * @param opts - Cipher layout and nonce-extension options. See {@link CipherOpts}.
+ * @returns Stream cipher function over byte arrays.
+ * @throws If the core callback, key size, counter, or output sizing is invalid. {@link Error}
+ */
+export function createCipher(core, opts) {
+    const { allowShortKeys, extendNonceFn, counterLength, counterRight, rounds } = checkOpts({ allowShortKeys: false, counterLength: 8, counterRight: false, rounds: 20 }, opts);
+    if (typeof core !== 'function')
+        throw new Error('core must be a function');
+    anumber(counterLength);
+    anumber(rounds);
+    abool(counterRight);
+    abool(allowShortKeys);
+    return (key, nonce, data, output, counter = 0) => {
+        abytes(key, undefined, 'key');
+        abytes(nonce, undefined, 'nonce');
+        abytes(data, undefined, 'data');
+        const len = data.length;
+        // Raw XorStream APIs return ciphertext/plaintext bytes directly, so caller-provided outputs
+        // must match the logical result length exactly instead of returning an oversized workspace.
+        const hasOutput = output !== undefined;
+        output = getOutput(len, output, false);
+        if (hasOutput)
+            complexOverlapBytes(data, output);
+        anumber(counter);
+        // See MAX_COUNTER policy note above: reject advanced explicit-counter requests before any wrap.
+        if (counter < 0 || counter >= MAX_COUNTER)
+            throw new Error('arx: counter overflow');
+        const toClean = [];
+        // Key & sigma
+        // key=16 -> sigma16, k=key|key
+        // key=32 -> sigma32, k=key
+        let l = key.length;
+        let k;
+        let sigma;
+        if (l === 32) {
+            // Copy caller keys too: big-endian normalization, extended-nonce subkey derivation, and
+            // final clean(...) all mutate or wipe the temporary buffer in place.
+            toClean.push((k = copyBytes(key)));
+            sigma = sigma32_32;
+        }
+        else if (l === 16 && allowShortKeys) {
+            k = new Uint8Array(32);
+            k.set(key);
+            k.set(key, 16);
+            sigma = sigma16_32;
+            toClean.push(k);
+        }
+        else {
+            abytes(key, 32, 'arx key');
+            throw new Error('invalid key size');
+            // throw new Error(`"arx key" expected Uint8Array of length 32, got length=${l}`);
+        }
+        // Nonce
+        // salsa20:      8   (8-byte counter)
+        // chacha20orig: 8   (8-byte counter)
+        // chacha20:     12  (4-byte counter)
+        // xsalsa20:     24  (16 -> hsalsa,  8 -> old nonce)
+        // xchacha20:    24  (16 -> hchacha, 8 -> old nonce)
+        // Copy before taking u32(...) views on misaligned inputs, and on big-endian so later
+        // swap32IfBE(...) never mutates caller nonce bytes in place.
+        if (!isLE || !isAligned32(nonce))
+            toClean.push((nonce = copyBytes(nonce)));
+        let k32 = u32(k);
+        // hsalsa & hchacha: handle extended nonce
+        if (extendNonceFn) {
+            if (nonce.length !== 24)
+                throw new Error('arx: extended nonce must be 24 bytes');
+            const n16 = nonce.subarray(0, 16);
+            if (isLE)
+                extendNonceFn(sigma, k32, u32(n16), k32);
+            else {
+                const sigmaRaw = swap32IfBE(Uint32Array.from(sigma));
+                extendNonceFn(sigmaRaw, k32, u32(n16), k32);
+                clean(sigmaRaw);
+                swap32IfBE(k32);
+            }
+            nonce = nonce.subarray(16);
+        }
+        else if (!isLE)
+            swap32IfBE(k32);
+        // Handle nonce counter
+        const nonceNcLen = 16 - counterLength;
+        if (nonceNcLen !== nonce.length)
+            throw new Error(`arx: nonce must be ${nonceNcLen} or 16 bytes`);
+        // Normalize 64-bit-nonce layouts to the 12-byte core input: ChaCha/XChaCha prefix 4 zero
+        // counter bytes, while Salsa/XSalsa append them after the nonce words.
+        if (nonceNcLen !== 12) {
+            const nc = new Uint8Array(12);
+            nc.set(nonce, counterRight ? 0 : 12 - nonce.length);
+            nonce = nc;
+            toClean.push(nonce);
+        }
+        const n32 = swap32IfBE(u32(nonce));
+        // Ensure temporary key/nonce copies are wiped even if the remaining
+        // runtime guard in runCipher(...) throws on counter overflow.
+        try {
+            runCipher(core, sigma, k32, n32, data, output, counter, rounds);
+            return output;
+        }
+        finally {
+            clean(...toClean);
+        }
+    };
+}
+/** Internal class which wraps chacha20 or chacha8 to create CSPRNG. */
+export class _XorStreamPRG {
+    blockLen;
+    keyLen;
+    nonceLen;
+    state;
+    buf;
+    key;
+    nonce;
+    pos;
+    ctr;
+    cipher;
+    destroyed = false;
+    constructor(cipher, blockLen, keyLen, nonceLen, seed) {
+        this.cipher = cipher;
+        this.blockLen = blockLen;
+        this.keyLen = keyLen;
+        this.nonceLen = nonceLen;
+        this.state = new Uint8Array(this.keyLen + this.nonceLen);
+        this.reseed(seed);
+        this.ctr = 0;
+        this.pos = this.blockLen;
+        this.buf = new Uint8Array(this.blockLen);
+        // Keep a single key||nonce backing buffer so reseed/addEntropy/clean update the live cipher
+        // inputs in place through these subarray views.
+        this.key = this.state.subarray(0, this.keyLen);
+        this.nonce = this.state.subarray(this.keyLen);
+    }
+    reseed(seed) {
+        abytes(seed, undefined, 'seed');
+        if (!seed || seed.length === 0)
+            throw new Error('entropy required');
+        // Mix variable-length entropy cyclically across the whole key||nonce state, then restart the
+        // keystream so buffered leftovers from the previous state are never reused.
+        for (let i = 0; i < seed.length; i++)
+            this.state[i % this.state.length] ^= seed[i];
+        this.ctr = 0;
+        this.pos = this.blockLen;
+    }
+    addEntropy(seed) {
+        if (this.destroyed)
+            throw new Error('cannot use destroyed PRG');
+        // Reject empty entropy before re-keying, otherwise a throwing call would still advance state.
+        abytes(seed, undefined, 'seed');
+        if (seed.length === 0)
+            throw new Error('entropy required');
+        // Re-key from the current stream first, then mix external entropy into the fresh key||nonce
+        // state through reseed() so stale buffered bytes are discarded.
+        this.state.set(this.randomBytes(this.state.length));
+        this.reseed(seed);
+    }
+    randomBytes(len) {
+        if (this.destroyed)
+            throw new Error('cannot use destroyed PRG');
+        anumber(len);
+        if (len === 0)
+            return new Uint8Array(0);
+        const avail = this.pos < this.blockLen ? this.blockLen - this.pos : 0;
+        const blocks = Math.ceil(Math.max(0, len - avail) / this.blockLen);
+        // Preflight overflow so failed reads don't partially consume keystream
+        // and leave the PRG repeating blocks.
+        if (blocks > 0 && this.ctr > MAX_COUNTER - blocks)
+            throw new Error('arx: counter overflow');
+        const out = new Uint8Array(len);
+        let outPos = 0;
+        // `out` starts zero-filled, and `buf.fill(0)` below does the same for leftovers: XOR-stream
+        // ciphers then emit raw keystream bytes directly into those buffers.
+        // Serve buffered leftovers first so split reads stay identical to one larger read.
+        if (this.pos < this.blockLen) {
+            const take = Math.min(len, this.blockLen - this.pos);
+            out.set(this.buf.subarray(this.pos, this.pos + take), 0);
+            this.pos += take;
+            outPos += take;
+            if (outPos === len)
+                return out; // fast path
+        }
+        // Full blocks directly to out
+        const full = Math.floor((len - outPos) / this.blockLen);
+        if (full > 0) {
+            const blockBytes = full * this.blockLen;
+            const b = out.subarray(outPos, outPos + blockBytes);
+            this.cipher(this.key, this.nonce, b, b, this.ctr);
+            this.ctr += full;
+            outPos += blockBytes;
+        }
+        // Save leftovers
+        const left = len - outPos;
+        if (left > 0) {
+            this.buf.fill(0);
+            // NOTE: cipher will handle overflow
+            this.cipher(this.key, this.nonce, this.buf, this.buf, this.ctr++);
+            out.set(this.buf.subarray(0, left), outPos);
+            this.pos = left;
+        }
+        return out;
+    }
+    // Clone seeds the new instance from this stream, so the source PRG advances too.
+    clone() {
+        if (this.destroyed)
+            throw new Error('cannot use destroyed PRG');
+        return new _XorStreamPRG(this.cipher, this.blockLen, this.keyLen, this.nonceLen, this.randomBytes(this.state.length));
+    }
+    // Zeroes the current state, leftover buffer, and marks the instance destroyed.
+    // The instance fails closed: any later randomBytes()/addEntropy()/clone() throws instead
+    // of silently continuing from zero key||nonce state (which would yield predictable output).
+    clean() {
+        this.pos = 0;
+        this.ctr = 0;
+        this.buf.fill(0);
+        this.state.fill(0);
+        this.destroyed = true;
+    }
+}
+/**
+ * Creates a PRG constructor from a stream cipher.
+ * @param cipher - Stream cipher used to fill output blocks.
+ * @param blockLen - Keystream block length in bytes.
+ * @param keyLen - Internal key length in bytes.
+ * @param nonceLen - Internal nonce length in bytes.
+ * @returns PRG factory for seeded concrete `_XorStreamPRG` instances.
+ * @example
+ * Builds a PRG from XChaCha20 and reads bytes from a randomly seeded instance.
+ * ```ts
+ * import { xchacha20 } from '@noble/ciphers/chacha.js';
+ * import { createPRG } from '@noble/ciphers/_arx.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const seed = randomBytes(32);
+ * const init = createPRG(xchacha20, 64, 32, 24);
+ * const prg = init(seed);
+ * prg.randomBytes(8);
+ * ```
+ */
+export const createPRG = (cipher, blockLen, keyLen, nonceLen) => {
+    return ((seed = randomBytes(32)) => new _XorStreamPRG(cipher, blockLen, keyLen, nonceLen, seed));
+};

backend/static/vendor/noble-ciphers/_poly1305.js

@@ -0,0 +1,318 @@
+/**
+ * Poly1305 ({@link https://cr.yp.to/mac/poly1305-20050329.pdf | PDF},
+ * {@link https://en.wikipedia.org/wiki/Poly1305 | wiki})
+ * is a fast and parallel secret-key message-authentication code suitable for
+ * a wide variety of applications. It was standardized in
+ * {@link https://www.rfc-editor.org/rfc/rfc8439 | RFC 8439} and is now used in TLS 1.3.
+ *
+ * Polynomial MACs are not perfect for every situation:
+ * they lack Random Key Robustness: the MAC can be forged, and can't be used in PAKE schemes.
+ * See {@link https://keymaterial.net/2020/09/07/invisible-salamanders-in-aes-gcm-siv/ | the invisible salamanders attack writeup}.
+ * To combat invisible salamanders, `hash(key)` can be included in ciphertext,
+ * however, this would violate ciphertext indistinguishability:
+ * an attacker would know which key was used - so `HKDF(key, i)`
+ * could be used instead.
+ *
+ * Check out the {@link https://cr.yp.to/mac.html | original website}.
+ * Based on public-domain {@link https://github.com/floodyberry/poly1305-donna | poly1305-donna}.
+ * @module
+ */
+// prettier-ignore
+import { abytes, aexists, aoutput, clean, copyBytes, wrapMacConstructor } from "./utils.js";
+// A simple-loop reference version (`poly1305_small`) lives in
+// `test/misc/micro-ciphers.ts`, provided for auditability.
+// Little-endian 2-byte load used by the Poly1305 limb decomposition.
+function u8to16(a, i) {
+    return (a[i++] & 0xff) | ((a[i++] & 0xff) << 8);
+}
+/**
+ * Incremental Poly1305 MAC state.
+ * Prefer `poly1305()` for one-shot use.
+ * @param key - 32-byte Poly1305 one-time key.
+ * @example
+ * Feeds one chunk into an incremental Poly1305 state with a fresh one-time key.
+ *
+ * ```ts
+ * import { Poly1305 } from '@noble/ciphers/_poly1305.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const mac = new Poly1305(key);
+ * mac.update(new Uint8Array([1, 2, 3]));
+ * mac.digest();
+ * ```
+ */
+export class Poly1305 {
+    blockLen = 16;
+    outputLen = 16;
+    buffer = new Uint8Array(16);
+    r = new Uint16Array(10); // Allocating 1 array with .subarray() here is slower than 3
+    h = new Uint16Array(10);
+    pad = new Uint16Array(8);
+    pos = 0;
+    finished = false;
+    destroyed = false;
+    // Can be speed-up using BigUint64Array, at the cost of complexity
+    constructor(key) {
+        key = copyBytes(abytes(key, 32, 'key'));
+        const t0 = u8to16(key, 0);
+        const t1 = u8to16(key, 2);
+        const t2 = u8to16(key, 4);
+        const t3 = u8to16(key, 6);
+        const t4 = u8to16(key, 8);
+        const t5 = u8to16(key, 10);
+        const t6 = u8to16(key, 12);
+        const t7 = u8to16(key, 14);
+        // RFC 8439 §2.5.1 / RFC 7539 §2.5.1 clamp r before multiplication.
+        // These masks unpack that clamped value into 13-bit limbs, while pad
+        // keeps the raw s half for finalize().
+        // {@link https://github.com/floodyberry/poly1305-donna/blob/e6ad6e091d30d7f4ec2d4f978be1fcfcbce72781/poly1305-donna-16.h#L47 | poly1305-donna reference}
+        this.r[0] = t0 & 0x1fff;
+        this.r[1] = ((t0 >>> 13) | (t1 << 3)) & 0x1fff;
+        this.r[2] = ((t1 >>> 10) | (t2 << 6)) & 0x1f03;
+        this.r[3] = ((t2 >>> 7) | (t3 << 9)) & 0x1fff;
+        this.r[4] = ((t3 >>> 4) | (t4 << 12)) & 0x00ff;
+        this.r[5] = (t4 >>> 1) & 0x1ffe;
+        this.r[6] = ((t4 >>> 14) | (t5 << 2)) & 0x1fff;
+        this.r[7] = ((t5 >>> 11) | (t6 << 5)) & 0x1f81;
+        this.r[8] = ((t6 >>> 8) | (t7 << 8)) & 0x1fff;
+        this.r[9] = (t7 >>> 5) & 0x007f;
+        for (let i = 0; i < 8; i++)
+            this.pad[i] = u8to16(key, 16 + 2 * i);
+    }
+    process(data, offset, isLast = false) {
+        // RFC 8439 §2.5 / §2.5.1 and RFC 7539 §2.5 / §2.5.1 add an extra high
+        // bit to every full 16-byte block. The final partial block gets its
+        // explicit `1` byte during digestInto(), so `hibit` stays zero there.
+        const hibit = isLast ? 0 : 1 << 11;
+        const { h, r } = this;
+        const r0 = r[0];
+        const r1 = r[1];
+        const r2 = r[2];
+        const r3 = r[3];
+        const r4 = r[4];
+        const r5 = r[5];
+        const r6 = r[6];
+        const r7 = r[7];
+        const r8 = r[8];
+        const r9 = r[9];
+        const t0 = u8to16(data, offset + 0);
+        const t1 = u8to16(data, offset + 2);
+        const t2 = u8to16(data, offset + 4);
+        const t3 = u8to16(data, offset + 6);
+        const t4 = u8to16(data, offset + 8);
+        const t5 = u8to16(data, offset + 10);
+        const t6 = u8to16(data, offset + 12);
+        const t7 = u8to16(data, offset + 14);
+        let h0 = h[0] + (t0 & 0x1fff);
+        let h1 = h[1] + (((t0 >>> 13) | (t1 << 3)) & 0x1fff);
+        let h2 = h[2] + (((t1 >>> 10) | (t2 << 6)) & 0x1fff);
+        let h3 = h[3] + (((t2 >>> 7) | (t3 << 9)) & 0x1fff);
+        let h4 = h[4] + (((t3 >>> 4) | (t4 << 12)) & 0x1fff);
+        let h5 = h[5] + ((t4 >>> 1) & 0x1fff);
+        let h6 = h[6] + (((t4 >>> 14) | (t5 << 2)) & 0x1fff);
+        let h7 = h[7] + (((t5 >>> 11) | (t6 << 5)) & 0x1fff);
+        let h8 = h[8] + (((t6 >>> 8) | (t7 << 8)) & 0x1fff);
+        let h9 = h[9] + ((t7 >>> 5) | hibit);
+        let c = 0;
+        let d0 = c + h0 * r0 + h1 * (5 * r9) + h2 * (5 * r8) + h3 * (5 * r7) + h4 * (5 * r6);
+        c = d0 >>> 13;
+        d0 &= 0x1fff;
+        d0 += h5 * (5 * r5) + h6 * (5 * r4) + h7 * (5 * r3) + h8 * (5 * r2) + h9 * (5 * r1);
+        c += d0 >>> 13;
+        d0 &= 0x1fff;
+        let d1 = c + h0 * r1 + h1 * r0 + h2 * (5 * r9) + h3 * (5 * r8) + h4 * (5 * r7);
+        c = d1 >>> 13;
+        d1 &= 0x1fff;
+        d1 += h5 * (5 * r6) + h6 * (5 * r5) + h7 * (5 * r4) + h8 * (5 * r3) + h9 * (5 * r2);
+        c += d1 >>> 13;
+        d1 &= 0x1fff;
+        let d2 = c + h0 * r2 + h1 * r1 + h2 * r0 + h3 * (5 * r9) + h4 * (5 * r8);
+        c = d2 >>> 13;
+        d2 &= 0x1fff;
+        d2 += h5 * (5 * r7) + h6 * (5 * r6) + h7 * (5 * r5) + h8 * (5 * r4) + h9 * (5 * r3);
+        c += d2 >>> 13;
+        d2 &= 0x1fff;
+        let d3 = c + h0 * r3 + h1 * r2 + h2 * r1 + h3 * r0 + h4 * (5 * r9);
+        c = d3 >>> 13;
+        d3 &= 0x1fff;
+        d3 += h5 * (5 * r8) + h6 * (5 * r7) + h7 * (5 * r6) + h8 * (5 * r5) + h9 * (5 * r4);
+        c += d3 >>> 13;
+        d3 &= 0x1fff;
+        let d4 = c + h0 * r4 + h1 * r3 + h2 * r2 + h3 * r1 + h4 * r0;
+        c = d4 >>> 13;
+        d4 &= 0x1fff;
+        d4 += h5 * (5 * r9) + h6 * (5 * r8) + h7 * (5 * r7) + h8 * (5 * r6) + h9 * (5 * r5);
+        c += d4 >>> 13;
+        d4 &= 0x1fff;
+        let d5 = c + h0 * r5 + h1 * r4 + h2 * r3 + h3 * r2 + h4 * r1;
+        c = d5 >>> 13;
+        d5 &= 0x1fff;
+        d5 += h5 * r0 + h6 * (5 * r9) + h7 * (5 * r8) + h8 * (5 * r7) + h9 * (5 * r6);
+        c += d5 >>> 13;
+        d5 &= 0x1fff;
+        let d6 = c + h0 * r6 + h1 * r5 + h2 * r4 + h3 * r3 + h4 * r2;
+        c = d6 >>> 13;
+        d6 &= 0x1fff;
+        d6 += h5 * r1 + h6 * r0 + h7 * (5 * r9) + h8 * (5 * r8) + h9 * (5 * r7);
+        c += d6 >>> 13;
+        d6 &= 0x1fff;
+        let d7 = c + h0 * r7 + h1 * r6 + h2 * r5 + h3 * r4 + h4 * r3;
+        c = d7 >>> 13;
+        d7 &= 0x1fff;
+        d7 += h5 * r2 + h6 * r1 + h7 * r0 + h8 * (5 * r9) + h9 * (5 * r8);
+        c += d7 >>> 13;
+        d7 &= 0x1fff;
+        let d8 = c + h0 * r8 + h1 * r7 + h2 * r6 + h3 * r5 + h4 * r4;
+        c = d8 >>> 13;
+        d8 &= 0x1fff;
+        d8 += h5 * r3 + h6 * r2 + h7 * r1 + h8 * r0 + h9 * (5 * r9);
+        c += d8 >>> 13;
+        d8 &= 0x1fff;
+        let d9 = c + h0 * r9 + h1 * r8 + h2 * r7 + h3 * r6 + h4 * r5;
+        c = d9 >>> 13;
+        d9 &= 0x1fff;
+        d9 += h5 * r4 + h6 * r3 + h7 * r2 + h8 * r1 + h9 * r0;
+        c += d9 >>> 13;
+        d9 &= 0x1fff;
+        c = ((c << 2) + c) | 0;
+        c = (c + d0) | 0;
+        d0 = c & 0x1fff;
+        c = c >>> 13;
+        d1 += c;
+        h[0] = d0;
+        h[1] = d1;
+        h[2] = d2;
+        h[3] = d3;
+        h[4] = d4;
+        h[5] = d5;
+        h[6] = d6;
+        h[7] = d7;
+        h[8] = d8;
+        h[9] = d9;
+    }
+    finalize() {
+        const { h, pad } = this;
+        const g = new Uint16Array(10);
+        let c = h[1] >>> 13;
+        h[1] &= 0x1fff;
+        for (let i = 2; i < 10; i++) {
+            h[i] += c;
+            c = h[i] >>> 13;
+            h[i] &= 0x1fff;
+        }
+        h[0] += c * 5;
+        c = h[0] >>> 13;
+        h[0] &= 0x1fff;
+        h[1] += c;
+        c = h[1] >>> 13;
+        h[1] &= 0x1fff;
+        h[2] += c;
+        // RFC 8439 §2.5 / RFC 7539 §2.5 reduce modulo 2^130-5 before repacking
+        // to 16-bit words and adding the raw s half.
+        g[0] = h[0] + 5;
+        c = g[0] >>> 13;
+        g[0] &= 0x1fff;
+        for (let i = 1; i < 10; i++) {
+            g[i] = h[i] + c;
+            c = g[i] >>> 13;
+            g[i] &= 0x1fff;
+        }
+        g[9] -= 1 << 13;
+        let mask = (c ^ 1) - 1;
+        for (let i = 0; i < 10; i++)
+            g[i] &= mask;
+        mask = ~mask;
+        for (let i = 0; i < 10; i++)
+            h[i] = (h[i] & mask) | g[i];
+        h[0] = (h[0] | (h[1] << 13)) & 0xffff;
+        h[1] = ((h[1] >>> 3) | (h[2] << 10)) & 0xffff;
+        h[2] = ((h[2] >>> 6) | (h[3] << 7)) & 0xffff;
+        h[3] = ((h[3] >>> 9) | (h[4] << 4)) & 0xffff;
+        h[4] = ((h[4] >>> 12) | (h[5] << 1) | (h[6] << 14)) & 0xffff;
+        h[5] = ((h[6] >>> 2) | (h[7] << 11)) & 0xffff;
+        h[6] = ((h[7] >>> 5) | (h[8] << 8)) & 0xffff;
+        h[7] = ((h[8] >>> 8) | (h[9] << 5)) & 0xffff;
+        let f = h[0] + pad[0];
+        h[0] = f & 0xffff;
+        for (let i = 1; i < 8; i++) {
+            f = (((h[i] + pad[i]) | 0) + (f >>> 16)) | 0;
+            h[i] = f & 0xffff;
+        }
+        clean(g);
+    }
+    update(data) {
+        aexists(this);
+        abytes(data);
+        data = copyBytes(data);
+        const { buffer, blockLen } = this;
+        const len = data.length;
+        for (let pos = 0; pos < len;) {
+            const take = Math.min(blockLen - this.pos, len - pos);
+            // Fast path: we have at least one block in input
+            if (take === blockLen) {
+                for (; blockLen <= len - pos; pos += blockLen)
+                    this.process(data, pos);
+                continue;
+            }
+            buffer.set(data.subarray(pos, pos + take), this.pos);
+            this.pos += take;
+            pos += take;
+            if (this.pos === blockLen) {
+                this.process(buffer, 0, false);
+                this.pos = 0;
+            }
+        }
+        return this;
+    }
+    destroy() {
+        // `aexists(this)` guards update/digest paths, so destroy must mark the instance unusable too.
+        this.destroyed = true;
+        clean(this.h, this.r, this.buffer, this.pad);
+    }
+    digestInto(out) {
+        aexists(this);
+        aoutput(out, this);
+        this.finished = true;
+        const { buffer, h } = this;
+        let { pos } = this;
+        if (pos) {
+            // RFC 8439 §2.5 / RFC 7539 §2.5: the final short block appends a
+            // single `0x01` byte and zero-fills the remaining bytes before the
+            // last multiplication step.
+            buffer[pos++] = 1;
+            for (; pos < 16; pos++)
+                buffer[pos] = 0;
+            this.process(buffer, 0, true);
+        }
+        this.finalize();
+        let opos = 0;
+        for (let i = 0; i < 8; i++) {
+            out[opos++] = h[i] >>> 0;
+            out[opos++] = h[i] >>> 8;
+        }
+    }
+    digest() {
+        const { buffer, outputLen } = this;
+        this.digestInto(buffer);
+        // Copy out before destroy() zeroes the internal buffer.
+        const res = buffer.slice(0, outputLen);
+        this.destroy();
+        return res;
+    }
+}
+/**
+ * Poly1305 MAC from RFC 8439.
+ * @param msg - Message bytes to authenticate.
+ * @param key - 32-byte Poly1305 one-time key.
+ * @returns 16-byte authentication tag.
+ * @example
+ * Authenticates one message with a one-shot Poly1305 call and a fresh key.
+ *
+ * ```ts
+ * import { poly1305 } from '@noble/ciphers/_poly1305.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * poly1305(new Uint8Array(), key);
+ * ```
+ */
+export const poly1305 = /* @__PURE__ */ wrapMacConstructor(32, (key) => new Poly1305(key));

backend/static/vendor/noble-ciphers/chacha.js

@@ -0,0 +1,460 @@
+/**
+ * ChaCha stream cipher, released
+ * in 2008. Developed after Salsa20, ChaCha aims to increase diffusion per round.
+ * It was standardized in
+ * {@link https://www.rfc-editor.org/rfc/rfc8439 | RFC 8439} and
+ * is now used in TLS 1.3.
+ *
+ * {@link https://datatracker.ietf.org/doc/html/draft-irtf-cfrg-xchacha | XChaCha20}
+ * extended-nonce variant is also provided. Similar to XSalsa, it's safe to use with
+ * randomly-generated nonces.
+ *
+ * Check out
+ * {@link http://cr.yp.to/chacha/chacha-20080128.pdf | PDF},
+ * {@link https://en.wikipedia.org/wiki/Salsa20 | wiki}, and
+ * {@link https://cr.yp.to/chacha.html | website}.
+ *
+ * @module
+ */
+import { createCipher, createPRG, rotl } from "./_arx.js";
+import { poly1305 } from "./_poly1305.js";
+import { abytes, clean, equalBytes, getOutput, isLE, swap32IfBE, u64Lengths, wrapCipher, } from "./utils.js";
+/**
+ * ChaCha core function. Uses an unrolled loop (chachaCore, hchacha) - 4x
+ * faster than a simple loop, but larger & harder to read. A simple-loop
+ * reference version lives in `test/misc/micro-ciphers.ts`;
+ * `test/arx.test.ts` keeps the two aligned.
+ * The specific implementation is selected in `createCipher` below.
+ */
+/** RFC 8439 §2.3 block core for `state = constants | key | counter | nonce`. */
+// prettier-ignore
+function chachaCore(s, k, n, out, cnt, rounds = 20) {
+    let y00 = s[0], y01 = s[1], y02 = s[2], y03 = s[3], // "expa"   "nd 3"  "2-by"  "te k"
+    y04 = k[0], y05 = k[1], y06 = k[2], y07 = k[3], // Key      Key     Key     Key
+    y08 = k[4], y09 = k[5], y10 = k[6], y11 = k[7], // Key      Key     Key     Key
+    y12 = cnt, y13 = n[0], y14 = n[1], y15 = n[2]; // Counter  Nonce   Nonce   Nonce
+    // Save state to temporary variables
+    let x00 = y00, x01 = y01, x02 = y02, x03 = y03, x04 = y04, x05 = y05, x06 = y06, x07 = y07, x08 = y08, x09 = y09, x10 = y10, x11 = y11, x12 = y12, x13 = y13, x14 = y14, x15 = y15;
+    for (let r = 0; r < rounds; r += 2) {
+        x00 = (x00 + x04) | 0;
+        x12 = rotl(x12 ^ x00, 16);
+        x08 = (x08 + x12) | 0;
+        x04 = rotl(x04 ^ x08, 12);
+        x00 = (x00 + x04) | 0;
+        x12 = rotl(x12 ^ x00, 8);
+        x08 = (x08 + x12) | 0;
+        x04 = rotl(x04 ^ x08, 7);
+        x01 = (x01 + x05) | 0;
+        x13 = rotl(x13 ^ x01, 16);
+        x09 = (x09 + x13) | 0;
+        x05 = rotl(x05 ^ x09, 12);
+        x01 = (x01 + x05) | 0;
+        x13 = rotl(x13 ^ x01, 8);
+        x09 = (x09 + x13) | 0;
+        x05 = rotl(x05 ^ x09, 7);
+        x02 = (x02 + x06) | 0;
+        x14 = rotl(x14 ^ x02, 16);
+        x10 = (x10 + x14) | 0;
+        x06 = rotl(x06 ^ x10, 12);
+        x02 = (x02 + x06) | 0;
+        x14 = rotl(x14 ^ x02, 8);
+        x10 = (x10 + x14) | 0;
+        x06 = rotl(x06 ^ x10, 7);
+        x03 = (x03 + x07) | 0;
+        x15 = rotl(x15 ^ x03, 16);
+        x11 = (x11 + x15) | 0;
+        x07 = rotl(x07 ^ x11, 12);
+        x03 = (x03 + x07) | 0;
+        x15 = rotl(x15 ^ x03, 8);
+        x11 = (x11 + x15) | 0;
+        x07 = rotl(x07 ^ x11, 7);
+        x00 = (x00 + x05) | 0;
+        x15 = rotl(x15 ^ x00, 16);
+        x10 = (x10 + x15) | 0;
+        x05 = rotl(x05 ^ x10, 12);
+        x00 = (x00 + x05) | 0;
+        x15 = rotl(x15 ^ x00, 8);
+        x10 = (x10 + x15) | 0;
+        x05 = rotl(x05 ^ x10, 7);
+        x01 = (x01 + x06) | 0;
+        x12 = rotl(x12 ^ x01, 16);
+        x11 = (x11 + x12) | 0;
+        x06 = rotl(x06 ^ x11, 12);
+        x01 = (x01 + x06) | 0;
+        x12 = rotl(x12 ^ x01, 8);
+        x11 = (x11 + x12) | 0;
+        x06 = rotl(x06 ^ x11, 7);
+        x02 = (x02 + x07) | 0;
+        x13 = rotl(x13 ^ x02, 16);
+        x08 = (x08 + x13) | 0;
+        x07 = rotl(x07 ^ x08, 12);
+        x02 = (x02 + x07) | 0;
+        x13 = rotl(x13 ^ x02, 8);
+        x08 = (x08 + x13) | 0;
+        x07 = rotl(x07 ^ x08, 7);
+        x03 = (x03 + x04) | 0;
+        x14 = rotl(x14 ^ x03, 16);
+        x09 = (x09 + x14) | 0;
+        x04 = rotl(x04 ^ x09, 12);
+        x03 = (x03 + x04) | 0;
+        x14 = rotl(x14 ^ x03, 8);
+        x09 = (x09 + x14) | 0;
+        x04 = rotl(x04 ^ x09, 7);
+    }
+    // RFC 8439 §2.3 / §2.3.1: add the original state words back in state order.
+    let oi = 0;
+    out[oi++] = (y00 + x00) | 0;
+    out[oi++] = (y01 + x01) | 0;
+    out[oi++] = (y02 + x02) | 0;
+    out[oi++] = (y03 + x03) | 0;
+    out[oi++] = (y04 + x04) | 0;
+    out[oi++] = (y05 + x05) | 0;
+    out[oi++] = (y06 + x06) | 0;
+    out[oi++] = (y07 + x07) | 0;
+    out[oi++] = (y08 + x08) | 0;
+    out[oi++] = (y09 + x09) | 0;
+    out[oi++] = (y10 + x10) | 0;
+    out[oi++] = (y11 + x11) | 0;
+    out[oi++] = (y12 + x12) | 0;
+    out[oi++] = (y13 + x13) | 0;
+    out[oi++] = (y14 + x14) | 0;
+    out[oi++] = (y15 + x15) | 0;
+}
+/**
+ * hchacha hashes key and nonce into key' and nonce' for xchacha20.
+ * Algorithmically identical to `hchacha_small` from `test/misc/micro-ciphers.ts`,
+ * but this exported path normalizes word order on big-endian hosts.
+ * Reuses `chachaCore` and subtracts its feed-forward, keeping the hot per-block
+ * path untouched.
+ * @param s - Sigma constants as 32-bit words.
+ * @param k - Key words.
+ * @param i - Nonce-prefix words.
+ * @param out - Output buffer for the derived subkey.
+ * @example
+ * Derives the XChaCha subkey from sigma, key, and nonce-prefix words.
+ *
+ * ```ts
+ * const sigma = new Uint32Array(4);
+ * const key = new Uint32Array(8);
+ * const nonce = new Uint32Array(4);
+ * const out = new Uint32Array(8);
+ * hchacha(sigma, key, nonce, out);
+ * ```
+ */
+// prettier-ignore
+export function hchacha(s, k, i, out) {
+    // Runs the shared chachaCore permutation, then subtracts the RFC 8439 feed-forward
+    // it applies, recovering the raw permutation words hchacha needs.
+    // LE hosts read the caller arrays in place (no copies); BE hosts get
+    // byte-swapped scratch copies, wiped before returning.
+    const s2 = isLE ? s : swap32IfBE(s.slice(0, 4));
+    const k2 = isLE ? k : swap32IfBE(k.slice(0, 8));
+    const i2 = isLE ? i : swap32IfBE(i.slice(0, 4));
+    const t = new Uint32Array(16);
+    chachaCore(s2, k2, i2.subarray(1), t, i2[0]);
+    let oi = 0;
+    out[oi++] = (t[0] - s2[0]) | 0;
+    out[oi++] = (t[1] - s2[1]) | 0;
+    out[oi++] = (t[2] - s2[2]) | 0;
+    out[oi++] = (t[3] - s2[3]) | 0;
+    out[oi++] = (t[12] - i2[0]) | 0;
+    out[oi++] = (t[13] - i2[1]) | 0;
+    out[oi++] = (t[14] - i2[2]) | 0;
+    out[oi++] = (t[15] - i2[3]) | 0;
+    swap32IfBE(out);
+    if (!isLE)
+        clean(s2, k2, i2);
+    clean(t);
+}
+/**
+ * Original, non-RFC chacha20 from DJB. 8-byte nonce, 8-byte counter.
+ * The nonce/counter layout still reserves 8 counter bytes internally, but the shared public
+ * `counter` argument follows noble's strict non-wrapping 32-bit policy. See `src/_arx.ts`
+ * near `MAX_COUNTER` for the full counter-policy rationale.
+ * @param key - 16-byte or 32-byte key.
+ * @param nonce - 8-byte nonce.
+ * @param data - Input bytes to xor with the keystream.
+ * @param output - Optional destination buffer.
+ * @param counter - Initial block counter.
+ * @returns Encrypted or decrypted bytes.
+ * @example
+ * Encrypts bytes with the original 8-byte-nonce ChaCha variant and a fresh key/nonce.
+ *
+ * ```ts
+ * import { chacha20orig } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(8);
+ * chacha20orig(key, nonce, new Uint8Array(4));
+ * ```
+ */
+export const chacha20orig = /* @__PURE__ */ createCipher(chachaCore, {
+    counterRight: false,
+    counterLength: 8,
+    allowShortKeys: true,
+});
+/**
+ * ChaCha stream cipher. Conforms to RFC 8439 (IETF, TLS). 12-byte nonce, 4-byte counter.
+ * With smaller nonce, it's not safe to make it random (CSPRNG), due to collision chance.
+ * @param key - 32-byte key.
+ * @param nonce - 12-byte nonce.
+ * @param data - Input bytes to xor with the keystream.
+ * @param output - Optional destination buffer.
+ * @param counter - Initial block counter.
+ * @returns Encrypted or decrypted bytes.
+ * @example
+ * Encrypts bytes with the RFC 8439 ChaCha20 stream cipher and a fresh key/nonce.
+ *
+ * ```ts
+ * import { chacha20 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(12);
+ * chacha20(key, nonce, new Uint8Array(4));
+ * ```
+ */
+export const chacha20 = /* @__PURE__ */ createCipher(chachaCore, {
+    counterRight: false,
+    counterLength: 4,
+    allowShortKeys: false,
+});
+/**
+ * XChaCha eXtended-nonce ChaCha. With 24-byte nonce, it's safe to make it random (CSPRNG).
+ * See {@link https://datatracker.ietf.org/doc/html/draft-irtf-cfrg-xchacha | the IRTF draft}.
+ * The nonce/counter layout still reserves 8 counter bytes internally, but the shared public
+ * `counter` argument follows noble's strict non-wrapping 32-bit policy. See `src/_arx.ts`
+ * near `MAX_COUNTER` for the full counter-policy rationale.
+ * @param key - 32-byte key.
+ * @param nonce - 24-byte extended nonce.
+ * @param data - Input bytes to xor with the keystream.
+ * @param output - Optional destination buffer.
+ * @param counter - Initial block counter.
+ * @returns Encrypted or decrypted bytes.
+ * @example
+ * Encrypts bytes with XChaCha20 using a fresh key and random 24-byte nonce.
+ *
+ * ```ts
+ * import { xchacha20 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(24);
+ * xchacha20(key, nonce, new Uint8Array(4));
+ * ```
+ */
+export const xchacha20 = /* @__PURE__ */ createCipher(chachaCore, {
+    counterRight: false,
+    counterLength: 8,
+    extendNonceFn: hchacha,
+    allowShortKeys: false,
+});
+/**
+ * Reduced 8-round chacha, described in original paper.
+ * @param key - 32-byte key.
+ * @param nonce - 12-byte nonce.
+ * @param data - Input bytes to xor with the keystream.
+ * @param output - Optional destination buffer.
+ * @param counter - Initial block counter.
+ * @returns Encrypted or decrypted bytes.
+ * @example
+ * Uses the reduced 8-round variant for non-critical workloads with a fresh key/nonce.
+ *
+ * ```ts
+ * import { chacha8 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(12);
+ * chacha8(key, nonce, new Uint8Array(4));
+ * ```
+ */
+export const chacha8 = /* @__PURE__ */ createCipher(chachaCore, {
+    counterRight: false,
+    counterLength: 4,
+    rounds: 8,
+});
+/**
+ * Reduced 12-round chacha, described in original paper.
+ * @param key - 32-byte key.
+ * @param nonce - 12-byte nonce.
+ * @param data - Input bytes to xor with the keystream.
+ * @param output - Optional destination buffer.
+ * @param counter - Initial block counter.
+ * @returns Encrypted or decrypted bytes.
+ * @example
+ * Uses the reduced 12-round variant for non-critical workloads with a fresh key/nonce.
+ *
+ * ```ts
+ * import { chacha12 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(12);
+ * chacha12(key, nonce, new Uint8Array(4));
+ * ```
+ */
+export const chacha12 = /* @__PURE__ */ createCipher(chachaCore, {
+    counterRight: false,
+    counterLength: 4,
+    rounds: 12,
+});
+// Test-only hook: exposes the unrolled production core so tests can compare it
+// with the simple/reference core from `test/misc/micro-ciphers.ts`.
+export const __TESTS = /* @__PURE__ */ Object.freeze({ chachaCore });
+// RFC 8439 §2.8.1 pad16(x): shared zero block for AAD/ciphertext padding.
+const ZEROS16 = /* @__PURE__ */ new Uint8Array(16);
+// RFC 8439 §2.8 / §2.8.1: aligned inputs add nothing, otherwise append 16-(len%16) zero bytes.
+const updatePadded = (h, msg) => {
+    h.update(msg);
+    const leftover = msg.length % 16;
+    if (leftover)
+        h.update(ZEROS16.subarray(leftover));
+};
+// RFC 8439 §2.6.1 poly1305_key_gen returns `block[0..31]`, so AEAD key
+// generation only needs 32 zero bytes.
+const ZEROS32 = /* @__PURE__ */ new Uint8Array(32);
+function computeTag(fn, key, nonce, ciphertext, AAD) {
+    if (AAD !== undefined)
+        abytes(AAD, undefined, 'AAD');
+    // RFC 8439 §2.6 / §2.8: derive the Poly1305 one-time key from counter 0,
+    // then MAC AAD || pad16(AAD) || ciphertext || pad16(ciphertext) || len(AAD) || len(ciphertext).
+    const authKey = fn(key, nonce, ZEROS32);
+    const lengths = u64Lengths(ciphertext.length, AAD ? AAD.length : 0, true);
+    // Methods below can be replaced with
+    // `return poly1305_computeTag_small(authKey, lengths, ciphertext, AAD)`
+    // from `test/misc/micro-ciphers.ts`.
+    const h = poly1305.create(authKey);
+    if (AAD)
+        updatePadded(h, AAD);
+    updatePadded(h, ciphertext);
+    h.update(lengths);
+    const res = h.digest();
+    clean(authKey, lengths);
+    return res;
+}
+/**
+ * AEAD algorithm from RFC 8439.
+ * Salsa20 and chacha (RFC 8439) use poly1305 differently.
+ * We could have composed them, but it's hard because of authKey:
+ * In salsa20, authKey changes position in salsa stream.
+ * In chacha, authKey can't be computed inside computeTag, it modifies the counter.
+ */
+export const _poly1305_aead = (xorStream) => (key, nonce, AAD) => {
+    // This borrows caller key/nonce/AAD buffers by reference; mutating them after construction
+    // changes future encrypt/decrypt results.
+    const tagLength = 16;
+    return {
+        encrypt(plaintext, output) {
+            const plength = plaintext.length;
+            output = getOutput(plength + tagLength, output, false);
+            output.set(plaintext);
+            const oPlain = output.subarray(0, -tagLength);
+            // RFC 8439 §2.8: payload encryption starts at counter 1 because counter 0 produced the OTK.
+            xorStream(key, nonce, oPlain, oPlain, 1);
+            const tag = computeTag(xorStream, key, nonce, oPlain, AAD);
+            output.set(tag, plength); // append tag
+            clean(tag);
+            return output;
+        },
+        decrypt(ciphertext, output) {
+            output = getOutput(ciphertext.length - tagLength, output, false);
+            const data = ciphertext.subarray(0, -tagLength);
+            const passedTag = ciphertext.subarray(-tagLength);
+            const tag = computeTag(xorStream, key, nonce, data, AAD);
+            // RFC 8439 §2.8 / §4: authenticate ciphertext before decrypting it, and compare tags with
+            // the constant-time equalBytes() helper rather than decrypting speculative plaintext first.
+            if (!equalBytes(passedTag, tag)) {
+                clean(tag);
+                throw new Error('invalid tag');
+            }
+            output.set(ciphertext.subarray(0, -tagLength));
+            // Actual decryption
+            xorStream(key, nonce, output, output, 1); // start stream with i=1
+            clean(tag);
+            return output;
+        },
+    };
+};
+/**
+ * ChaCha20-Poly1305 from RFC 8439.
+ *
+ * Unsafe to use random nonces under the same key, due to collision chance.
+ * Prefer XChaCha instead.
+ * @param key - 32-byte key.
+ * @param nonce - 12-byte nonce.
+ * @param AAD - Additional authenticated data.
+ * @returns AEAD cipher instance.
+ * @example
+ * Encrypts and authenticates plaintext with a fresh key and nonce.
+ *
+ * ```ts
+ * import { chacha20poly1305 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(12);
+ * const aad = new TextEncoder().encode('session metadata');
+ * const cipher = chacha20poly1305(key, nonce, aad);
+ * cipher.encrypt(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export const chacha20poly1305 = /* @__PURE__ */ wrapCipher({ blockSize: 64, nonceLength: 12, tagLength: 16, withAAD: true }, 
+/* @__PURE__ */ _poly1305_aead(chacha20));
+/**
+ * XChaCha20-Poly1305 extended-nonce chacha.
+ *
+ * Can be safely used with random nonces (CSPRNG).
+ * See {@link https://datatracker.ietf.org/doc/html/draft-irtf-cfrg-xchacha | the IRTF draft}.
+ * @param key - 32-byte key.
+ * @param nonce - 24-byte nonce.
+ * @param AAD - Additional authenticated data.
+ * @returns AEAD cipher instance.
+ * @example
+ * Encrypts and authenticates plaintext with a fresh key and random 24-byte nonce.
+ *
+ * ```ts
+ * import { xchacha20poly1305 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const key = randomBytes(32);
+ * const nonce = randomBytes(24);
+ * const aad = new TextEncoder().encode('session metadata');
+ * const cipher = xchacha20poly1305(key, nonce, aad);
+ * cipher.encrypt(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export const xchacha20poly1305 = /* @__PURE__ */ wrapCipher({ blockSize: 64, nonceLength: 24, tagLength: 16, withAAD: true }, 
+/* @__PURE__ */ _poly1305_aead(xchacha20));
+/**
+ * Chacha20 CSPRNG (cryptographically secure pseudorandom number generator).
+ * It's best to limit usage to non-production, non-critical cases: for example, test-only.
+ * Compatible with libtomcrypt. It does not have a specification, so unclear how secure it is.
+ * @param seed - Optional seed bytes mixed into the internal `key || nonce` state. When omitted,
+ * only 32 random bytes are mixed into the 40-byte state.
+ * @returns Seeded concrete `_XorStreamPRG` instance, including `clone()`.
+ * @example
+ * Seeds the test-only ChaCha20 DRBG from fresh entropy.
+ *
+ * ```ts
+ * import { rngChacha20 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const seed = randomBytes(32);
+ * const prg = rngChacha20(seed);
+ * prg.randomBytes(8);
+ * ```
+ */
+export const rngChacha20 = /* @__PURE__ */ createPRG(chacha20orig, 64, 32, 8);
+/**
+ * Chacha20/8 CSPRNG (cryptographically secure pseudorandom number generator).
+ * It's best to limit usage to non-production, non-critical cases: for example, test-only.
+ * Faster than `rngChacha20`.
+ * @param seed - Optional seed bytes mixed into the internal `key || nonce` state. When omitted,
+ * only 32 random bytes are mixed into the 44-byte state.
+ * @returns Seeded concrete `_XorStreamPRG` instance, including `clone()`.
+ * @example
+ * Seeds the faster test-only ChaCha8 DRBG from fresh entropy.
+ *
+ * ```ts
+ * import { rngChacha8 } from '@noble/ciphers/chacha.js';
+ * import { randomBytes } from '@noble/ciphers/utils.js';
+ * const seed = randomBytes(32);
+ * const prg = rngChacha8(seed);
+ * prg.randomBytes(8);
+ * ```
+ */
+export const rngChacha8 = /* @__PURE__ */ createPRG(chacha8, 64, 32, 12);

backend/static/vendor/noble-ciphers/utils.js

@@ -0,0 +1,896 @@
+/**
+ * Utilities for hex, bytes, CSPRNG.
+ * @module
+ */
+/*! noble-ciphers - MIT License (c) 2023 Paul Miller (paulmillr.com) */
+export function aarray(item, title, inner = () => { }) {
+    if (!Array.isArray(item))
+        throw new TypeError(`"${title}" expected array, got type=${typeof item}`);
+    for (let i = 0; i < item.length; i++)
+        inner(item[i], `${title}[${i}]`);
+    return item;
+}
+/**
+ * Checks if something is Uint8Array. Be careful: nodejs Buffer will return true.
+ * @param a - Value to inspect.
+ * @returns `true` when the value is a Uint8Array view, including Node's `Buffer`.
+ * @example
+ * Guards a value before treating it as raw key material.
+ *
+ * ```ts
+ * isBytes(new Uint8Array());
+ * ```
+ */
+export function isBytes(a) {
+    // Plain `instanceof Uint8Array` is too strict for some Buffer / proxy /
+    // cross-realm cases. The fallback still requires a real ArrayBuffer view
+    // so plain JSON-deserialized `{ constructor: ... }`
+    // spoofing is rejected, and `BYTES_PER_ELEMENT === 1` keeps the fallback on byte-oriented views.
+    return (a instanceof Uint8Array ||
+        (ArrayBuffer.isView(a) &&
+            a.constructor.name === 'Uint8Array' &&
+            'BYTES_PER_ELEMENT' in a &&
+            a.BYTES_PER_ELEMENT === 1));
+}
+// Shared error-message prefix builder. Only called on throw paths, so assert
+// success paths never pay for the string concatenation.
+const atitle = (title) => (title ? `"${title}" ` : '');
+/**
+ * Asserts something is boolean.
+ * @param value - Value to validate.
+ * @returns The validated boolean.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Validates a boolean option before branching on it.
+ *
+ * ```ts
+ * abool(true);
+ * ```
+ */
+export function abool(value, title = '') {
+    if (typeof value !== 'boolean')
+        throw new TypeError(atitle(title) + 'expected boolean, got type=' + typeof value);
+    return value;
+}
+/**
+ * Asserts something is a non-negative safe integer.
+ * @param n - Value to validate.
+ * @returns The validated number.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Validates a non-negative length or counter.
+ *
+ * ```ts
+ * anumber(1);
+ * ```
+ */
+export function anumber(n, title = '') {
+    if (typeof n !== 'number')
+        throw new TypeError(atitle(title) + 'expected number, got ' + typeof n);
+    if (!Number.isSafeInteger(n) || n < 0)
+        throw new RangeError(atitle(title) + 'expected integer >= 0, got ' + n);
+    return n;
+}
+/**
+ * Asserts something is Uint8Array.
+ * @param value - Value to validate.
+ * @param length - Expected byte length.
+ * @param title - Optional label used in error messages.
+ * @returns The validated byte array.
+ * On Node, `Buffer` is accepted too because it is a Uint8Array view.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument lengths. {@link RangeError}
+ * @example
+ * Validates a fixed-length nonce or key buffer.
+ *
+ * ```ts
+ * abytes(new Uint8Array([1, 2]), 2);
+ * ```
+ */
+export function abytes(value, length, title = '') {
+    // Success path first: this runs at the start of every update() / digestInto(), and the
+    // common `abytes(data)` form must not pay for length handling it does not use.
+    if (isBytes(value) && (length === undefined || value.length === length))
+        return value;
+    // Error path: recompute freely to build the exact message.
+    if (length !== undefined)
+        anumber(length, 'length');
+    const bytes = isBytes(value);
+    const ofLen = length !== undefined ? ` of length ${length}` : '';
+    const got = bytes ? `length=${value.length}` : `type=${typeof value}`;
+    const message = atitle(title) + 'expected Uint8Array' + ofLen + ', got ' + got;
+    if (!bytes)
+        throw new TypeError(message);
+    throw new RangeError(message);
+}
+const aobject = (value, label) => {
+    if (value === null || typeof value !== 'object' || Array.isArray(value))
+        throw new TypeError(label === 'object'
+            ? 'expected valid options object'
+            : `"${label}" expected object, got type=${typeof value}`);
+};
+/**
+ * Asserts a hash- or MAC-like instance has not been destroyed or finished.
+ * @param instance - Stateful instance to validate.
+ * @param checkFinished - Whether to reject finished instances.
+ * When `false`, only `destroyed` is checked.
+ * @throws If the hash instance has already been destroyed or finalized. {@link Error}
+ * @example
+ * Guards against calling `update()` or `digest()` on a finished hash.
+ *
+ * ```ts
+ * aexists({ destroyed: false, finished: false });
+ * ```
+ */
+export function aexists(instance, checkFinished = true) {
+    // Runs on every update()/digestInto(); the flags are library-owned booleans, so only their
+    // truthiness is checked - re-validating their type per call was pure hot-path overhead.
+    if (instance.destroyed)
+        throw new Error('hash was destroyed');
+    if (checkFinished && instance.finished)
+        throw new Error('digest() was already called');
+}
+/**
+ * Asserts output is a sufficiently-sized byte array.
+ * @param out - Output buffer to validate.
+ * @param instance - Hash-like instance providing `outputLen`.
+ * This is the relaxed `digestInto()`-style contract: output must be at least `outputLen`,
+ * unlike one-shot cipher helpers elsewhere in the repo that often require exact lengths.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong output buffer lengths. {@link RangeError}
+ * @example
+ * Verifies that a caller-provided output buffer is large enough.
+ *
+ * ```ts
+ * aoutput(new Uint8Array(16), { outputLen: 16 });
+ * ```
+ */
+export function aoutput(out, instance) {
+    abytes(out, undefined, 'output');
+    // `outputLen` is a library-owned readonly number; the negated comparison keeps failing fast
+    // when it is missing/NaN (comparisons with undefined/NaN are false) without an anumber() call.
+    const min = instance.outputLen;
+    if (!(out.length >= min)) {
+        throw new RangeError('"output" expected length >= ' + min);
+    }
+}
+/**
+ * Asserts output is a sufficiently-sized, 4-byte-aligned byte array.
+ * {@link aoutput} plus an {@link isAligned32} check, for `digestInto()` paths
+ * that write through zero-allocation `u32` word views.
+ * @param out - Output buffer to validate.
+ * @param instance - Hash-like instance providing `outputLen`.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong output buffer lengths. {@link RangeError}
+ * @throws On wrong output buffer alignment. {@link Error}
+ * @example
+ * Verifies that a caller-provided output buffer is large enough and aligned.
+ *
+ * ```ts
+ * aoutput32(new Uint8Array(16), { outputLen: 16 });
+ * ```
+ */
+export function aoutput32(out, instance) {
+    aoutput(out, instance);
+    if (!isAligned32(out))
+        throw new Error('invalid output, must be aligned');
+}
+/**
+ * Casts a typed-array view to Uint8Array.
+ * @param arr - Typed-array view to reinterpret.
+ * @returns Uint8Array view over the same bytes.
+ * @example
+ * Views 32-bit words as raw bytes without copying.
+ *
+ * ```ts
+ * u8(new Uint32Array([1]));
+ * ```
+ */
+export function u8(arr) {
+    return new Uint8Array(arr.buffer, arr.byteOffset, arr.byteLength);
+}
+/**
+ * Casts a typed-array view to Uint32Array.
+ * @param arr - Typed-array view to reinterpret.
+ * @returns Uint32Array view over the same bytes. Callers are expected to provide a
+ * 4-byte-aligned offset; trailing `1..3` bytes are silently dropped.
+ * @example
+ * Views a byte buffer as 32-bit words for block processing.
+ *
+ * ```ts
+ * u32(new Uint8Array(4));
+ * ```
+ */
+export function u32(arr) {
+    return new Uint32Array(arr.buffer, arr.byteOffset, Math.floor(arr.byteLength / 4));
+}
+/**
+ * Zeroizes typed arrays in place.
+ * Warning: JS provides no guarantees.
+ * @param arrays - Arrays to wipe.
+ * @example
+ * Wipes a temporary key buffer after use.
+ *
+ * ```ts
+ * const bytes = new Uint8Array([1]);
+ * clean(bytes);
+ * ```
+ */
+export function clean(...arrays) {
+    for (let i = 0; i < arrays.length; i++) {
+        arrays[i].fill(0);
+    }
+}
+/**
+ * Creates a DataView for byte-level manipulation.
+ * @param arr - Typed-array view to wrap.
+ * @returns DataView over the same bytes.
+ * @example
+ * Creates an endian-aware view for length encoding.
+ *
+ * ```ts
+ * createView(new Uint8Array(4));
+ * ```
+ */
+export function createView(arr) {
+    return new DataView(arr.buffer, arr.byteOffset, arr.byteLength);
+}
+/**
+ * Whether the current platform is little-endian.
+ * Most are; some IBM systems are not.
+ */
+export const isLE = /* @__PURE__ */ (() => new Uint8Array(new Uint32Array([0x11223344]).buffer)[0] === 0x44)();
+/**
+ * Reverses byte order of one 32-bit word.
+ * @param word - Unsigned 32-bit word to swap.
+ * @returns The same word with bytes reversed.
+ * @example
+ * Swaps a big-endian word into little-endian byte order.
+ *
+ * ```ts
+ * byteSwap(0x11223344);
+ * ```
+ */
+export function byteSwap(word) {
+    return (((word << 24) & 0xff000000) |
+        ((word << 8) & 0xff0000) |
+        ((word >>> 8) & 0xff00) |
+        ((word >>> 24) & 0xff));
+}
+/**
+ * Normalizes one 32-bit word to the little-endian representation expected by cipher cores.
+ * @param n - Unsigned 32-bit word to normalize.
+ * @returns Little-endian normalized word on big-endian hosts, else the input word unchanged.
+ * @example
+ * Normalizes a host-endian word before passing it into an ARX/AES core.
+ *
+ * ```ts
+ * swap8IfBE(0x11223344);
+ * ```
+ */
+export const swap8IfBE = isLE
+    ? (n) => n
+    : (n) => byteSwap(n) >>> 0;
+/**
+ * Byte-swaps every word of a Uint32Array in place.
+ * @param arr - Uint32Array whose words should be swapped.
+ * @returns The same array after in-place byte swapping.
+ * @example
+ * Swaps every 32-bit word in a word-view buffer.
+ *
+ * ```ts
+ * byteSwap32(new Uint32Array([0x11223344]));
+ * ```
+ */
+export function byteSwap32(arr) {
+    for (let i = 0; i < arr.length; i++) {
+        arr[i] = byteSwap(arr[i]);
+    }
+    return arr;
+}
+/**
+ * Normalizes a Uint32Array view to the little-endian representation expected by cipher cores.
+ * @param u - Word view to normalize in place.
+ * @returns Little-endian normalized word view.
+ * @example
+ * Normalizes a word-view buffer before block processing.
+ *
+ * ```ts
+ * swap32IfBE(new Uint32Array([0x11223344]));
+ * ```
+ */
+export const swap32IfBE = isLE
+    ? (u) => u
+    : byteSwap32;
+// Built-in hex conversion:
+// {@link https://caniuse.com/mdn-javascript_builtins_uint8array_fromhex | caniuse entry}
+const hasHexBuiltin = /* @__PURE__ */ (() => 
+// @ts-ignore
+typeof Uint8Array.from([]).toHex === 'function' && typeof Uint8Array.fromHex === 'function')();
+// Array where index 0xf0 (240) is mapped to string 'f0'
+const hexes = /* @__PURE__ */ Array.from({ length: 256 }, (_, i) => i.toString(16).padStart(2, '0'));
+/**
+ * Convert byte array to hex string. Uses built-in function, when available.
+ * @param bytes - Bytes to encode.
+ * @returns Lowercase hexadecimal string.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Formats ciphertext bytes for logs or test vectors.
+ *
+ * ```ts
+ * bytesToHex(Uint8Array.from([0xca, 0xfe, 0x01, 0x23])); // 'cafe0123'
+ * ```
+ */
+export function bytesToHex(bytes) {
+    abytes(bytes);
+    // @ts-ignore
+    if (hasHexBuiltin)
+        return bytes.toHex();
+    // pre-caching improves the speed 6x
+    let hex = '';
+    for (let i = 0; i < bytes.length; i++) {
+        hex += hexes[bytes[i]];
+    }
+    return hex;
+}
+// Strict ASCII nibble parser: non-ASCII hex lookalikes are rejected as undefined.
+// ASCII codes: '0'..'9' = 48..57, 'A'..'F' = 65..70, 'a'..'f' = 97..102.
+// prettier-ignore
+function asciiToBase16(ch) {
+    return ch >= 48 && ch <= 57 ? ch - 48 // '2' => 50-48
+        : ch >= 65 && ch <= 70 ? ch - (65 - 10) // 'B' => 66-(65-10)
+            : ch >= 97 && ch <= 102 ? ch - (97 - 10) // 'b' => 98-(97-10)
+                : undefined;
+}
+/**
+ * Convert hex string to byte array. Uses built-in function, when available.
+ * @param hex - hexadecimal string to decode
+ * @returns Decoded bytes.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Decode lowercase hexadecimal into bytes.
+ * ```ts
+ * hexToBytes('cafe0123'); // Uint8Array.from([0xca, 0xfe, 0x01, 0x23])
+ * ```
+ */
+export function hexToBytes(hex) {
+    if (typeof hex !== 'string')
+        throw new TypeError('hex string expected, got ' + typeof hex);
+    if (hasHexBuiltin) {
+        try {
+            return Uint8Array.fromHex(hex);
+        }
+        catch (error) {
+            if (error instanceof SyntaxError)
+                throw new RangeError(error.message);
+            throw error;
+        }
+    }
+    const hl = hex.length;
+    const al = hl / 2;
+    if (hl % 2)
+        throw new RangeError('hex string expected, got unpadded hex of length ' + hl);
+    const array = new Uint8Array(al);
+    for (let ai = 0, hi = 0; ai < al; ai++, hi += 2) {
+        const n1 = asciiToBase16(hex.charCodeAt(hi)); // parse first char, multiply it by 16
+        const n2 = asciiToBase16(hex.charCodeAt(hi + 1)); // parse second char
+        if (n1 === undefined || n2 === undefined) {
+            const char = hex[hi] + hex[hi + 1];
+            throw new RangeError('hex string expected, got non-hex character "' + char + '" at index ' + hi);
+        }
+        array[ai] = n1 * 16 + n2; // example: 'A9' => 10*16 + 9
+    }
+    return array;
+}
+const _0n = /* @__PURE__ */ BigInt(0);
+// Used in ff1, via bytesToNumberBE
+/**
+ * Converts a big-endian hex string into bigint.
+ * @param hex - Hexadecimal string without `0x`.
+ * @returns Parsed bigint value. The empty string is treated as `0n`.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Parses a big-endian field element or counter from hex.
+ *
+ * ```ts
+ * hexToNumber('ff');
+ * ```
+ */
+export function hexToNumber(hex) {
+    if (typeof hex !== 'string')
+        throw new TypeError('hex string expected, got ' + typeof hex);
+    // Numeric parser, not byte-hex decoder: odd-length forms like 'f' are valid,
+    // and malformed syntax follows BigInt's native error behavior.
+    return hex === '' ? _0n : BigInt('0x' + hex); // Big Endian
+}
+// Used in ff1
+// BE: Big Endian, LE: Little Endian
+/**
+ * Converts big-endian bytes into bigint.
+ * @param bytes - Big-endian bytes.
+ * @returns Parsed bigint value. Empty input is treated as `0n`.
+ * @throws On invalid byte input passed to the internal hex conversion. {@link TypeError}
+ * @example
+ * Reads a big-endian integer from serialized bytes.
+ *
+ * ```ts
+ * bytesToNumberBE(new Uint8Array([1, 0]));
+ * ```
+ */
+export function bytesToNumberBE(bytes) {
+    return hexToNumber(bytesToHex(bytes));
+}
+/**
+ * Validates that a value is a non-negative bigint or safe integer.
+ * @param n - Value to validate.
+ * @returns The same validated value.
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ */
+function abignumber(n) {
+    if (typeof n === 'bigint') {
+        if (!(_0n <= n))
+            throw new RangeError('positive bigint expected, got ' + n);
+    }
+    else
+        anumber(n);
+    return n;
+}
+// Used in ff1
+/**
+ * Converts a number into big-endian bytes of fixed length.
+ * @param n - Number to encode.
+ * @param len - Output length in bytes. Must be greater than zero.
+ * @returns Big-endian bytes padded to `len`.
+ * Negative values, `len = 0`, and values that do not fit are rejected before
+ * downstream hex parsing.
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @throws If a documented runtime validation or state check fails. {@link Error}
+ * @example
+ * Encodes a counter as fixed-width big-endian bytes.
+ *
+ * ```ts
+ * numberToBytesBE(1, 2);
+ * ```
+ */
+export function numberToBytesBE(n, len) {
+    anumber(len);
+    if (len === 0)
+        throw new Error('zero output length is invalid');
+    n = abignumber(n);
+    const expectedLen = len * 2;
+    const hex = n.toString(16);
+    // Detect overflow before hex parsing so oversized values don't leak the shared odd-hex error.
+    if (hex.length > expectedLen)
+        throw new RangeError('number is too large');
+    return hexToBytes(hex.padStart(expectedLen, '0'));
+}
+/**
+ * Converts string to bytes using UTF8 encoding.
+ * @param str - String to encode.
+ * @returns UTF-8 bytes in a detached fresh Uint8Array copy.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Encodes application text before encryption or MACing.
+ *
+ * ```ts
+ * utf8ToBytes('abc'); // new Uint8Array([97, 98, 99])
+ * ```
+ */
+export function utf8ToBytes(str) {
+    if (typeof str !== 'string')
+        throw new TypeError('string expected');
+    return new Uint8Array(new TextEncoder().encode(str)); // {@link https://bugzil.la/1681809 | Firefox bug 1681809}
+}
+/**
+ * Converts bytes to string using UTF8 encoding.
+ * @param bytes - UTF-8 bytes.
+ * @returns Decoded string. Input validation is delegated to `TextDecoder`, and malformed
+ * UTF-8 is replacement-decoded instead of rejected.
+ * @example
+ * Decodes UTF-8 plaintext back into a string.
+ *
+ * ```ts
+ * bytesToUtf8(new Uint8Array([97, 98, 99])); // 'abc'
+ * ```
+ */
+export function bytesToUtf8(bytes) {
+    return new TextDecoder().decode(bytes);
+}
+/**
+ * Checks if two U8A use same underlying buffer and overlaps.
+ * This is invalid and can corrupt data.
+ * @param a - First byte view.
+ * @param b - Second byte view.
+ * @returns `true` when the views overlap in memory.
+ * @example
+ * Detects whether two slices alias the same backing buffer.
+ *
+ * ```ts
+ * overlapBytes(new Uint8Array(4), new Uint8Array(4));
+ * ```
+ */
+export function overlapBytes(a, b) {
+    // Zero-length views cannot overwrite anything, even if their offset sits inside another range.
+    if (!a.byteLength || !b.byteLength)
+        return false;
+    return (a.buffer === b.buffer && // best we can do, may fail with an obscure Proxy
+        a.byteOffset < b.byteOffset + b.byteLength && // a starts before b end
+        b.byteOffset < a.byteOffset + a.byteLength // b starts before a end
+    );
+}
+/**
+ * If input and output overlap and input starts before output, we will overwrite end of input before
+ * we start processing it, so this is not supported by forward-processing ciphers.
+ * @param input - Input bytes.
+ * @param output - Output bytes.
+ * @throws If the output view would overwrite unread input bytes. {@link Error}
+ * @example
+ * Rejects an in-place layout that would overwrite unread input bytes.
+ *
+ * ```ts
+ * const buffer = new Uint8Array(8);
+ * complexOverlapBytes(buffer.subarray(0, 4), buffer.subarray(2, 6));
+ * ```
+ */
+export function complexOverlapBytes(input, output) {
+    // This is very cursed. It works somehow, but I'm completely unsure,
+    // reasoning about overlapping aligned windows is very hard.
+    if (overlapBytes(input, output) && input.byteOffset < output.byteOffset)
+        throw new Error('complex overlap of input and output is not supported');
+}
+/**
+ * Copies several Uint8Arrays into one.
+ * @param arrays - Byte arrays to concatenate.
+ * @returns Combined byte array.
+ * @throws On wrong argument types inside the byte-array list. {@link TypeError}
+ * @example
+ * Builds a `nonce || ciphertext` style buffer.
+ *
+ * ```ts
+ * concatBytes(new Uint8Array([1]), new Uint8Array([2]));
+ * ```
+ */
+export function concatBytes(...arrays) {
+    let sum = 0;
+    for (let i = 0; i < arrays.length; i++) {
+        const a = arrays[i];
+        abytes(a);
+        sum += a.length;
+    }
+    const res = new Uint8Array(sum);
+    for (let i = 0, pad = 0; i < arrays.length; i++) {
+        const a = arrays[i];
+        res.set(a, pad);
+        pad += a.length;
+    }
+    return res;
+}
+/**
+ * Merges user options into defaults.
+ * @param defaults - Default option values.
+ * @param opts - User-provided overrides.
+ * @returns Combined options object.
+ * `defaults` is a library-owned mutable object; user-provided `opts` only need to be
+ * object-shaped, since "plain object" checks reject valid proxy/cross-realm containers.
+ * The merge mutates `defaults` in place and returns the same object, so direct callers
+ * should pass a fresh defaults object unless they intentionally want shared state updated.
+ * @throws If options are missing or not an object. {@link Error}
+ * @example
+ * Applies user overrides to the default cipher options.
+ *
+ * ```ts
+ * checkOpts({ rounds: 20 }, { rounds: 8 });
+ * ```
+ */
+export function checkOpts(defaults, opts) {
+    aobject(defaults, 'defaults');
+    aobject(opts, 'opts');
+    // Mutates defaults by design. __proto__ follows Object.assign semantics here and can only
+    // affect this local option object; callers already control this low-level options surface.
+    const merged = Object.assign(defaults, opts);
+    return merged;
+}
+/**
+ * Compares two byte arrays in kinda constant time once lengths already match.
+ * @param a - First byte array.
+ * @param b - Second byte array.
+ * @returns `true` when the arrays contain the same bytes. Different lengths still return early.
+ * @example
+ * Compares an expected authentication tag with the received one.
+ *
+ * ```ts
+ * equalBytes(new Uint8Array([1]), new Uint8Array([1]));
+ * ```
+ */
+export function equalBytes(a, b) {
+    a = abytes(a);
+    b = abytes(b);
+    if (a.length !== b.length)
+        return false;
+    let diff = 0;
+    for (let i = 0; i < a.length; i++)
+        diff |= a[i] ^ b[i];
+    return diff === 0;
+}
+/**
+ * Wraps a keyed MAC constructor into a one-shot helper with `.create()`.
+ * @param keyLen - Valid probe-key length used to read static metadata once.
+ * The probe key is only used for `outputLen` / `blockLen`, so callers with several valid key sizes
+ * can pass any representative size as long as those values stay fixed.
+ * @param macCons - Keyed MAC constructor or factory.
+ * @param fromMsg - Optional adapter that derives extra constructor args from the one-shot message.
+ * @returns Callable MAC helper with `.create()`.
+ */
+export function wrapMacConstructor(keyLen, macCons, fromMsg) {
+    const mac = macCons;
+    const getArgs = (fromMsg || (() => []));
+    const macC = (msg, key) => mac(key, ...getArgs(msg))
+        .update(msg)
+        .digest();
+    const tmp = mac(new Uint8Array(keyLen), ...getArgs(new Uint8Array(0)));
+    macC.outputLen = tmp.outputLen;
+    macC.blockLen = tmp.blockLen;
+    macC.create = (key, ...args) => mac(key, ...args);
+    return macC;
+}
+/**
+ * Wraps a cipher: validates args, ensures encrypt() can only be called once.
+ * Used internally by the exported cipher constructors.
+ * Output-buffer support is inferred from the wrapped `encrypt` / `decrypt`
+ * arity (`fn.length === 2`), so wrapped output-capable methods must use a normal
+ * second parameter, not a default/rest parameter. AAD support is explicit in
+ * `params.withAAD`; optional AAD starts after the nonce slot when one is present.
+ * @__NO_SIDE_EFFECTS__
+ * @param params - Static cipher metadata. See {@link CipherParams}.
+ * @param constructor - Cipher constructor.
+ * @returns Wrapped constructor with validation.
+ */
+export const wrapCipher = (params, constructor) => {
+    function wrappedCipher(key, ...args) {
+        // Validate key
+        abytes(key, undefined, 'key');
+        // Validate nonce if nonceLength is present
+        if (params.nonceLength !== undefined) {
+            const nonce = args[0];
+            abytes(nonce, params.varSizeNonce ? undefined : params.nonceLength, 'nonce');
+        }
+        // Keep tag length available for decrypt-size checks after constructor validation.
+        const tagl = params.tagLength;
+        const aadStart = params.nonceLength !== undefined ? 1 : 0;
+        // No-AAD constructors otherwise silently ignore byte args meant as AAD.
+        if (!params.withAAD) {
+            for (let i = aadStart; i < args.length; i++)
+                if (isBytes(args[i]))
+                    throw new Error('AAD not supported');
+        }
+        // Validate the first AAD slot early; rest-arg AAD constructors validate the tail themselves.
+        if (params.withAAD && args[aadStart] !== undefined)
+            abytes(args[aadStart], undefined, 'AAD');
+        const cipher = constructor(key, ...args);
+        const checkOutput = (fnLength, output) => {
+            if (output !== undefined) {
+                if (fnLength !== 2)
+                    throw new Error('cipher output not supported');
+                abytes(output, undefined, 'output');
+            }
+        };
+        // Create wrapped cipher with validation and single-use encryption
+        let called = false;
+        const wrCipher = {
+            encrypt(data, output) {
+                if (called)
+                    throw new Error('cannot encrypt() twice with same key + nonce');
+                // Any encrypt attempt consumes the instance, even if validation rejects below.
+                called = true;
+                abytes(data, undefined, 'data');
+                checkOutput(cipher.encrypt.length, output);
+                return cipher.encrypt(data, output);
+            },
+            decrypt(data, output) {
+                abytes(data, undefined, 'data');
+                if (tagl && data.length < tagl)
+                    throw new Error('"ciphertext" expected length >= tagLength=' + tagl);
+                checkOutput(cipher.decrypt.length, output);
+                return cipher.decrypt(data, output);
+            },
+        };
+        return wrCipher;
+    }
+    Object.assign(wrappedCipher, params);
+    return wrappedCipher;
+};
+/**
+ * By default, returns u8a of length.
+ * When out is available, it checks it for validity and uses it.
+ * @param expectedLength - Required output length.
+ * @param out - Optional destination buffer.
+ * @param onlyAligned - Whether `out` must be 4-byte aligned.
+ * @returns Output buffer ready for writing.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws If the provided output buffer has the wrong size. {@link RangeError}
+ * @throws If the provided output buffer has the wrong alignment. {@link Error}
+ * @example
+ * Reuses a caller-provided output buffer when lengths match.
+ *
+ * ```ts
+ * getOutput(16, new Uint8Array(16));
+ * ```
+ */
+export function getOutput(expectedLength, out, onlyAligned = true) {
+    if (out === undefined)
+        return new Uint8Array(expectedLength);
+    // Keep Buffer/cross-realm Uint8Array support here instead of trusting a shape-compatible object.
+    abytes(out, expectedLength, 'output');
+    if (onlyAligned && !isAligned32(out))
+        throw new Error('invalid output, must be aligned');
+    return out;
+}
+/**
+ * Encodes data and AAD lengths into a 16-byte buffer.
+ * @param dataLength - Data length. Units are caller-defined: GCM passes bit
+ * lengths, ChaCha20-Poly1305 passes byte lengths — the helper writes the raw values.
+ * @param aadLength - AAD length, same unit convention as `dataLength`.
+ * The serialized block is still `aadLength || dataLength`, matching GCM/Poly1305
+ * conventions even though the helper parameter order is `(dataLength, aadLength)`.
+ * @param isLE - Whether to encode lengths as little-endian.
+ * @returns 16-byte length block.
+ * @throws On wrong argument types passed to the endian validator. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Builds the length block appended by GCM and Poly1305.
+ *
+ * ```ts
+ * u64Lengths(16, 8, true);
+ * ```
+ */
+export function u64Lengths(dataLength, aadLength, isLE) {
+    // Reject coercible non-number lengths like '10' and true before BigInt(...) accepts them.
+    anumber(dataLength);
+    anumber(aadLength);
+    abool(isLE);
+    const num = new Uint8Array(16);
+    const view = createView(num);
+    view.setBigUint64(0, BigInt(aadLength), isLE);
+    view.setBigUint64(8, BigInt(dataLength), isLE);
+    return num;
+}
+/**
+ * Checks whether a byte array is aligned to a 4-byte offset.
+ * @param bytes - Byte array to inspect.
+ * @returns `true` when the view is 4-byte aligned.
+ * @example
+ * Checks whether a buffer can be safely viewed as Uint32Array.
+ *
+ * ```ts
+ * isAligned32(new Uint8Array(4));
+ * ```
+ */
+export function isAligned32(bytes) {
+    return bytes.byteOffset % 4 === 0;
+}
+/**
+ * Copies bytes into a new Uint8Array.
+ * @param bytes - Bytes to copy.
+ * @returns Copied byte array.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Copies input into an aligned Uint8Array before block processing.
+ *
+ * ```ts
+ * copyBytes(new Uint8Array([1, 2]));
+ * ```
+ */
+export function copyBytes(bytes) {
+    // `Uint8Array.from(...)` would also accept arrays / other typed arrays. Keep this helper strict
+    // because callers use it at byte-validation boundaries before mutating the detached copy.
+    return Uint8Array.from(abytes(bytes));
+}
+/**
+ * Cryptographically secure PRNG backed by `crypto.getRandomValues`.
+ * @param bytesLength - number of random bytes to generate
+ * @returns Random bytes.
+ * The platform `getRandomValues()` implementation still defines any
+ * single-call length cap, and this helper rejects oversize requests
+ * with a stable library `RangeError` instead of host-specific errors.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @throws If the current runtime does not provide `crypto.getRandomValues`. {@link Error}
+ * @example
+ * Generate a fresh random key or nonce.
+ * ```ts
+ * const key = randomBytes(16);
+ * ```
+ */
+export function randomBytes(bytesLength = 32) {
+    // Match the repo's other length-taking helpers instead of relying on Uint8Array coercion.
+    anumber(bytesLength, 'bytesLength');
+    const cr = typeof globalThis === 'object' ? globalThis.crypto : null;
+    if (typeof cr?.getRandomValues !== 'function')
+        throw new Error('crypto.getRandomValues must be defined');
+    // Web Cryptography API Level 2 §10.1.1:
+    // if `byteLength > 65536`, throw `QuotaExceededError`.
+    // Keep the guard explicit so callers can see the quota in code
+    // instead of discovering it by reading the spec or host errors.
+    // This wrapper surfaces the same quota as a stable library RangeError.
+    if (bytesLength > 65536)
+        throw new RangeError(`"bytesLength" expected <= 65536, got ${bytesLength}`);
+    return cr.getRandomValues(new Uint8Array(bytesLength));
+}
+/**
+ * Uses CSPRNG for nonce, nonce injected in ciphertext.
+ * For `encrypt`, a `nonceBytes`-length buffer is fetched from CSPRNG and
+ * prepended to encrypted ciphertext. For `decrypt`, first `nonceBytes` of ciphertext
+ * are treated as nonce. The wrapper always allocates a fresh `nonce || ciphertext`
+ * buffer on encrypt and intentionally does not support caller-provided destination buffers.
+ * Too-short decrypt inputs are split into short/empty nonce views and then delegated
+ * to the wrapped cipher instead of being rejected here first.
+ *
+ * NOTE: Under the same key, using random nonces (e.g. `managedNonce`) with AES-GCM and ChaCha
+ * should be limited to `2**23` (8M) messages to get a collision chance of
+ * `2**-50`. Stretching to `2**32` (4B) messages would raise that chance to
+ * `2**-33`, still negligible but creeping up.
+ * @param fn - Cipher constructor that expects a nonce.
+ * @param randomBytes_ - Random-byte source used for nonce generation.
+ * @returns Cipher constructor that prepends the nonce to ciphertext.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On invalid nonce lengths observed at wrapper construction or use. {@link RangeError}
+ * @example
+ * Prepends a fresh random nonce to every ciphertext.
+ *
+ * ```ts
+ * import { gcm } from '@noble/ciphers/aes.js';
+ * import { managedNonce, randomBytes } from '@noble/ciphers/utils.js';
+ * const wrapped = managedNonce(gcm);
+ * const key = randomBytes(16);
+ * const ciphertext = wrapped(key).encrypt(new Uint8Array([1, 2, 3]));
+ * wrapped(key).decrypt(ciphertext);
+ * ```
+ */
+export function managedNonce(fn, randomBytes_ = randomBytes) {
+    if (typeof fn !== 'function')
+        throw new TypeError('"fn" expected cipher constructor, got type=' + typeof fn);
+    if (typeof randomBytes_ !== 'function')
+        throw new TypeError('"randomBytes_" expected function, got type=' + typeof randomBytes_);
+    const { nonceLength } = fn;
+    anumber(nonceLength, 'fn.nonceLength');
+    const addNonce = (nonce, ciphertext, plaintext) => {
+        const out = concatBytes(nonce, ciphertext);
+        // Wrapped ciphers may alias caller plaintext on encrypt(); never zero
+        // caller-owned buffers here.
+        if (!overlapBytes(plaintext, ciphertext))
+            ciphertext.fill(0);
+        return out;
+    };
+    // NOTE: we cannot support DST here, it would be mistake:
+    // - we don't know how much dst length cipher requires
+    // - nonce may unalign dst and break everything
+    // - we create new u8a anyway (concatBytes)
+    // - previously we passed all args to cipher, but that was mistake!
+    const res = ((key, ...args) => ({
+        encrypt(plaintext) {
+            abytes(plaintext, undefined, 'data');
+            const nonce = randomBytes_(nonceLength);
+            const encrypted = fn(key, nonce, ...args).encrypt(plaintext);
+            // @ts-ignore
+            if (encrypted instanceof Promise)
+                return encrypted.then((ct) => addNonce(nonce, ct, plaintext));
+            return addNonce(nonce, encrypted, plaintext);
+        },
+        decrypt(ciphertext) {
+            abytes(ciphertext, undefined, 'data');
+            const nonce = ciphertext.subarray(0, nonceLength);
+            const decrypted = ciphertext.subarray(nonceLength);
+            return fn(key, nonce, ...args).decrypt(decrypted);
+        },
+    }));
+    // Auto-nonce wrappers still preserve the wrapped payload geometry.
+    if ('blockSize' in fn)
+        res.blockSize = fn.blockSize;
+    if ('tagLength' in fn)
+        res.tagLength = fn.tagLength;
+    if ('withAAD' in fn)
+        res.withAAD = fn.withAAD;
+    return res;
+}

backend/static/vendor/noble-hashes/_blake.js

@@ -0,0 +1,59 @@
+/**
+ * Internal helpers for blake hash.
+ * @module
+ */
+import { rotr } from "./utils.js";
+// Unrealized speed-up: a file-local copy of rotr measured ~1-2% faster blake2s/blake256/blake3
+// on Node 24 (V8 does not inline it into G1s/G2s across the module boundary). Reused from
+// utils for deduplication.
+/**
+ * Internal blake permutation table.
+ * Rows `0..9` serve BLAKE2s, rows `0..11` serve BLAKE2b with `10..11 = 0..1`, and Blake1 also
+ * reuses the later rows shown below. Blake1 expands rounds `10..15` as `SIGMA[i % 10]`, so rows
+ * `10..15` intentionally repeat rows `0..5` for the 14-round (256) and 16-round (512) variants.
+ */
+// prettier-ignore
+export const BSIGMA = /* @__PURE__ */ Uint8Array.from([
+    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+    14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3,
+    11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4,
+    7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8,
+    9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13,
+    2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9,
+    12, 5, 1, 15, 14, 13, 4, 10, 0, 7, 6, 3, 9, 2, 8, 11,
+    13, 11, 7, 14, 12, 1, 3, 9, 5, 0, 15, 4, 8, 6, 2, 10,
+    6, 15, 14, 9, 11, 3, 0, 8, 12, 2, 13, 7, 1, 4, 10, 5,
+    10, 2, 8, 4, 7, 6, 1, 5, 15, 11, 9, 14, 3, 12, 13, 0,
+    0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15,
+    14, 10, 4, 8, 9, 15, 13, 6, 1, 12, 0, 2, 11, 7, 5, 3,
+    // Blake1, unused in others
+    11, 8, 12, 0, 5, 2, 15, 13, 10, 14, 3, 6, 7, 1, 9, 4,
+    7, 9, 3, 1, 13, 12, 11, 14, 2, 6, 5, 10, 4, 0, 15, 8,
+    9, 0, 5, 7, 2, 4, 10, 15, 14, 1, 11, 12, 6, 8, 3, 13,
+    2, 12, 6, 10, 0, 11, 8, 3, 4, 13, 7, 5, 15, 14, 1, 9,
+]);
+// Unrealized speed-up: G1s/G2s return a fresh {a,b,c,d} per call (160×/blake2s block), which
+// V8's escape analysis does not eliminate when G stays a real call. Writing into one reused
+// module-level slot instead measured +4-6% blake2s/blake256 1MB throughput on Node 24 — but
+// ~2% slower blake3 (whose profile inlines G and scalar-replaces the fresh object) and no
+// difference on 32-byte inputs. Kept simple.
+// 32-bit / BLAKE2s first half of G, with the fixed `(16, 12)` rotation pair.
+// Parameter `x` is the RFC 7693 first-half message word, or Blake1's pre-mixed
+// `m[sigma[r][2i]] ^ u[sigma[r][2i+1]]` addend in the 32-bit path.
+export function G1s(a, b, c, d, x) {
+    a = (a + b + x) | 0;
+    d = rotr(d ^ a, 16);
+    c = (c + d) | 0;
+    b = rotr(b ^ c, 12);
+    return { a, b, c, d };
+}
+// 32-bit / BLAKE2s second half of G.
+// Parameter `x` is the RFC 7693 second-half (`y`) message word, or Blake1's pre-mixed
+// `m[sigma[r][2i + 1]] ^ u[sigma[r][2i]]` addend in the 32-bit path.
+export function G2s(a, b, c, d, x) {
+    a = (a + b + x) | 0;
+    d = rotr(d ^ a, 8);
+    c = (c + d) | 0;
+    b = rotr(b ^ c, 7);
+    return { a, b, c, d };
+}

backend/static/vendor/noble-hashes/_md.js

@@ -0,0 +1,209 @@
+/**
+ * Internal Merkle-Damgard hash utils.
+ * @module
+ */
+import { setU64FromNum } from "./_u64.js";
+import { abytes, aexists, aoutput, createView } from "./utils.js";
+/**
+ * Shared 32-bit conditional boolean primitive reused by SHA-256, SHA-1, and MD5 `F`.
+ * Returns bits from `b` when `a` is set, otherwise from `c`.
+ * The XOR form is equivalent to MD5's `F(X,Y,Z) = XY v not(X)Z` because the masked terms never
+ * set the same bit.
+ * @param a - selector word
+ * @param b - word chosen when selector bit is set
+ * @param c - word chosen when selector bit is clear
+ * @returns Mixed 32-bit word.
+ * @example
+ * Combine three words with the shared 32-bit choice primitive.
+ * ```ts
+ * Chi(0xffffffff, 0x12345678, 0x87654321);
+ * ```
+ */
+export function Chi(a, b, c) {
+    return (a & b) ^ (~a & c);
+}
+/**
+ * Shared 32-bit majority primitive reused by SHA-256 and SHA-1.
+ * Returns bits shared by at least two inputs.
+ * @param a - first input word
+ * @param b - second input word
+ * @param c - third input word
+ * @returns Mixed 32-bit word.
+ * @example
+ * Combine three words with the shared 32-bit majority primitive.
+ * ```ts
+ * Maj(0xffffffff, 0x12345678, 0x87654321);
+ * ```
+ */
+export function Maj(a, b, c) {
+    return (a & b) ^ (a & c) ^ (b & c);
+}
+/**
+ * Merkle-Damgard hash construction base class.
+ * Could be used to create MD5, RIPEMD, SHA1, SHA2.
+ * Accepts only byte-aligned `Uint8Array` input, even when the underlying spec describes bit
+ * strings with partial-byte tails.
+ * @param blockLen - internal block size in bytes
+ * @param outputLen - digest size in bytes
+ * @param padOffset - trailing length field size in bytes
+ * @param isLE - whether length and state words are encoded in little-endian
+ * @example
+ * Use a concrete subclass to get the shared Merkle-Damgard update/digest flow.
+ * ```ts
+ * import { _SHA1 } from '@noble/hashes/legacy.js';
+ * const hash = new _SHA1();
+ * hash.update(new Uint8Array([97, 98, 99]));
+ * hash.digest();
+ * ```
+ */
+export class HashMD {
+    blockLen;
+    outputLen;
+    canXOF = false;
+    padOffset;
+    isLE;
+    // For partial updates less than block size
+    buffer;
+    view;
+    finished = false;
+    length = 0;
+    pos = 0;
+    destroyed = false;
+    constructor(blockLen, outputLen, padOffset, isLE) {
+        this.blockLen = blockLen;
+        this.outputLen = outputLen;
+        this.padOffset = padOffset;
+        this.isLE = isLE;
+        this.buffer = new Uint8Array(blockLen);
+        this.view = createView(this.buffer);
+    }
+    update(data) {
+        aexists(this);
+        abytes(data);
+        const { view, buffer, blockLen } = this;
+        const len = data.length;
+        let processed = false;
+        for (let pos = 0; pos < len;) {
+            const take = Math.min(blockLen - this.pos, len - pos);
+            // Fast path only when there is no buffered partial block: `take === blockLen` implies
+            // `this.pos === 0`, so we can process full blocks directly from the input view.
+            if (take === blockLen) {
+                const dataView = createView(data);
+                for (; blockLen <= len - pos; pos += blockLen)
+                    this.process(dataView, pos);
+                processed = true;
+                continue;
+            }
+            // When the whole input is buffered in one go (common for short messages), passing `data`
+            // directly avoids allocating a subarray view.
+            buffer.set(pos === 0 && take === len ? data : data.subarray(pos, pos + take), this.pos);
+            this.pos += take;
+            pos += take;
+            if (this.pos === blockLen) {
+                this.process(view, 0);
+                this.pos = 0;
+                processed = true;
+            }
+        }
+        this.length += data.length;
+        // Shared schedule buffers only pick up input-derived words inside process(); if everything
+        // was buffered without processing, there is nothing to zero.
+        if (processed)
+            this.roundClean();
+        return this;
+    }
+    digestInto(out) {
+        aexists(this);
+        aoutput(out, this);
+        this.finished = true;
+        // Padding
+        // We can avoid allocation of buffer for padding completely if it
+        // was previously not allocated here. But it won't change performance.
+        const { buffer, view, blockLen, isLE } = this;
+        let { pos } = this;
+        // append the bit '1' to the message, then zero-pad the rest of the block
+        buffer[pos++] = 0b10000000;
+        buffer.fill(0, pos);
+        // we have less than padOffset left in buffer, so we cannot put length in
+        // current block, need process it and pad again
+        if (this.padOffset > blockLen - pos) {
+            this.process(view, 0);
+            buffer.fill(0);
+        }
+        // `padOffset` reserves the whole length field. For SHA-384/512 the high 64 bits stay zero from
+        // the padding fill above, and JS will overflow before user input can make that half non-zero.
+        // So we only need to write the low 64 bits here (`length * 8` only scales the exponent of an
+        // integer below 2**53, so the split inside the helper stays exact).
+        setU64FromNum(view, blockLen - 8, this.length * 8, isLE);
+        this.process(view, 0);
+        // The final block above is processed outside update(), so the shared message-schedule
+        // buffers (e.g. SHA256_W) would otherwise retain input-derived words after digest().
+        this.roundClean();
+        // digest() passes our own `buffer` as `out`; reuse its cached view instead of allocating one.
+        const oview = out === buffer ? view : createView(out);
+        const len = this.outputLen;
+        // NOTE: we do division by 4 later, which must be fused in single op with modulo by JIT
+        const outLen = len / 4;
+        const state = this.get();
+        // Subclass-misconfiguration invariant: outputLen must be 32-bit aligned and fit the state.
+        if (len % 4 || outLen > state.length)
+            throw new Error('invalid outputLen');
+        for (let i = 0; i < outLen; i++)
+            oview.setUint32(4 * i, state[i], isLE);
+    }
+    digest() {
+        const { buffer, outputLen } = this;
+        this.digestInto(buffer);
+        // Copy before destroy(): subclasses wipe `buffer` during cleanup, but `digest()` must return
+        // fresh bytes to the caller.
+        const res = buffer.slice(0, outputLen);
+        this.destroy();
+        return res;
+    }
+    _cloneIntoMeta(to) {
+        const { buffer, length, finished, destroyed, pos } = this;
+        to.destroyed = destroyed;
+        to.finished = finished;
+        to.length = length;
+        to.pos = pos;
+        // Only partial-block bytes need copying: when `length % blockLen === 0`, `pos === 0` and
+        // later `update()` / `digestInto()` overwrite `to.buffer` from the start before reading it.
+        if (pos)
+            to.buffer.set(buffer); // Avoid a hot modulo guard.
+        return to;
+    }
+    clone() {
+        return this._cloneInto();
+    }
+}
+/**
+ * Initial SHA-2 state: fractional parts of square roots of first 16 primes 2..53.
+ * Check out `test/misc/sha2-gen-iv.js` for recomputation guide.
+ */
+/** Initial SHA256 state from RFC 6234 §6.1: the first 32 bits of the fractional parts of the
+ * square roots of the first eight prime numbers. Exported as a shared table; callers must treat
+ * it as read-only because constructors copy words from it by index. */
+export const SHA256_IV = /* @__PURE__ */ Uint32Array.from([
+    0x6a09e667, 0xbb67ae85, 0x3c6ef372, 0xa54ff53a, 0x510e527f, 0x9b05688c, 0x1f83d9ab, 0x5be0cd19,
+]);
+/** Initial SHA224 state `H(0)` from RFC 6234 §6.1. Exported as a shared table; callers must
+ * treat it as read-only because constructors copy words from it by index. */
+export const SHA224_IV = /* @__PURE__ */ Uint32Array.from([
+    0xc1059ed8, 0x367cd507, 0x3070dd17, 0xf70e5939, 0xffc00b31, 0x68581511, 0x64f98fa7, 0xbefa4fa4,
+]);
+/** Initial SHA384 state from RFC 6234 §6.3: eight RFC 64-bit `H(0)` words stored as sixteen
+ * big-endian 32-bit halves. Derived from the fractional parts of the square roots of the ninth
+ * through sixteenth prime numbers. Exported as a shared table; callers must treat it as read-only
+ * because constructors copy halves from it by index. */
+export const SHA384_IV = /* @__PURE__ */ Uint32Array.from([
+    0xcbbb9d5d, 0xc1059ed8, 0x629a292a, 0x367cd507, 0x9159015a, 0x3070dd17, 0x152fecd8, 0xf70e5939,
+    0x67332667, 0xffc00b31, 0x8eb44a87, 0x68581511, 0xdb0c2e0d, 0x64f98fa7, 0x47b5481d, 0xbefa4fa4,
+]);
+/** Initial SHA512 state from RFC 6234 §6.3: eight RFC 64-bit `H(0)` words stored as sixteen
+ * big-endian 32-bit halves. Derived from the fractional parts of the square roots of the first
+ * eight prime numbers. Exported as a shared table; callers must treat it as read-only because
+ * constructors copy halves from it by index. */
+export const SHA512_IV = /* @__PURE__ */ Uint32Array.from([
+    0x6a09e667, 0xf3bcc908, 0xbb67ae85, 0x84caa73b, 0x3c6ef372, 0xfe94f82b, 0xa54ff53a, 0x5f1d36f1,
+    0x510e527f, 0xade682d1, 0x9b05688c, 0x2b3e6c1f, 0x1f83d9ab, 0xfb41bd6b, 0x5be0cd19, 0x137e2179,
+]);

backend/static/vendor/noble-hashes/_u64.js

@@ -0,0 +1,77 @@
+const U32_MASK64 = /* @__PURE__ */ (() => BigInt(2 ** 32 - 1))();
+const _32n = /* @__PURE__ */ BigInt(32);
+// Split bigint into two 32-bit halves. With `le=true`, returned fields become `{ h: low, l: high
+// }` to match little-endian word order rather than the property names.
+function fromBig(n, le = false) {
+    if (le)
+        return { h: Number(n & U32_MASK64), l: Number((n >> _32n) & U32_MASK64) };
+    return { h: Number((n >> _32n) & U32_MASK64) | 0, l: Number(n & U32_MASK64) | 0 };
+}
+// Split bigint list into `[highWords, lowWords]` when `le=false`; with `le=true`, the first array
+// holds the low halves because `fromBig(...)` swaps the semantic meaning of `h` and `l`.
+function split(lst, le = false) {
+    const len = lst.length;
+    let Ah = new Uint32Array(len);
+    let Al = new Uint32Array(len);
+    for (let i = 0; i < len; i++) {
+        const { h, l } = fromBig(lst[i], le);
+        [Ah[i], Al[i]] = [h, l];
+    }
+    return [Ah, Al];
+}
+// Combine explicit `(high, low)` 32-bit halves into a bigint; `>>> 0` normalizes signed JS
+// bitwise results back to uint32 first, and little-endian callers must swap.
+const toBig = (h, l) => (BigInt(h >>> 0) << _32n) | BigInt(l >>> 0);
+// Split a JS number into u32 halves without a BigInt allocation. Exact only for integers
+// `0 <= n < 2**53`; callers use it on byte / bit counters, which JS length math caps far below
+// that (an ArrayBuffer cannot exceed 2**53 - 1 bytes).
+const fromNumH = (n) => (n / 2 ** 32) | 0;
+const fromNumL = (n) => n >>> 0;
+// Drop-in replacement for `view.setBigUint64(byteOffset, BigInt(n), isLE)` without the per-call
+// BigInt allocation. Same `n < 2**53` precondition as `fromNumH`/`fromNumL`.
+function setU64FromNum(view, byteOffset, n, isLE) {
+    const h = fromNumH(n);
+    const l = fromNumL(n);
+    view.setUint32(byteOffset, isLE ? l : h, isLE);
+    view.setUint32(byteOffset + 4, isLE ? h : l, isLE);
+}
+// High 32-bit half of a 64-bit logical right shift for `s` in `0..31`.
+const shrSH = (h, _l, s) => h >>> s;
+// Low 32-bit half of a 64-bit logical right shift, valid for `s` in `1..31`.
+const shrSL = (h, l, s) => (h << (32 - s)) | (l >>> s);
+// High 32-bit half of a 64-bit right rotate, valid for `s` in `1..31`.
+const rotrSH = (h, l, s) => (h >>> s) | (l << (32 - s));
+// Low 32-bit half of a 64-bit right rotate, valid for `s` in `1..31`.
+const rotrSL = (h, l, s) => (h << (32 - s)) | (l >>> s);
+// High 32-bit half of a 64-bit right rotate, valid for `s` in `33..63`; `32` uses `rotr32*`.
+const rotrBH = (h, l, s) => (h << (64 - s)) | (l >>> (s - 32));
+// Low 32-bit half of a 64-bit right rotate, valid for `s` in `33..63`; `32` uses `rotr32*`.
+const rotrBL = (h, l, s) => (h >>> (s - 32)) | (l << (64 - s));
+// High 32-bit half of a 64-bit right rotate for `s === 32`; this is just the swapped low half.
+const rotr32H = (_h, l) => l;
+// Low 32-bit half of a 64-bit right rotate for `s === 32`; this is just the swapped high half.
+const rotr32L = (h, _l) => h;
+// 64-bit left rotates (rotl*) are not defined here: sha3.ts, their only consumer, keeps
+// local copies so V8 inlines them into keccakP.
+// Add two split 64-bit words and return the split `{ h, l }` sum.
+// JS uses 32-bit signed integers for bitwise operations, so we cannot simply shift the carry out
+// of the low sum and instead use division.
+function add(Ah, Al, Bh, Bl) {
+    const l = (Al >>> 0) + (Bl >>> 0);
+    return { h: (Ah + Bh + ((l / 2 ** 32) | 0)) | 0, l: l | 0 };
+}
+// Addition with more than 2 elements
+// Unmasked low-word accumulator for 3-way addition; pass the raw result into `add3H(...)`.
+const add3L = (Al, Bl, Cl) => (Al >>> 0) + (Bl >>> 0) + (Cl >>> 0);
+// High-word finalize step for 3-way addition; `low` must be the untruncated output of `add3L(...)`.
+const add3H = (low, Ah, Bh, Ch) => (Ah + Bh + Ch + ((low / 2 ** 32) | 0)) | 0;
+// Unmasked low-word accumulator for 4-way addition; pass the raw result into `add4H(...)`.
+const add4L = (Al, Bl, Cl, Dl) => (Al >>> 0) + (Bl >>> 0) + (Cl >>> 0) + (Dl >>> 0);
+// High-word finalize step for 4-way addition; `low` must be the untruncated output of `add4L(...)`.
+const add4H = (low, Ah, Bh, Ch, Dh) => (Ah + Bh + Ch + Dh + ((low / 2 ** 32) | 0)) | 0;
+// Unmasked low-word accumulator for 5-way addition; pass the raw result into `add5H(...)`.
+const add5L = (Al, Bl, Cl, Dl, El) => (Al >>> 0) + (Bl >>> 0) + (Cl >>> 0) + (Dl >>> 0) + (El >>> 0);
+// High-word finalize step for 5-way addition; `low` must be the untruncated output of `add5L(...)`.
+const add5H = (low, Ah, Bh, Ch, Dh, Eh) => (Ah + Bh + Ch + Dh + Eh + ((low / 2 ** 32) | 0)) | 0;
+// prettier-ignore
+export { add, add3H, add3L, add4H, add4L, add5H, add5L, fromBig, fromNumH, fromNumL, rotr32H, rotr32L, rotrBH, rotrBL, rotrSH, rotrSL, setU64FromNum, shrSH, shrSL, split, toBig };

backend/static/vendor/noble-hashes/argon2.js

@@ -0,0 +1,577 @@
+/**
+ * Argon2 KDF from RFC 9106. Can be used to create a key from password and salt.
+ * We suggest to use Scrypt. JS Argon is 2-10x slower than native code because of 64-bitness:
+ * * argon uses uint64, but JS doesn't have fast uint64array
+ * * uint64 multiplication is 1/3 of time
+ * * `P` function would be very nice with u64, because most of value will be in registers,
+ *   hovewer with u32 it will require 32 registers, which is too much.
+ * * JS arrays do slow bound checks, so reading from `A2_BUF` slows it down
+ * @module
+ */
+import { blake2b } from "./blake2.js";
+import { anumber, checkOpts, clean, kdfInputToBytes, nextTick, swap32IfBE, swap8IfBE, u32, u8, } from "./utils.js";
+// RFC 9106 §3.1 type `y`: 0 = Argon2d, 1 = Argon2i, 2 = Argon2id. The numeric values are the
+// spec-bound part here; the object keys are internal labels.
+const AT = { Argon2d: 0, Argon2i: 1, Argon2id: 2 };
+// RFC 9106 sync points constant `SL = 4`, fixed by the design rather than exposed as a tuning knob.
+const ARGON2_SYNC_POINTS = 4;
+// Preserve Argon2's `LE32(len(X)) || X` encoding for omitted
+// optional fields by emitting empty bytes.
+const abytesOrZero = (buf, errorTitle = '') => {
+    if (buf === undefined)
+        return Uint8Array.of();
+    return kdfInputToBytes(buf, errorTitle);
+};
+// Temporary block buffer.
+// 1024-byte block: 256 u32 = 128 interleaved low/high halves = RFC's
+// 8x8 matrix of 16-byte registers.
+const A2_BUF = new Uint32Array(256);
+// Quarter-round over 64-bit word indices into `A2_BUF`; each index maps to adjacent low/high u32s.
+// Each BlaMka step `X = X + Y + 2 * trunc(X) * trunc(Y)` (trunc = low 32 bits) starts
+// with an exact low product from `Math.imul`. The rounded double product is within 1024 of the
+// exact u64 product; subtracting that exact low half and rounding to the nearest multiple of
+// 2^32 therefore recovers the exact high half. RFC 9106 Figure 19 GB then rotates by 32, 24,
+// 16, and 63 bits after each XOR.
+function G(a, b, c, d) {
+    let Al = A2_BUF[2 * a], Ah = A2_BUF[2 * a + 1]; // prettier-ignore
+    let Bl = A2_BUF[2 * b], Bh = A2_BUF[2 * b + 1]; // prettier-ignore
+    let Cl = A2_BUF[2 * c], Ch = A2_BUF[2 * c + 1]; // prettier-ignore
+    let Dl = A2_BUF[2 * d], Dh = A2_BUF[2 * d + 1]; // prettier-ignore
+    let ml = 0, mh = 0, rl = 0, xh = 0, xl = 0; // prettier-ignore
+    // A = blamka(A, B); D = rotr64(D ^ A, 32)
+    ml = Math.imul(Al, Bl);
+    mh = (((Al >>> 0) * (Bl >>> 0) - (ml >>> 0)) / 0x100000000 + 0.5) | 0; // prettier-ignore
+    rl = (Al >>> 0) + (Bl >>> 0) + ((ml << 1) >>> 0);
+    Ah = (Ah + Bh + ((mh << 1) | (ml >>> 31)) + ((rl / 0x100000000) | 0)) | 0;
+    Al = rl | 0; // prettier-ignore
+    xh = Dh ^ Ah;
+    xl = Dl ^ Al; // prettier-ignore
+    Dh = xl;
+    Dl = xh; // prettier-ignore
+    // C = blamka(C, D); B = rotr64(B ^ C, 24)
+    ml = Math.imul(Cl, Dl);
+    mh = (((Cl >>> 0) * (Dl >>> 0) - (ml >>> 0)) / 0x100000000 + 0.5) | 0; // prettier-ignore
+    rl = (Cl >>> 0) + (Dl >>> 0) + ((ml << 1) >>> 0);
+    Ch = (Ch + Dh + ((mh << 1) | (ml >>> 31)) + ((rl / 0x100000000) | 0)) | 0;
+    Cl = rl | 0; // prettier-ignore
+    xh = Bh ^ Ch;
+    xl = Bl ^ Cl; // prettier-ignore
+    Bh = (xh >>> 24) | (xl << 8);
+    Bl = (xh << 8) | (xl >>> 24); // prettier-ignore
+    // A = blamka(A, B); D = rotr64(D ^ A, 16)
+    ml = Math.imul(Al, Bl);
+    mh = (((Al >>> 0) * (Bl >>> 0) - (ml >>> 0)) / 0x100000000 + 0.5) | 0; // prettier-ignore
+    rl = (Al >>> 0) + (Bl >>> 0) + ((ml << 1) >>> 0);
+    Ah = (Ah + Bh + ((mh << 1) | (ml >>> 31)) + ((rl / 0x100000000) | 0)) | 0;
+    Al = rl | 0; // prettier-ignore
+    xh = Dh ^ Ah;
+    xl = Dl ^ Al; // prettier-ignore
+    Dh = (xh >>> 16) | (xl << 16);
+    Dl = (xh << 16) | (xl >>> 16); // prettier-ignore
+    // C = blamka(C, D); B = rotr64(B ^ C, 63)
+    ml = Math.imul(Cl, Dl);
+    mh = (((Cl >>> 0) * (Dl >>> 0) - (ml >>> 0)) / 0x100000000 + 0.5) | 0; // prettier-ignore
+    rl = (Cl >>> 0) + (Dl >>> 0) + ((ml << 1) >>> 0);
+    Ch = (Ch + Dh + ((mh << 1) | (ml >>> 31)) + ((rl / 0x100000000) | 0)) | 0;
+    Cl = rl | 0; // prettier-ignore
+    xh = Bh ^ Ch;
+    xl = Bl ^ Cl; // prettier-ignore
+    Bh = (xh << 1) | (xl >>> 31);
+    Bl = (xh >>> 31) | (xl << 1); // prettier-ignore
+    ((A2_BUF[2 * a] = Al), (A2_BUF[2 * a + 1] = Ah));
+    ((A2_BUF[2 * b] = Bl), (A2_BUF[2 * b + 1] = Bh));
+    ((A2_BUF[2 * c] = Cl), (A2_BUF[2 * c + 1] = Ch));
+    ((A2_BUF[2 * d] = Dl), (A2_BUF[2 * d + 1] = Dh));
+}
+// Argon2 permutation over 16 register indices into `A2_BUF`, not the register values themselves.
+// RFC 9106 Figure 17: these arguments are the 16 `v0..v15` 64-bit word
+// indices inside eight 16-byte inputs, not copied word values.
+// prettier-ignore
+function P(v00, v01, v02, v03, v04, v05, v06, v07, v08, v09, v10, v11, v12, v13, v14, v15) {
+    // RFC 9106 Figure 18: first apply GB across rows, then across columns of the 8x8 register matrix.
+    G(v00, v04, v08, v12);
+    G(v01, v05, v09, v13);
+    G(v02, v06, v10, v14);
+    G(v03, v07, v11, v15);
+    G(v00, v05, v10, v15);
+    G(v01, v06, v11, v12);
+    G(v02, v07, v08, v13);
+    G(v03, v04, v09, v14);
+}
+function block(x, xPos, yPos, outPos, needXor) {
+    // Stage R = X xor Y in the destination before permuting it. This avoids rereading both source
+    // blocks when folding R into the permuted scratch block below.
+    if (needXor) {
+        for (let i = 0; i < 256; i++) {
+            const r = x[xPos + i] ^ x[yPos + i];
+            A2_BUF[i] = r;
+            x[outPos + i] ^= r;
+        }
+    }
+    else {
+        for (let i = 0; i < 256; i++) {
+            const r = x[xPos + i] ^ x[yPos + i];
+            A2_BUF[i] = r;
+            x[outPos + i] = r;
+        }
+    }
+    // rows (8 consecutive 16-register groups)
+    for (let i = 0; i < 128; i += 16) {
+        // prettier-ignore
+        P(i, i + 1, i + 2, i + 3, i + 4, i + 5, i + 6, i + 7, i + 8, i + 9, i + 10, i + 11, i + 12, i + 13, i + 14, i + 15);
+    }
+    // columns (8 strided 16-register groups)
+    for (let i = 0; i < 16; i += 2) {
+        // prettier-ignore
+        P(i, i + 1, i + 16, i + 17, i + 32, i + 33, i + 48, i + 49, i + 64, i + 65, i + 80, i + 81, i + 96, i + 97, i + 112, i + 113);
+    }
+    // RFC 9106 step 6: passes after the first XOR the old destination block into the new G(X, Y).
+    for (let i = 0; i < 256; i++)
+        x[outPos + i] ^= A2_BUF[i];
+    clean(A2_BUF);
+}
+// Variable-Length Hash Function H'
+// Returns bytes, not words; 1024-byte block callers explicitly reinterpret with `u32(...)`.
+function Hp(A, dkLen) {
+    const A8 = u8(A);
+    const T = new Uint32Array(1);
+    const T8 = u8(T);
+    // Argon2 H' prefixes dkLen as LE32; native Uint32Array writes would serialize as BE on s390x.
+    T[0] = swap8IfBE(dkLen);
+    // Fast path
+    if (dkLen <= 64)
+        return blake2b.create({ dkLen }).update(T8).update(A8).digest();
+    const out = new Uint8Array(dkLen);
+    let V = blake2b.create({}).update(T8).update(A8).digest();
+    let pos = 0;
+    // RFC 9106 Figure 8: each intermediate `V_i` contributes only `W_i`, its first 32 bytes; only
+    // `V_{r+1}` is emitted in full at the remaining length.
+    out.set(V.subarray(0, 32));
+    pos += 32;
+    // Rest blocks
+    for (; dkLen - pos > 64; pos += 32) {
+        const Vh = blake2b.create({}).update(V);
+        Vh.digestInto(V);
+        Vh.destroy();
+        out.set(V.subarray(0, 32), pos);
+    }
+    // Last block
+    out.set(blake2b(V, { dkLen: dkLen - pos }), pos);
+    clean(V, T);
+    // H' is byte-oriented; returning `u32(out)` would silently drop dkLen % 4 tail bytes.
+    return out;
+}
+// Used only inside argon2Blocks!
+function indexAlpha(r, s, laneLen, segmentLen, index, randL, sameLane = false) {
+    // RFC 9106 §3.4.2 Figures 12-13: map `J1` / `J2` into the current lane's reference area `W`.
+    let area;
+    if (r === 0) {
+        if (s === 0)
+            area = index - 1;
+        else if (sameLane)
+            area = s * segmentLen + index - 1;
+        else
+            area = s * segmentLen + (index == 0 ? -1 : 0);
+    }
+    else if (sameLane)
+        area = laneLen - segmentLen + index - 1;
+    else
+        area = laneLen - segmentLen + (index == 0 ? -1 : 0);
+    const startPos = r !== 0 && s !== ARGON2_SYNC_POINTS - 1 ? (s + 1) * segmentLen : 0;
+    // Use the same exact-high recovery as G. `areaHigh` is floor(|W| * J1^2/2^32 / 2^32).
+    const randLow = Math.imul(randL, randL);
+    const randHigh = (((randL >>> 0) * (randL >>> 0) - (randLow >>> 0)) / 0x100000000 + 0.5) | 0;
+    const areaLow = Math.imul(area, randHigh);
+    const areaHigh = (((area >>> 0) * (randHigh >>> 0) - (areaLow >>> 0)) / 0x100000000 + 0.5) | 0;
+    const rel = area - 1 - areaHigh;
+    return (startPos + rel) % laneLen;
+}
+// Exclusive `2^32` sentinel used by `isU32(...)`, not the inclusive maximum u32 value.
+const maxUint32 = Math.pow(2, 32);
+const ARGON2_DEFAULT_MEMORY = 1024 ** 2; // KiB: 1 GiB
+const ARGON2_DEFAULT_MAXMEM = ARGON2_DEFAULT_MEMORY * 1024;
+// Validate safe JS integers in `[0, 2^32 - 1]`.
+function isU32(num) {
+    return Number.isSafeInteger(num) && num >= 0 && num < maxUint32;
+}
+function argon2Opts(opts = {}) {
+    opts = checkOpts({}, opts);
+    const merged = {
+        t: 3,
+        m: ARGON2_DEFAULT_MEMORY,
+        p: 1,
+        version: 0x13,
+        dkLen: 32,
+        maxmem: ARGON2_DEFAULT_MAXMEM,
+        asyncTick: 10,
+    };
+    // Unknown keys are copied through unchanged here and later ignored unless
+    // destructuring consumes them.
+    for (let [k, v] of Object.entries(opts))
+        if (v !== undefined)
+            merged[k] = v;
+    const { dkLen, p, m, t, version, onProgress, asyncTick } = merged;
+    // RFC 9106 §3.1: tag length `T` MUST be an integer number of bytes from 4 to 2^32-1.
+    if (!isU32(dkLen) || dkLen < 4)
+        throw new Error('"dkLen" must be 4..');
+    if (!isU32(p) || p < 1 || p >= Math.pow(2, 24))
+        throw new Error('"p" must be 1..2^24');
+    if (!isU32(m))
+        throw new Error('"m" must be 0..2^32');
+    if (!isU32(t) || t < 1)
+        throw new Error('"t" (iterations) must be 1..2^32');
+    if (onProgress !== undefined && typeof onProgress !== 'function')
+        throw new Error('"onProgress" must be a function');
+    anumber(asyncTick, 'asyncTick');
+    /*
+    Memory size m MUST be an integer number of kibibytes from 8*p
+    to 2^(32)-1. The actual number of blocks is m', which is m
+    rounded down to the nearest multiple of 4*p.
+    */
+    if (!isU32(m) || m < 8 * p)
+        throw new Error('"m" (memory) must be at least 8*p bytes');
+    // Accept legacy `0x10` for compatibility even though RFC 9106 profiles standardize `0x13`.
+    if (version !== 0x10 && version !== 0x13)
+        throw new Error('"version" must be 0x10 or 0x13, got ' + version);
+    return merged;
+}
+function argon2InitialHash(password, salt, type, opts) {
+    const ownedInputs = [];
+    const BUF = new Uint32Array(1);
+    const BUF8 = u8(BUF);
+    let h;
+    let H0;
+    let succeeded = false;
+    const rememberOwned = (input, bytes) => {
+        if (typeof input === 'string')
+            ownedInputs.push(bytes);
+        return bytes;
+    };
+    try {
+        const passwordBytes = rememberOwned(password, kdfInputToBytes(password, 'password'));
+        const saltBytes = rememberOwned(salt, kdfInputToBytes(salt, 'salt'));
+        if (!isU32(passwordBytes.length))
+            throw new Error('"password" must be less of length 1..4Gb');
+        // RFC 9106 §3.1 only requires S <= 2^32-1 bytes and says 16 bytes is RECOMMENDED for password
+        // hashing; this library intentionally takes the stricter common >=8-byte salt path.
+        if (!isU32(saltBytes.length) || saltBytes.length < 8)
+            throw new Error('"salt" must be of length 8..4Gb');
+        if (!Object.values(AT).includes(type))
+            throw new Error('"type" was invalid');
+        let { p, dkLen, m, t, version, key, personalization, maxmem, onProgress, asyncTick } = argon2Opts(opts);
+        // Validation
+        const keyInput = key;
+        key = rememberOwned(keyInput, abytesOrZero(keyInput, 'key'));
+        const personalizationInput = personalization;
+        personalization = rememberOwned(personalizationInput, abytesOrZero(personalizationInput, 'personalization'));
+        // H_0 = H^(64)(LE32(p) || LE32(T) || LE32(m) || LE32(t) ||
+        //       LE32(v) || LE32(y) || LE32(length(P)) || P ||
+        //       LE32(length(S)) || S ||  LE32(length(K)) || K ||
+        //       LE32(length(X)) || X)
+        h = blake2b.create();
+        for (let item of [p, dkLen, m, t, version, type]) {
+            // RFC 9106 H0 encodes these scalars as LE32, so normalize the host word before
+            // exposing bytes.
+            BUF[0] = swap8IfBE(item);
+            h.update(BUF8);
+        }
+        for (let i of [passwordBytes, saltBytes, key, personalization]) {
+            BUF[0] = swap8IfBE(i.length); // BUF is u32 array, this is valid once normalized to LE bytes
+            h.update(BUF8).update(i);
+        }
+        // Reserve two extra LE32 words after the 64-byte `H_0` so Figures 3-4 can append
+        // `LE32(0 or 1) || LE32(i)` in place for the lane-starting blocks.
+        H0 = new Uint32Array(18);
+        h.digestInto(u8(H0));
+        succeeded = true;
+        return { H0, p, dkLen, m, t, version, maxmem, onProgress, asyncTick };
+    }
+    finally {
+        // digestInto() does not destroy BLAKE2 state. It and all string-derived inputs contain secrets.
+        if (h)
+            h.destroy();
+        clean(BUF, ...ownedInputs);
+        if (!succeeded && H0)
+            clean(H0);
+    }
+}
+function argon2Init(password, salt, type, opts) {
+    const { H0, p, dkLen, m, t, version, maxmem, onProgress, asyncTick } = argon2InitialHash(password, salt, type, opts);
+    // 256 u32 = 1024 (BLOCK_SIZE), fills A2_BUF on processing
+    try {
+        // Params
+        const lanes = p;
+        // m' = 4 * p * floor (m / 4p)
+        const mP = 4 * p * Math.floor(m / (ARGON2_SYNC_POINTS * p));
+        //q = m' / p columns
+        const laneLen = Math.floor(mP / p);
+        const segmentLen = Math.floor(laneLen / ARGON2_SYNC_POINTS);
+        // `maxmem` is documented in bytes; compare against the actual 1024-byte block allocation.
+        const memUsed = mP * 1024;
+        if (!isU32(maxmem))
+            throw new Error('"maxmem" expected <2**32, got ' + maxmem);
+        if (memUsed > maxmem)
+            throw new Error('"maxmem" limit was hit: memUsed(mP*1024)=' + memUsed + ', maxmem=' + maxmem);
+        const B = new Uint32Array(memUsed / 4);
+        // Fill first blocks
+        for (let l = 0; l < p; l++) {
+            const i = 256 * laneLen * l;
+            // B[i][0] = H'^(1024)(H_0 || LE32(0) || LE32(i))
+            H0[17] = swap8IfBE(l);
+            H0[16] = swap8IfBE(0);
+            B.set(swap32IfBE(u32(Hp(H0, 1024))), i);
+            // B[i][1] = H'^(1024)(H_0 || LE32(1) || LE32(i))
+            H0[16] = swap8IfBE(1);
+            B.set(swap32IfBE(u32(Hp(H0, 1024))), i + 256);
+        }
+        let perBlock = () => { };
+        if (onProgress) {
+            // The first segment of the first pass skips two preinitialized blocks per lane.
+            const totalBlock = t * ARGON2_SYNC_POINTS * p * segmentLen - 2 * p;
+            // Invoke callback if progress changes from 10.01 to 10.02
+            // Allows to draw smooth progress bar on up to 8K screen
+            const callbackPer = Math.max(Math.floor(totalBlock / 10000), 1);
+            let blockCnt = 0;
+            perBlock = () => {
+                blockCnt++;
+                if (onProgress && (!(blockCnt % callbackPer) || blockCnt === totalBlock))
+                    onProgress(blockCnt / totalBlock);
+            };
+        }
+        return { type, mP, p, t, version, B, laneLen, lanes, segmentLen, dkLen, perBlock, asyncTick };
+    }
+    finally {
+        clean(H0);
+    }
+}
+function argon2Output(B, p, laneLen, dkLen) {
+    const B_final = new Uint32Array(256);
+    for (let l = 0; l < p; l++)
+        for (let j = 0; j < 256; j++)
+            B_final[j] ^= B[256 * (laneLen * l + laneLen - 1) + j];
+    // RFC 9106 steps 7-8 feed the byte string `C` into `H'^T(C)`, so normalize the xor'ed words
+    // back to spec byte order before `Hp(...)` reinterprets them as bytes.
+    const res = Hp(swap32IfBE(B_final), dkLen);
+    // Wipe both the xor scratch and the full working matrix once final digest bytes exist.
+    // JS cleanup is still only best-effort, but this local buffer is no longer needed here.
+    clean(B, B_final);
+    return res;
+}
+/**
+ * Fills every Argon2 block for all passes / slices / lanes, yielding once per
+ * processed block so callers control pacing: the sync driver just drains the
+ * generator, while the async driver awaits `nextTick()` between time slices.
+ */
+function* argon2Blocks(ctx, address) {
+    const { type, mP, p, t, version, B, laneLen, lanes, segmentLen, perBlock } = ctx;
+    // [address, input, zero_block] format so we can pass single U32 to block function
+    address[256 + 6] = mP;
+    address[256 + 8] = t;
+    address[256 + 10] = type;
+    for (let r = 0; r < t; r++) {
+        // RFC 9106 step 6 applies the XOR-on-later-passes rule only for version `0x13`; legacy
+        // `0x10` keeps the older overwrite behavior used by the v16 test vectors.
+        const needXor = r !== 0 && version === 0x13;
+        address[256 + 0] = r;
+        for (let s = 0; s < ARGON2_SYNC_POINTS; s++) {
+            address[256 + 4] = s;
+            // RFC 9106 §3.4.1.3: Argon2id uses Argon2i's data-independent `J1` / `J2` generation only
+            // in pass 0, slices 0 and 1; Argon2i uses it in every segment.
+            const dataIndependent = type == AT.Argon2i || (type == AT.Argon2id && r === 0 && s < 2);
+            for (let l = 0; l < p; l++) {
+                address[256 + 2] = l;
+                address[256 + 12] = 0;
+                let startPos = 0;
+                if (r === 0 && s === 0) {
+                    startPos = 2;
+                    if (dataIndependent) {
+                        address[256 + 12]++;
+                        block(address, 256, 2 * 256, 0, false);
+                        block(address, 0, 2 * 256, 0, false);
+                    }
+                }
+                // current block position
+                let offset = l * laneLen + s * segmentLen + startPos;
+                for (let index = startPos; index < segmentLen; index++, offset++) {
+                    perBlock();
+                    // Previous block position: wraps to the lane's last block only at lane start,
+                    // which can happen here only for the first block of slice 0 on passes > 0.
+                    const prev = offset % laneLen ? offset - 1 : offset + laneLen - 1;
+                    let randL, randH;
+                    if (dataIndependent) {
+                        let i128 = index % 128;
+                        // RFC 9106 §3.4.1.2: each 1024-byte address block yields 128 `(J1, J2)` pairs, so
+                        // regenerate it whenever the segment index crosses a multiple of 128.
+                        if (i128 === 0) {
+                            address[256 + 12]++;
+                            block(address, 256, 2 * 256, 0, false);
+                            block(address, 0, 2 * 256, 0, false);
+                        }
+                        randL = address[2 * i128];
+                        randH = address[2 * i128 + 1];
+                    }
+                    else {
+                        const T = 256 * prev;
+                        randL = B[T];
+                        randH = B[T + 1];
+                    }
+                    // Address-block path selects `J1` / `J2`, then maps them to the reference
+                    // lane/block per RFC 9106 §3.4.
+                    const refLane = r === 0 && s === 0 ? l : randH % lanes;
+                    const refPos = indexAlpha(r, s, laneLen, segmentLen, index, randL, refLane == l);
+                    const refBlock = laneLen * refLane + refPos;
+                    // B[i][j] = G(B[i][j-1], B[l][z])
+                    block(B, 256 * prev, 256 * refBlock, offset * 256, needXor);
+                    yield;
+                }
+            }
+        }
+    }
+    clean(address);
+}
+function argon2(type, password, salt, opts) {
+    const ctx = argon2Init(password, salt, type, opts);
+    const blocks = argon2Blocks(ctx, new Uint32Array(3 * 256));
+    while (!blocks.next().done) { }
+    return argon2Output(ctx.B, ctx.p, ctx.laneLen, ctx.dkLen);
+}
+/**
+ * Argon2d GPU-resistant version.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2d.
+ * ```ts
+ * argon2d('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ * @example
+ * Derive a key with optional secret and scheduler controls.
+ * ```ts
+ * const progressLog: number[] = [];
+ * argon2d('password', 'salt1234', {
+ *   t: 1,
+ *   m: 8,
+ *   p: 1,
+ *   dkLen: 32,
+ *   version: 0x13,
+ *   key: 'secret',
+ *   personalization: 'application',
+ *   maxmem: 1024 * 1024,
+ *   asyncTick: 1,
+ *   onProgress(progress) {
+ *     progressLog.push(progress);
+ *   },
+ * });
+ * ```
+ */
+export const argon2d = (password, salt, opts = {}) => argon2(AT.Argon2d, password, salt, opts);
+/**
+ * Argon2i side-channel-resistant version.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2i.
+ * ```ts
+ * argon2i('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ */
+export const argon2i = (password, salt, opts = {}) => argon2(AT.Argon2i, password, salt, opts);
+/**
+ * Argon2id, combining i+d, the most popular version from RFC 9106.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2id.
+ * ```ts
+ * argon2id('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ */
+export const argon2id = (password, salt, opts = {}) => argon2(AT.Argon2id, password, salt, opts);
+async function argon2Async(type, password, salt, opts) {
+    const ctx = argon2Init(password, salt, type, opts);
+    // Keep the generator-local address block reachable so an aborted scheduler yield can wipe it.
+    const address = new Uint32Array(3 * 256);
+    const blocks = argon2Blocks(ctx, address);
+    const abort = () => clean(address, ctx.B);
+    let ts = Date.now();
+    while (!blocks.next().done) {
+        // Date.now() is not monotonic. If the clock goes backwards,
+        // still yield control.
+        const diff = Date.now() - ts;
+        if (diff >= 0 && diff < ctx.asyncTick)
+            continue;
+        await nextTick(abort);
+        // Scheduler delay is outside the synchronous work budget.
+        ts = Date.now();
+    }
+    return argon2Output(ctx.B, ctx.p, ctx.laneLen, ctx.dkLen);
+}
+/**
+ * Argon2d async GPU-resistant version.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Promise resolving to derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2d asynchronously.
+ * ```ts
+ * await argon2dAsync('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ * @example
+ * Derive a key asynchronously with optional secret and scheduler controls.
+ * ```ts
+ * await argon2dAsync('password', 'salt1234', {
+ *   t: 1,
+ *   m: 8,
+ *   p: 1,
+ *   dkLen: 32,
+ *   version: 0x13,
+ *   key: 'secret',
+ *   personalization: 'application',
+ *   maxmem: 1024 * 1024,
+ *   asyncTick: 1,
+ *   onProgress(progress) {
+ *     if (progress > 1) throw new Error('invalid progress');
+ *   },
+ * });
+ * ```
+ */
+export const argon2dAsync = (password, salt, opts = {}) => argon2Async(AT.Argon2d, password, salt, opts);
+/**
+ * Argon2i async side-channel-resistant version.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Promise resolving to derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2i asynchronously.
+ * ```ts
+ * await argon2iAsync('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ */
+export const argon2iAsync = (password, salt, opts = {}) => argon2Async(AT.Argon2i, password, salt, opts);
+/**
+ * Argon2id async, combining i+d, the most popular version from RFC 9106.
+ * @param password - password or input key material
+ * @param salt - unique salt value
+ * @param opts - Argon2 cost and optional tuning parameters. See {@link ArgonOpts}.
+ * @returns Promise resolving to derived key bytes.
+ * @throws If the Argon2 input or cost parameters are invalid. {@link Error}
+ * @example
+ * Derive a key with Argon2id asynchronously.
+ * ```ts
+ * await argon2idAsync('password', 'salt1234', { t: 1, m: 8, p: 1, dkLen: 32 });
+ * ```
+ */
+export const argon2idAsync = (password, salt, opts = {}) => argon2Async(AT.Argon2id, password, salt, opts);

backend/static/vendor/noble-hashes/blake2.js

@@ -0,0 +1,573 @@
+/**
+ * blake2b (64-bit) & blake2s (8 to 32-bit) hash functions.
+ * b could have been faster, but there is no fast u64 in js, so s is 1.5x faster.
+ * @module
+ */
+import { BSIGMA, G1s, G2s } from "./_blake.js";
+import { SHA256_IV } from "./_md.js";
+import * as u64 from "./_u64.js";
+// prettier-ignore
+import { abytes, aexists, anumber, aoutput, checkOpts, clean, copyBytes, createHasher, swap32IfBE, swap8IfBE, u32 } from "./utils.js";
+// Same IV words as `SHA512_IV`, but endian-swapped into LE u32 low/high halves
+// for the BLAKE2b u64 helpers below.
+const B2B_IV = /* @__PURE__ */ Uint32Array.from([
+    0xf3bcc908, 0x6a09e667, 0x84caa73b, 0xbb67ae85, 0xfe94f82b, 0x3c6ef372, 0x5f1d36f1, 0xa54ff53a,
+    0xade682d1, 0x510e527f, 0x2b3e6c1f, 0x9b05688c, 0xfb41bd6b, 0x1f83d9ab, 0x137e2179, 0x5be0cd19,
+]);
+// Shared synchronous BLAKE2b work vector as LE u32 low/high halves.
+const BBUF = /* @__PURE__ */ new Uint32Array(32);
+// BLAKE2b G mix split into two half-rounds over LE u32 low/high limbs.
+function G1b(a, b, c, d, msg, x) {
+    // NOTE: V is LE here
+    const Xl = msg[x], Xh = msg[x + 1]; // prettier-ignore
+    let Al = BBUF[2 * a], Ah = BBUF[2 * a + 1]; // prettier-ignore
+    let Bl = BBUF[2 * b], Bh = BBUF[2 * b + 1]; // prettier-ignore
+    let Cl = BBUF[2 * c], Ch = BBUF[2 * c + 1]; // prettier-ignore
+    let Dl = BBUF[2 * d], Dh = BBUF[2 * d + 1]; // prettier-ignore
+    // v[a] = (v[a] + v[b] + x) | 0;
+    const ll = u64.add3L(Al, Bl, Xl);
+    Ah = u64.add3H(ll, Ah, Bh, Xh);
+    Al = ll | 0;
+    // v[d] = rotr(v[d] ^ v[a], 32)
+    let xh = Dh ^ Ah, xl = Dl ^ Al; // prettier-ignore
+    Dh = u64.rotr32H(xh, xl);
+    Dl = u64.rotr32L(xh, xl);
+    // v[c] = (v[c] + v[d]) | 0;
+    ({ h: Ch, l: Cl } = u64.add(Ch, Cl, Dh, Dl));
+    // v[b] = rotr(v[b] ^ v[c], 24)
+    xh = Bh ^ Ch;
+    xl = Bl ^ Cl;
+    Bh = u64.rotrSH(xh, xl, 24);
+    Bl = u64.rotrSL(xh, xl, 24);
+    BBUF[2 * a] = Al;
+    BBUF[2 * a + 1] = Ah;
+    BBUF[2 * b] = Bl;
+    BBUF[2 * b + 1] = Bh;
+    BBUF[2 * c] = Cl;
+    BBUF[2 * c + 1] = Ch;
+    BBUF[2 * d] = Dl;
+    BBUF[2 * d + 1] = Dh;
+}
+// Second half-round of the same LE-limb BLAKE2b G mix; `x` is the message word offset.
+function G2b(a, b, c, d, msg, x) {
+    // NOTE: V is LE here
+    const Xl = msg[x], Xh = msg[x + 1]; // prettier-ignore
+    let Al = BBUF[2 * a], Ah = BBUF[2 * a + 1]; // prettier-ignore
+    let Bl = BBUF[2 * b], Bh = BBUF[2 * b + 1]; // prettier-ignore
+    let Cl = BBUF[2 * c], Ch = BBUF[2 * c + 1]; // prettier-ignore
+    let Dl = BBUF[2 * d], Dh = BBUF[2 * d + 1]; // prettier-ignore
+    // v[a] = (v[a] + v[b] + x) | 0;
+    const ll = u64.add3L(Al, Bl, Xl);
+    Ah = u64.add3H(ll, Ah, Bh, Xh);
+    Al = ll | 0;
+    // v[d] = rotr(v[d] ^ v[a], 16)
+    let xh = Dh ^ Ah, xl = Dl ^ Al; // prettier-ignore
+    Dh = u64.rotrSH(xh, xl, 16);
+    Dl = u64.rotrSL(xh, xl, 16);
+    // v[c] = (v[c] + v[d]) | 0;
+    ({ h: Ch, l: Cl } = u64.add(Ch, Cl, Dh, Dl));
+    // v[b] = rotr(v[b] ^ v[c], 63)
+    xh = Bh ^ Ch;
+    xl = Bl ^ Cl;
+    Bh = u64.rotrBH(xh, xl, 63);
+    Bl = u64.rotrBL(xh, xl, 63);
+    BBUF[2 * a] = Al;
+    BBUF[2 * a + 1] = Ah;
+    BBUF[2 * b] = Bl;
+    BBUF[2 * b + 1] = Bh;
+    BBUF[2 * c] = Cl;
+    BBUF[2 * c + 1] = Ch;
+    BBUF[2 * d] = Dl;
+    BBUF[2 * d + 1] = Dh;
+}
+function checkBlake2Opts(outputLen, opts = {}, keyLen, saltLen, persLen) {
+    anumber(keyLen);
+    // RFC 7693 §2.1 requires digest length nn in 1..keyLen (keyLen doubles as
+    // the per-variant max for both key and digest lengths: 64 for b, 32 for s).
+    if (outputLen <= 0 || outputLen > keyLen)
+        throw new Error('"dkLen" must be 1..' + keyLen + ', got ' + outputLen);
+    const { key, salt, personalization } = opts;
+    // This API uses `undefined` for the RFC 7693 `kk = 0` case, so a provided key must be non-empty.
+    if (key !== undefined && (key.length < 1 || key.length > keyLen))
+        throw new Error('"key" expected to be undefined or of length=1..' + keyLen);
+    if (salt !== undefined)
+        abytes(salt, saltLen, 'salt');
+    if (personalization !== undefined)
+        abytes(personalization, persLen, 'personalization');
+}
+/** Internal base class for BLAKE2. */
+export class _BLAKE2 {
+    buffer;
+    buffer32;
+    finished = false;
+    destroyed = false;
+    length = 0;
+    pos = 0;
+    blockLen;
+    outputLen;
+    canXOF = false;
+    constructor(blockLen, outputLen) {
+        anumber(blockLen);
+        anumber(outputLen);
+        this.blockLen = blockLen;
+        this.outputLen = outputLen;
+        this.buffer = new Uint8Array(blockLen);
+        this.buffer32 = u32(this.buffer);
+    }
+    update(data) {
+        aexists(this);
+        abytes(data);
+        // Main difference with other hashes: there is flag for last block,
+        // so we cannot process current block before we know that there
+        // is the next one. This significantly complicates logic and reduces ability
+        // to do zero-copy processing
+        const { blockLen, buffer, buffer32 } = this;
+        const len = data.length;
+        const offset = data.byteOffset;
+        const buf = data.buffer;
+        for (let pos = 0; pos < len;) {
+            // If buffer is full and we still have input (don't process last block, same as blake2s)
+            if (this.pos === blockLen) {
+                swap32IfBE(buffer32);
+                this.compress(buffer32, 0, false);
+                swap32IfBE(buffer32);
+                this.pos = 0;
+            }
+            const take = Math.min(blockLen - this.pos, len - pos);
+            const dataOffset = offset + pos;
+            // Zero-copy only for full, 4-byte-aligned, non-final blocks.
+            if (take === blockLen && !(dataOffset % 4) && pos + take < len) {
+                const data32 = new Uint32Array(buf, dataOffset, Math.floor((len - pos) / 4));
+                swap32IfBE(data32);
+                for (let pos32 = 0; pos + blockLen < len; pos32 += buffer32.length, pos += blockLen) {
+                    this.length += blockLen;
+                    this.compress(data32, pos32, false);
+                }
+                swap32IfBE(data32);
+                continue;
+            }
+            // When the whole input is buffered in one go (common for short messages), passing `data`
+            // directly avoids allocating a subarray view.
+            buffer.set(pos === 0 && take === len ? data : data.subarray(pos, pos + take), this.pos);
+            this.pos += take;
+            this.length += take;
+            pos += take;
+        }
+        return this;
+    }
+    digestInto(out) {
+        aexists(this);
+        aoutput(out, this);
+        // Reject unaligned views explicitly instead of hiding them behind a full scratch copy.
+        if (out.byteOffset & 3)
+            throw new RangeError('"output" expected 4-byte aligned byteOffset, got ' + out.byteOffset);
+        const { pos, buffer32 } = this;
+        this.finished = true;
+        // Padding
+        this.buffer.fill(0, pos);
+        swap32IfBE(buffer32);
+        this.compress(buffer32, 0, true);
+        swap32IfBE(buffer32);
+        const state = this.get();
+        // digest() passes our own `buffer` as `out`; reuse its cached u32 view instead of allocating.
+        const out32 = out === this.buffer ? buffer32 : u32(out);
+        const full = Math.floor(this.outputLen / 4);
+        for (let i = 0; i < full; i++)
+            out32[i] = swap8IfBE(state[i]);
+        const tail = this.outputLen % 4;
+        if (!tail)
+            return;
+        const off = full * 4;
+        const word = state[full];
+        for (let i = 0; i < tail; i++)
+            out[off + i] = word >>> (8 * i);
+    }
+    digest() {
+        const { buffer, outputLen } = this;
+        this.digestInto(buffer);
+        // Return a copy so callers do not alias the instance scratch buffer used during finalization.
+        const res = buffer.slice(0, outputLen);
+        this.destroy();
+        return res;
+    }
+    _cloneInto(to) {
+        const { buffer, length, finished, destroyed, outputLen, pos } = this;
+        // Recreate only `dkLen`; key/salt/personalization are already absorbed into the copied state.
+        to ||= new this.constructor({ dkLen: outputLen });
+        to.set(...this.get());
+        // Last-block-aware lazy compression keeps the pending block live even when full.
+        to.buffer.set(buffer);
+        to.destroyed = destroyed;
+        to.finished = finished;
+        to.length = length;
+        to.pos = pos;
+        // @ts-ignore
+        to.outputLen = outputLen;
+        return to;
+    }
+    clone() {
+        return this._cloneInto();
+    }
+}
+/** Internal blake2b hash class with state stored as LE u32 low/high halves. */
+export class _BLAKE2b extends _BLAKE2 {
+    // Same IV words as SHA-512 / BLAKE2b, encoded as LE u32 low/high halves.
+    v0l = B2B_IV[0] | 0;
+    v0h = B2B_IV[1] | 0;
+    v1l = B2B_IV[2] | 0;
+    v1h = B2B_IV[3] | 0;
+    v2l = B2B_IV[4] | 0;
+    v2h = B2B_IV[5] | 0;
+    v3l = B2B_IV[6] | 0;
+    v3h = B2B_IV[7] | 0;
+    v4l = B2B_IV[8] | 0;
+    v4h = B2B_IV[9] | 0;
+    v5l = B2B_IV[10] | 0;
+    v5h = B2B_IV[11] | 0;
+    v6l = B2B_IV[12] | 0;
+    v6h = B2B_IV[13] | 0;
+    v7l = B2B_IV[14] | 0;
+    v7h = B2B_IV[15] | 0;
+    constructor(opts = {}) {
+        opts = checkOpts({}, opts);
+        const olen = opts.dkLen === undefined ? 64 : opts.dkLen;
+        super(128, olen);
+        checkBlake2Opts(olen, opts, 64, 16, 16);
+        let { key, personalization, salt } = opts;
+        let keyLength = 0;
+        if (key !== undefined) {
+            abytes(key, undefined, 'key');
+            keyLength = key.length;
+        }
+        // RFC 7693 §2.5: xor `p[0] = 0x0101kknn` into the low 32 bits of `h[0]`;
+        // the high 32 bits stay at `IV[0]`.
+        this.v0l ^= this.outputLen | (keyLength << 8) | (0x01 << 16) | (0x01 << 24);
+        if (salt !== undefined) {
+            abytes(salt, undefined, 'salt');
+            // Copy: u32() would throw on views with byteOffset not divisible by 4.
+            const slt = u32(copyBytes(salt));
+            this.v4l ^= swap8IfBE(slt[0]);
+            this.v4h ^= swap8IfBE(slt[1]);
+            this.v5l ^= swap8IfBE(slt[2]);
+            this.v5h ^= swap8IfBE(slt[3]);
+        }
+        if (personalization !== undefined) {
+            abytes(personalization, undefined, 'personalization');
+            // Copy: u32() would throw on views with byteOffset not divisible by 4.
+            const pers = u32(copyBytes(personalization));
+            this.v6l ^= swap8IfBE(pers[0]);
+            this.v6h ^= swap8IfBE(pers[1]);
+            this.v7l ^= swap8IfBE(pers[2]);
+            this.v7h ^= swap8IfBE(pers[3]);
+        }
+        if (key !== undefined) {
+            // Pad to blockLen and update
+            const tmp = new Uint8Array(this.blockLen);
+            tmp.set(key);
+            this.update(tmp);
+            // The padded copy holds key material; buffer/state keep what they need.
+            clean(tmp);
+        }
+    }
+    // prettier-ignore
+    get() {
+        let { v0l, v0h, v1l, v1h, v2l, v2h, v3l, v3h, v4l, v4h, v5l, v5h, v6l, v6h, v7l, v7h } = this;
+        return [v0l, v0h, v1l, v1h, v2l, v2h, v3l, v3h, v4l, v4h, v5l, v5h, v6l, v6h, v7l, v7h];
+    }
+    // prettier-ignore
+    set(v0l, v0h, v1l, v1h, v2l, v2h, v3l, v3h, v4l, v4h, v5l, v5h, v6l, v6h, v7l, v7h) {
+        this.v0l = v0l | 0;
+        this.v0h = v0h | 0;
+        this.v1l = v1l | 0;
+        this.v1h = v1h | 0;
+        this.v2l = v2l | 0;
+        this.v2h = v2h | 0;
+        this.v3l = v3l | 0;
+        this.v3h = v3h | 0;
+        this.v4l = v4l | 0;
+        this.v4h = v4h | 0;
+        this.v5l = v5l | 0;
+        this.v5h = v5h | 0;
+        this.v6l = v6l | 0;
+        this.v6h = v6h | 0;
+        this.v7l = v7l | 0;
+        this.v7h = v7h | 0;
+    }
+    compress(msg, offset, isLast) {
+        // First half from state. Direct writes: get() would allocate an array +
+        // closure per block.
+        // prettier-ignore
+        const { v0l, v0h, v1l, v1h, v2l, v2h, v3l, v3h, v4l, v4h, v5l, v5h, v6l, v6h, v7l, v7h } = this;
+        // prettier-ignore
+        {
+            BBUF[0] = v0l;
+            BBUF[1] = v0h;
+            BBUF[2] = v1l;
+            BBUF[3] = v1h;
+            BBUF[4] = v2l;
+            BBUF[5] = v2h;
+            BBUF[6] = v3l;
+            BBUF[7] = v3h;
+            BBUF[8] = v4l;
+            BBUF[9] = v4h;
+            BBUF[10] = v5l;
+            BBUF[11] = v5h;
+            BBUF[12] = v6l;
+            BBUF[13] = v6h;
+            BBUF[14] = v7l;
+            BBUF[15] = v7h;
+        }
+        BBUF.set(B2B_IV, 16); // Second half from IV.
+        const l = u64.fromNumL(this.length);
+        const h = u64.fromNumH(this.length);
+        BBUF[24] = B2B_IV[8] ^ l; // Low word of the offset.
+        BBUF[25] = B2B_IV[9] ^ h; // High word.
+        // Invert all bits for last block
+        if (isLast) {
+            BBUF[28] = ~BBUF[28];
+            BBUF[29] = ~BBUF[29];
+        }
+        let j = 0;
+        const s = BSIGMA;
+        // SIGMA selects 64-bit message words; multiply by 2 because `msg` stores
+        // each word as [low32, high32].
+        for (let i = 0; i < 12; i++) {
+            G1b(0, 4, 8, 12, msg, offset + 2 * s[j++]);
+            G2b(0, 4, 8, 12, msg, offset + 2 * s[j++]);
+            G1b(1, 5, 9, 13, msg, offset + 2 * s[j++]);
+            G2b(1, 5, 9, 13, msg, offset + 2 * s[j++]);
+            G1b(2, 6, 10, 14, msg, offset + 2 * s[j++]);
+            G2b(2, 6, 10, 14, msg, offset + 2 * s[j++]);
+            G1b(3, 7, 11, 15, msg, offset + 2 * s[j++]);
+            G2b(3, 7, 11, 15, msg, offset + 2 * s[j++]);
+            G1b(0, 5, 10, 15, msg, offset + 2 * s[j++]);
+            G2b(0, 5, 10, 15, msg, offset + 2 * s[j++]);
+            G1b(1, 6, 11, 12, msg, offset + 2 * s[j++]);
+            G2b(1, 6, 11, 12, msg, offset + 2 * s[j++]);
+            G1b(2, 7, 8, 13, msg, offset + 2 * s[j++]);
+            G2b(2, 7, 8, 13, msg, offset + 2 * s[j++]);
+            G1b(3, 4, 9, 14, msg, offset + 2 * s[j++]);
+            G2b(3, 4, 9, 14, msg, offset + 2 * s[j++]);
+        }
+        this.v0l ^= BBUF[0] ^ BBUF[16];
+        this.v0h ^= BBUF[1] ^ BBUF[17];
+        this.v1l ^= BBUF[2] ^ BBUF[18];
+        this.v1h ^= BBUF[3] ^ BBUF[19];
+        this.v2l ^= BBUF[4] ^ BBUF[20];
+        this.v2h ^= BBUF[5] ^ BBUF[21];
+        this.v3l ^= BBUF[6] ^ BBUF[22];
+        this.v3h ^= BBUF[7] ^ BBUF[23];
+        this.v4l ^= BBUF[8] ^ BBUF[24];
+        this.v4h ^= BBUF[9] ^ BBUF[25];
+        this.v5l ^= BBUF[10] ^ BBUF[26];
+        this.v5h ^= BBUF[11] ^ BBUF[27];
+        this.v6l ^= BBUF[12] ^ BBUF[28];
+        this.v6h ^= BBUF[13] ^ BBUF[29];
+        this.v7l ^= BBUF[14] ^ BBUF[30];
+        this.v7h ^= BBUF[15] ^ BBUF[31];
+        clean(BBUF);
+    }
+    destroy() {
+        this.destroyed = true;
+        clean(this.buffer32);
+        this.set(0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0);
+    }
+}
+/**
+ * Blake2b hash function. 64-bit. 1.5x slower than blake2s in JS.
+ * @param msg - message that would be hashed
+ * @param opts - Optional output, MAC, salt, and personalization settings.
+ *   `dkLen` must be 1..64 bytes; `salt` and `personalization`, if present,
+ *   must be 16 bytes each. See {@link Blake2Opts}.
+ * @returns Digest bytes.
+ * @example
+ * Hash a message with Blake2b.
+ * ```ts
+ * blake2b(new Uint8Array([97, 98, 99]));
+ * ```
+ * @example
+ * Hash a message with Blake2b while selecting output, MAC, salt, and personalization settings.
+ * ```ts
+ * blake2b(new Uint8Array([97, 98, 99]), {
+ *   dkLen: 32,
+ *   key: new Uint8Array(32),
+ *   salt: new Uint8Array(16),
+ *   personalization: new Uint8Array(16),
+ * });
+ * ```
+ */
+export const blake2b = /* @__PURE__ */ createHasher((opts) => new _BLAKE2b(opts));
+/**
+ * BLAKE2-compress core method.
+ * Runs only the round function over a caller-supplied local vector; callers initialize `v0..v15`
+ * and apply the final `h[i] ^= v[i] ^ v[i + 8]` fold themselves.
+ * @param s - flattened sigma schedule bytes
+ * @param offset - starting word offset inside `msg`, not a byte offset
+ * @param msg - message words
+ * @param rounds - round count to execute
+ * @param v0 - state word 0
+ * @param v1 - state word 1
+ * @param v2 - state word 2
+ * @param v3 - state word 3
+ * @param v4 - state word 4
+ * @param v5 - state word 5
+ * @param v6 - state word 6
+ * @param v7 - state word 7
+ * @param v8 - state word 8
+ * @param v9 - state word 9
+ * @param v10 - state word 10
+ * @param v11 - state word 11
+ * @param v12 - state word 12
+ * @param v13 - state word 13
+ * @param v14 - state word 14
+ * @param v15 - state word 15
+ * @returns Updated compression state words.
+ * @example
+ * Run the BLAKE2 compression core on zeroed state and message words.
+ * ```ts
+ * import { _compress } from '@noble/hashes/blake2.js';
+ * const state = _compress(
+ *   new Uint8Array(16),
+ *   0,
+ *   new Uint32Array(16),
+ *   1,
+ *   0, 0, 0, 0, 0, 0, 0, 0,
+ *   0, 0, 0, 0, 0, 0, 0, 0
+ * );
+ * state.v0;
+ * ```
+ */
+// prettier-ignore
+export function _compress(s, offset, msg, rounds, v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15) {
+    let j = 0;
+    for (let i = 0; i < rounds; i++) {
+        ({ a: v0, b: v4, c: v8, d: v12 } = G1s(v0, v4, v8, v12, msg[offset + s[j++]]));
+        ({ a: v0, b: v4, c: v8, d: v12 } = G2s(v0, v4, v8, v12, msg[offset + s[j++]]));
+        ({ a: v1, b: v5, c: v9, d: v13 } = G1s(v1, v5, v9, v13, msg[offset + s[j++]]));
+        ({ a: v1, b: v5, c: v9, d: v13 } = G2s(v1, v5, v9, v13, msg[offset + s[j++]]));
+        ({ a: v2, b: v6, c: v10, d: v14 } = G1s(v2, v6, v10, v14, msg[offset + s[j++]]));
+        ({ a: v2, b: v6, c: v10, d: v14 } = G2s(v2, v6, v10, v14, msg[offset + s[j++]]));
+        ({ a: v3, b: v7, c: v11, d: v15 } = G1s(v3, v7, v11, v15, msg[offset + s[j++]]));
+        ({ a: v3, b: v7, c: v11, d: v15 } = G2s(v3, v7, v11, v15, msg[offset + s[j++]]));
+        ({ a: v0, b: v5, c: v10, d: v15 } = G1s(v0, v5, v10, v15, msg[offset + s[j++]]));
+        ({ a: v0, b: v5, c: v10, d: v15 } = G2s(v0, v5, v10, v15, msg[offset + s[j++]]));
+        ({ a: v1, b: v6, c: v11, d: v12 } = G1s(v1, v6, v11, v12, msg[offset + s[j++]]));
+        ({ a: v1, b: v6, c: v11, d: v12 } = G2s(v1, v6, v11, v12, msg[offset + s[j++]]));
+        ({ a: v2, b: v7, c: v8, d: v13 } = G1s(v2, v7, v8, v13, msg[offset + s[j++]]));
+        ({ a: v2, b: v7, c: v8, d: v13 } = G2s(v2, v7, v8, v13, msg[offset + s[j++]]));
+        ({ a: v3, b: v4, c: v9, d: v14 } = G1s(v3, v4, v9, v14, msg[offset + s[j++]]));
+        ({ a: v3, b: v4, c: v9, d: v14 } = G2s(v3, v4, v9, v14, msg[offset + s[j++]]));
+    }
+    return { v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15 };
+}
+// Blake2s reuses the SHA-256 IV words as-is.
+const B2S_IV = /* @__PURE__ */ SHA256_IV.slice();
+/** Internal blake2s hash class. */
+export class _BLAKE2s extends _BLAKE2 {
+    // Internal state, same as SHA-256
+    v0 = B2S_IV[0] | 0;
+    v1 = B2S_IV[1] | 0;
+    v2 = B2S_IV[2] | 0;
+    v3 = B2S_IV[3] | 0;
+    v4 = B2S_IV[4] | 0;
+    v5 = B2S_IV[5] | 0;
+    v6 = B2S_IV[6] | 0;
+    v7 = B2S_IV[7] | 0;
+    constructor(opts = {}) {
+        opts = checkOpts({}, opts);
+        const olen = opts.dkLen === undefined ? 32 : opts.dkLen;
+        super(64, olen);
+        checkBlake2Opts(olen, opts, 32, 8, 8);
+        let { key, personalization, salt } = opts;
+        let keyLength = 0;
+        if (key !== undefined) {
+            abytes(key, undefined, 'key');
+            keyLength = key.length;
+        }
+        // RFC 7693 §2.5: xor `p[0] = 0x0101kknn` directly into `h[0]`, since
+        // BLAKE2s stores each state word as one `u32`.
+        this.v0 ^= this.outputLen | (keyLength << 8) | (0x01 << 16) | (0x01 << 24);
+        if (salt !== undefined) {
+            abytes(salt, undefined, 'salt');
+            // Copy: u32() would throw on views with byteOffset not divisible by 4.
+            const slt = u32(copyBytes(salt));
+            this.v4 ^= swap8IfBE(slt[0]);
+            this.v5 ^= swap8IfBE(slt[1]);
+        }
+        if (personalization !== undefined) {
+            abytes(personalization, undefined, 'personalization');
+            // Copy: u32() would throw on views with byteOffset not divisible by 4.
+            const pers = u32(copyBytes(personalization));
+            this.v6 ^= swap8IfBE(pers[0]);
+            this.v7 ^= swap8IfBE(pers[1]);
+        }
+        if (key !== undefined) {
+            // Pad to blockLen and update
+            const tmp = new Uint8Array(this.blockLen);
+            tmp.set(key);
+            this.update(tmp);
+            // The padded copy holds key material; buffer/state keep what they need.
+            clean(tmp);
+        }
+    }
+    get() {
+        const { v0, v1, v2, v3, v4, v5, v6, v7 } = this;
+        return [v0, v1, v2, v3, v4, v5, v6, v7];
+    }
+    // prettier-ignore
+    set(v0, v1, v2, v3, v4, v5, v6, v7) {
+        this.v0 = v0 | 0;
+        this.v1 = v1 | 0;
+        this.v2 = v2 | 0;
+        this.v3 = v3 | 0;
+        this.v4 = v4 | 0;
+        this.v5 = v5 | 0;
+        this.v6 = v6 | 0;
+        this.v7 = v7 | 0;
+    }
+    compress(msg, offset, isLast) {
+        const l = u64.fromNumL(this.length);
+        const h = u64.fromNumH(this.length);
+        // Seed v8..v15 from the IV, xor the low/high 32-bit byte counter into
+        // v12/v13, and invert v14 on the final block.
+        // prettier-ignore
+        const { v0, v1, v2, v3, v4, v5, v6, v7, v8, v9, v10, v11, v12, v13, v14, v15 } = _compress(BSIGMA, offset, msg, 10, this.v0, this.v1, this.v2, this.v3, this.v4, this.v5, this.v6, this.v7, B2S_IV[0], B2S_IV[1], B2S_IV[2], B2S_IV[3], l ^ B2S_IV[4], h ^ B2S_IV[5], isLast ? ~B2S_IV[6] : B2S_IV[6], B2S_IV[7]);
+        this.v0 ^= v0 ^ v8;
+        this.v1 ^= v1 ^ v9;
+        this.v2 ^= v2 ^ v10;
+        this.v3 ^= v3 ^ v11;
+        this.v4 ^= v4 ^ v12;
+        this.v5 ^= v5 ^ v13;
+        this.v6 ^= v6 ^ v14;
+        this.v7 ^= v7 ^ v15;
+    }
+    destroy() {
+        this.destroyed = true;
+        clean(this.buffer32);
+        this.set(0, 0, 0, 0, 0, 0, 0, 0);
+    }
+}
+/**
+ * Blake2s hash function. Focuses on 8-bit to 32-bit platforms. 1.5x faster than blake2b in JS.
+ * @param msg - message that would be hashed
+ * @param opts - Optional output, MAC, salt, and personalization settings.
+ *   `dkLen` must be 1..32 bytes; `salt` and `personalization`, if present,
+ *   must be 8 bytes each. See {@link Blake2Opts}.
+ * @returns Digest bytes.
+ * @example
+ * Hash a message with Blake2s.
+ * ```ts
+ * blake2s(new Uint8Array([97, 98, 99]));
+ * ```
+ * @example
+ * Hash a message with Blake2s while selecting output, MAC, salt, and personalization settings.
+ * ```ts
+ * blake2s(new Uint8Array([97, 98, 99]), {
+ *   dkLen: 16,
+ *   key: new Uint8Array(32),
+ *   salt: new Uint8Array(8),
+ *   personalization: new Uint8Array(8),
+ * });
+ * ```
+ */
+export const blake2s = /* @__PURE__ */ createHasher((opts) => new _BLAKE2s(opts));

backend/static/vendor/noble-hashes/utils.js

@@ -0,0 +1,687 @@
+/**
+ * Checks if something is Uint8Array. Be careful: nodejs Buffer will return true.
+ * @param a - value to test
+ * @returns `true` when the value is a Uint8Array-compatible view.
+ * @example
+ * Check whether a value is a Uint8Array-compatible view.
+ * ```ts
+ * isBytes(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export function isBytes(a) {
+    // Plain `instanceof Uint8Array` is too strict for some Buffer / proxy / cross-realm cases.
+    // The fallback still requires a real ArrayBuffer view, so plain
+    // JSON-deserialized `{ constructor: ... }` spoofing is rejected, and
+    // `BYTES_PER_ELEMENT === 1` keeps the fallback on byte-oriented views.
+    return (a instanceof Uint8Array ||
+        (ArrayBuffer.isView(a) &&
+            a.constructor.name === 'Uint8Array' &&
+            'BYTES_PER_ELEMENT' in a &&
+            a.BYTES_PER_ELEMENT === 1));
+}
+// Shared error-message prefix builder. Only called on throw paths, so assert
+// success paths never pay for the string concatenation.
+const atitle = (title) => (title ? `"${title}" ` : '');
+/**
+ * Asserts something is a non-negative integer.
+ * @param n - number to validate
+ * @param title - label included in thrown errors
+ * @returns The validated number.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Validate a non-negative integer option.
+ * ```ts
+ * anumber(32, 'length');
+ * ```
+ */
+export function anumber(n, title = '') {
+    if (typeof n !== 'number')
+        throw new TypeError(atitle(title) + 'expected number, got ' + typeof n);
+    if (!Number.isSafeInteger(n) || n < 0)
+        throw new RangeError(atitle(title) + 'expected integer >= 0, got ' + n);
+    return n;
+}
+/**
+ * Asserts something is a boolean.
+ * @param value - value to validate
+ * @param title - label included in thrown errors
+ * @returns The validated boolean.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Validate a boolean option.
+ * ```ts
+ * abool(true, 'enableXOF');
+ * ```
+ */
+export function abool(value, title = '') {
+    if (typeof value !== 'boolean')
+        throw new TypeError(atitle(title) + 'expected boolean, got type=' + typeof value);
+    return value;
+}
+/**
+ * Asserts something is Uint8Array.
+ * @param value - value to validate
+ * @param length - optional exact length constraint
+ * @param title - label included in thrown errors
+ * @returns The validated byte array.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Validate that a value is a byte array.
+ * ```ts
+ * abytes(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export function abytes(value, length, title = '') {
+    // Success path first: this runs at the start of every update() / digestInto(), and the
+    // common `abytes(data)` form must not pay for length handling it does not use.
+    if (isBytes(value) && (length === undefined || value.length === length))
+        return value;
+    // Error path: recompute freely to build the exact message.
+    if (length !== undefined)
+        anumber(length, 'length');
+    const bytes = isBytes(value);
+    const ofLen = length !== undefined ? ` of length ${length}` : '';
+    const got = bytes ? `length=${value.length}` : `type=${typeof value}`;
+    const message = atitle(title) + 'expected Uint8Array' + ofLen + ', got ' + got;
+    if (!bytes)
+        throw new TypeError(message);
+    throw new RangeError(message);
+}
+/**
+ * Copies bytes into a fresh Uint8Array.
+ * Buffer-style slices can alias the same backing store, so callers that need ownership should copy.
+ * @param bytes - source bytes to clone
+ * @returns Freshly allocated copy of `bytes`.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Clone a byte array before mutating it.
+ * ```ts
+ * const copy = copyBytes(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export function copyBytes(bytes) {
+    // `Uint8Array.from(...)` would also accept arrays / other typed arrays. Keep this helper strict
+    // because callers use it at byte-validation boundaries before mutating the detached copy.
+    return Uint8Array.from(abytes(bytes));
+}
+/**
+ * Asserts something is a wrapped hash constructor.
+ * @param h - hash constructor to validate
+ * @throws On wrong argument types or invalid hash wrapper shape. {@link TypeError}
+ * @throws On invalid hash metadata ranges or values. {@link RangeError}
+ * @throws If the hash metadata allows empty outputs or block sizes. {@link Error}
+ * @example
+ * Validate a callable hash wrapper.
+ * ```ts
+ * import { ahash } from '@noble/hashes/utils.js';
+ * import { sha256 } from '@noble/hashes/sha2.js';
+ * ahash(sha256);
+ * ```
+ */
+export function ahash(h) {
+    if (typeof h !== 'function' || typeof h.create !== 'function')
+        throw new TypeError('expected hash wrapped by utils.createHasher');
+    anumber(h.outputLen);
+    anumber(h.blockLen);
+    // HMAC and KDF callers treat these as real byte lengths; allowing zero lets fake wrappers pass
+    // validation and can produce empty outputs instead of failing fast.
+    if (h.outputLen < 1 || h.blockLen < 1)
+        throw new Error('hash blockLen / outputLen must be >= 1');
+}
+const aobject = (value, label) => {
+    if (value === null || typeof value !== 'object' || Array.isArray(value))
+        throw new TypeError((label === 'object' ? '' : `"${label}" `) + 'expected object, got type=' + typeof value);
+};
+const aopts = (value, label) => {
+    aobject(value, label);
+    const proto = Object.getPrototypeOf(value);
+    if (proto !== Object.prototype && proto !== null)
+        throw new TypeError(`"${label}" expected plain object`);
+    // Object.assign() treats an own "__proto__" source key as a write to the target's legacy
+    // prototype setter. Reject it before merging so inherited option values cannot be injected.
+    if (Object.hasOwn(value, '__proto__'))
+        throw new TypeError(`"${label}.__proto__" is not allowed`);
+};
+/**
+ * Asserts a hash instance has not been destroyed or finished.
+ * @param instance - hash instance to validate
+ * @param checkFinished - whether to reject finalized instances
+ * @throws If the hash instance has already been destroyed or finalized. {@link Error}
+ * @example
+ * Validate that a hash instance is still usable.
+ * ```ts
+ * import { aexists } from '@noble/hashes/utils.js';
+ * import { sha256 } from '@noble/hashes/sha2.js';
+ * const hash = sha256.create();
+ * aexists(hash);
+ * ```
+ */
+export function aexists(instance, checkFinished = true) {
+    // Runs on every update()/digestInto(); the flags are library-owned booleans, so only their
+    // truthiness is checked - re-validating their type per call was pure hot-path overhead.
+    if (instance.destroyed)
+        throw new Error('hash was destroyed');
+    if (checkFinished && instance.finished)
+        throw new Error('digest() was already called');
+}
+/**
+ * Asserts output is a sufficiently-sized byte array.
+ * @param out - destination buffer
+ * @param instance - hash instance providing output length
+ * Oversized buffers are allowed; downstream code only promises to fill the first `outputLen` bytes.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Validate a caller-provided digest buffer.
+ * ```ts
+ * import { aoutput } from '@noble/hashes/utils.js';
+ * import { sha256 } from '@noble/hashes/sha2.js';
+ * const hash = sha256.create();
+ * aoutput(new Uint8Array(hash.outputLen), hash);
+ * ```
+ */
+export function aoutput(out, instance) {
+    abytes(out, undefined, 'output');
+    // `outputLen` is a library-owned readonly number; the negated comparison keeps failing fast
+    // when it is missing/NaN (comparisons with undefined/NaN are false) without an anumber() call.
+    const min = instance.outputLen;
+    if (!(out.length >= min)) {
+        throw new RangeError('"output" expected length >= ' + min);
+    }
+}
+/**
+ * Casts a typed array view to Uint8Array.
+ * @param arr - source typed array
+ * @returns Uint8Array view over the same buffer.
+ * @example
+ * Reinterpret a typed array as bytes.
+ * ```ts
+ * u8(new Uint32Array([1, 2]));
+ * ```
+ */
+export function u8(arr) {
+    return new Uint8Array(arr.buffer, arr.byteOffset, arr.byteLength);
+}
+/**
+ * Casts a typed array view to Uint32Array.
+ * `arr.byteOffset` must already be 4-byte aligned or the platform
+ * Uint32Array constructor will throw.
+ * @param arr - source typed array
+ * @returns Uint32Array view over the same buffer.
+ * @example
+ * Reinterpret a byte array as 32-bit words.
+ * ```ts
+ * u32(new Uint8Array(8));
+ * ```
+ */
+export function u32(arr) {
+    return new Uint32Array(arr.buffer, arr.byteOffset, Math.floor(arr.byteLength / 4));
+}
+/**
+ * Zeroizes typed arrays in place. Warning: JS provides no guarantees.
+ * @param arrays - arrays to overwrite with zeros
+ * @example
+ * Zeroize sensitive buffers in place.
+ * ```ts
+ * clean(new Uint8Array([1, 2, 3]));
+ * ```
+ */
+export function clean(...arrays) {
+    for (let i = 0; i < arrays.length; i++) {
+        arrays[i].fill(0);
+    }
+}
+/**
+ * Creates a DataView for byte-level manipulation.
+ * @param arr - source typed array
+ * @returns DataView over the same buffer region.
+ * @example
+ * Create a DataView over an existing buffer.
+ * ```ts
+ * createView(new Uint8Array(4));
+ * ```
+ */
+export function createView(arr) {
+    return new DataView(arr.buffer, arr.byteOffset, arr.byteLength);
+}
+/**
+ * Rotate-right operation for uint32 values.
+ * @param word - source word
+ * @param shift - shift amount in bits
+ * @returns Rotated word.
+ * @example
+ * Rotate a 32-bit word to the right.
+ * ```ts
+ * rotr(0x12345678, 8);
+ * ```
+ */
+export function rotr(word, shift) {
+    return (word << (32 - shift)) | (word >>> shift);
+}
+/**
+ * Rotate-left operation for uint32 values.
+ * @param word - source word
+ * @param shift - shift amount in bits
+ * @returns Rotated word.
+ * @example
+ * Rotate a 32-bit word to the left.
+ * ```ts
+ * rotl(0x12345678, 8);
+ * ```
+ */
+export function rotl(word, shift) {
+    return (word << shift) | ((word >>> (32 - shift)) >>> 0);
+}
+/** Whether the current platform is little-endian. */
+export const isLE = /* @__PURE__ */ (() => new Uint8Array(new Uint32Array([0x11223344]).buffer)[0] === 0x44)();
+/**
+ * Byte-swap operation for uint32 values.
+ * @param word - source word
+ * @returns Word with reversed byte order.
+ * @example
+ * Reverse the byte order of a 32-bit word.
+ * ```ts
+ * byteSwap(0x11223344);
+ * ```
+ */
+export function byteSwap(word) {
+    return (((word << 24) & 0xff000000) |
+        ((word << 8) & 0xff0000) |
+        ((word >>> 8) & 0xff00) |
+        ((word >>> 24) & 0xff));
+}
+/**
+ * Conditionally byte-swaps one 32-bit word on big-endian platforms.
+ * @param n - source word
+ * @returns Original or byte-swapped word depending on platform endianness.
+ * @example
+ * Normalize a 32-bit word for host endianness.
+ * ```ts
+ * swap8IfBE(0x11223344);
+ * ```
+ */
+export const swap8IfBE = isLE
+    ? (n) => n
+    : (n) => byteSwap(n) >>> 0;
+/**
+ * Byte-swaps every word of a Uint32Array in place.
+ * @param arr - array to mutate
+ * @returns The same array after mutation; callers pass live state arrays here.
+ * @example
+ * Reverse the byte order of every word in place.
+ * ```ts
+ * byteSwap32(new Uint32Array([0x11223344]));
+ * ```
+ */
+export function byteSwap32(arr) {
+    for (let i = 0; i < arr.length; i++) {
+        arr[i] = byteSwap(arr[i]);
+    }
+    return arr;
+}
+/**
+ * Conditionally byte-swaps a Uint32Array on big-endian platforms.
+ * @param u - array to normalize for host endianness
+ * @returns Original or byte-swapped array depending on platform endianness.
+ *   On big-endian runtimes this mutates `u` in place via `byteSwap32(...)`.
+ * @example
+ * Normalize a word array for host endianness.
+ * ```ts
+ * swap32IfBE(new Uint32Array([0x11223344]));
+ * ```
+ */
+export const swap32IfBE = isLE
+    ? (u) => u
+    : byteSwap32;
+// Built-in hex conversion https://caniuse.com/mdn-javascript_builtins_uint8array_fromhex
+const hasHexBuiltin = /* @__PURE__ */ (() => 
+// @ts-ignore
+typeof Uint8Array.from([]).toHex === 'function' && typeof Uint8Array.fromHex === 'function')();
+// Array where index 0xf0 (240) is mapped to string 'f0'
+const hexes = /* @__PURE__ */ Array.from({ length: 256 }, (_, i) => i.toString(16).padStart(2, '0'));
+/**
+ * Convert byte array to hex string.
+ * Uses the built-in function when available and assumes it matches the tested
+ * fallback semantics.
+ * @param bytes - bytes to encode
+ * @returns Lowercase hexadecimal string.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Convert bytes to lowercase hexadecimal.
+ * ```ts
+ * bytesToHex(Uint8Array.from([0xca, 0xfe, 0x01, 0x23])); // 'cafe0123'
+ * ```
+ */
+export function bytesToHex(bytes) {
+    abytes(bytes);
+    // @ts-ignore
+    if (hasHexBuiltin)
+        return bytes.toHex();
+    // pre-caching improves the speed 6x
+    let hex = '';
+    for (let i = 0; i < bytes.length; i++) {
+        hex += hexes[bytes[i]];
+    }
+    return hex;
+}
+// Strict ASCII nibble parser: non-ASCII hex lookalikes are rejected as undefined.
+// ASCII codes: '0'..'9' = 48..57, 'A'..'F' = 65..70, 'a'..'f' = 97..102.
+// prettier-ignore
+function asciiToBase16(ch) {
+    return ch >= 48 && ch <= 57 ? ch - 48 // '2' => 50-48
+        : ch >= 65 && ch <= 70 ? ch - (65 - 10) // 'B' => 66-(65-10)
+            : ch >= 97 && ch <= 102 ? ch - (97 - 10) // 'b' => 98-(97-10)
+                : undefined;
+}
+/**
+ * Convert hex string to byte array. Uses built-in function, when available.
+ * @param hex - hexadecimal string to decode
+ * @returns Decoded bytes.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Decode lowercase hexadecimal into bytes.
+ * ```ts
+ * hexToBytes('cafe0123'); // Uint8Array.from([0xca, 0xfe, 0x01, 0x23])
+ * ```
+ */
+export function hexToBytes(hex) {
+    if (typeof hex !== 'string')
+        throw new TypeError('hex string expected, got ' + typeof hex);
+    if (hasHexBuiltin) {
+        try {
+            return Uint8Array.fromHex(hex);
+        }
+        catch (error) {
+            if (error instanceof SyntaxError)
+                throw new RangeError(error.message);
+            throw error;
+        }
+    }
+    const hl = hex.length;
+    const al = hl / 2;
+    if (hl % 2)
+        throw new RangeError('hex string expected, got unpadded hex of length ' + hl);
+    const array = new Uint8Array(al);
+    for (let ai = 0, hi = 0; ai < al; ai++, hi += 2) {
+        const n1 = asciiToBase16(hex.charCodeAt(hi)); // parse first char, multiply it by 16
+        const n2 = asciiToBase16(hex.charCodeAt(hi + 1)); // parse second char
+        if (n1 === undefined || n2 === undefined) {
+            const char = hex[hi] + hex[hi + 1];
+            throw new RangeError('hex string expected, got non-hex character "' + char + '" at index ' + hi);
+        }
+        array[ai] = n1 * 16 + n2; // example: 'A9' => 10*16 + 9
+    }
+    return array;
+}
+/**
+ * Yields to the host task scheduler so timers, I/O, and rendering can make progress.
+ * Uses the Web Scheduling API when available or `setTimeout` as a cross-platform fallback.
+ * Host-task yields are much slower than microtasks (roughly 1ms with the timer fallback), so
+ * async loops should use `asyncTick >= 10` to amortize scheduling overhead to about 10% while
+ * still allowing other event-loop work to proceed.
+ * @param onReject - optional cleanup invoked only if the host yield fails
+ * @example
+ * Yield to the next scheduler tick.
+ * ```ts
+ * await nextTick();
+ * ```
+ */
+export function nextTick(onReject) {
+    const host = globalThis;
+    if (typeof host.scheduler?.yield === 'function') {
+        const promise = host.scheduler.yield();
+        // Keep the original scheduler rejection; this handler exists only for cleanup.
+        if (onReject)
+            promise.catch(onReject);
+        return promise;
+    }
+    return new Promise((resolve) => host.setTimeout(resolve, 0));
+}
+/**
+ * Returns control to the host event loop every `tick` milliseconds to avoid blocking long loops.
+ * @param iters - number of loop iterations to run
+ * @param tick - maximum time slice in milliseconds
+ * @param cb - callback executed on each iteration
+ * @param onReject - optional cleanup invoked only if a host yield fails
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @example
+ * Run a loop that periodically yields back to the event loop.
+ * ```ts
+ * await asyncLoop(2, 0, () => {});
+ * ```
+ */
+export async function asyncLoop(iters, tick, cb, onReject) {
+    anumber(iters, 'iters');
+    anumber(tick, 'tick');
+    if (typeof cb !== 'function')
+        throw new TypeError('callback must be a function');
+    // Callback is synchronous by contract; asyncLoop only yields between sync work windows.
+    let ts = Date.now();
+    for (let i = 0; i < iters; i++) {
+        cb(i);
+        // Date.now() is not monotonic, so in case if clock goes backwards we return return control too
+        const diff = Date.now() - ts;
+        if (diff >= 0 && diff < tick)
+            continue;
+        await nextTick(onReject);
+        // Track only synchronous work time; scheduler delay after yielding is outside our budget.
+        ts = Date.now();
+    }
+}
+/**
+ * Converts string to bytes using UTF8 encoding.
+ * Built-in doesn't validate input to be string: we do the check.
+ * Non-ASCII details are delegated to the platform `TextEncoder`.
+ * @param str - string to encode
+ * @returns UTF-8 encoded bytes.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Encode a string as UTF-8 bytes.
+ * ```ts
+ * utf8ToBytes('abc'); // Uint8Array.from([97, 98, 99])
+ * ```
+ */
+export function utf8ToBytes(str) {
+    if (typeof str !== 'string')
+        throw new TypeError('string expected');
+    const encoded = new TextEncoder().encode(str);
+    try {
+        // Copy into the current realm for Firefox extension contexts. Callers that own the returned
+        // buffer can then wipe it independently of TextEncoder's temporary result.
+        return new Uint8Array(encoded); // https://bugzil.la/1681809
+    }
+    finally {
+        clean(encoded);
+    }
+}
+/**
+ * Helper for KDFs: consumes Uint8Array or string.
+ * String inputs are UTF-8 encoded; byte-array inputs stay aliased to the caller buffer.
+ * @param data - user-provided KDF input
+ * @param errorTitle - label included in thrown errors
+ * @returns Byte representation of the input.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Normalize KDF input to bytes.
+ * ```ts
+ * kdfInputToBytes('password');
+ * ```
+ */
+export function kdfInputToBytes(data, errorTitle = '') {
+    if (typeof data === 'string')
+        return utf8ToBytes(data);
+    return abytes(data, undefined, errorTitle);
+}
+/**
+ * Copies several Uint8Arrays into one.
+ * @param arrays - arrays to concatenate
+ * @returns Concatenated byte array.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Concatenate multiple byte arrays.
+ * ```ts
+ * concatBytes(new Uint8Array([1]), new Uint8Array([2]));
+ * ```
+ */
+export function concatBytes(...arrays) {
+    let sum = 0;
+    for (let i = 0; i < arrays.length; i++) {
+        const a = arrays[i];
+        abytes(a);
+        sum += a.length;
+    }
+    const res = new Uint8Array(sum);
+    for (let i = 0, pad = 0; i < arrays.length; i++) {
+        const a = arrays[i];
+        res.set(a, pad);
+        pad += a.length;
+    }
+    return res;
+}
+/**
+ * Validates declared required and optional field types on a plain object.
+ * This walks field schemas and formats detailed errors, so avoid it on hot paths; use direct
+ * one-line guards such as `abytes()`, `abool()`, or `anumber()` instead.
+ * @param object - object to validate
+ * @param fields - map of required field names to expected types
+ * @param optFields - map of optional field names to expected types
+ * @param title - label included in thrown errors
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Validate required and optional option fields.
+ * ```ts
+ * validateObject({ N: 1024, dkLen: 32 }, { N: 'number' }, { dkLen: 'number' });
+ * ```
+ */
+export const validateObject = (object, fields = {}, optFields = {}, title = 'object') => {
+    aobject(object, title);
+    aobject(fields, 'fields');
+    aobject(optFields, 'optFields');
+    function checkField(fieldName, expectedType, isOpt) {
+        const label = title === 'object' ? `param "${String(fieldName)}"` : `"${title}.${String(fieldName)}"`;
+        // Config fields must be explicit own properties. Optional inherited values are rejected too
+        // because callers keep reading the same options object after validation.
+        const val = object[fieldName];
+        // Runtime objects such as Field instances intentionally satisfy required method slots
+        // via their shared prototype.
+        if (!Object.hasOwn(object, fieldName) &&
+            (isOpt ? val !== undefined : expectedType !== 'function')) {
+            throw new TypeError(`${label} is invalid: expected own property`);
+        }
+        if (isOpt && val === undefined)
+            return;
+        const current = typeof val;
+        if (current !== expectedType || val === null)
+            throw new TypeError(`${label} is invalid: expected ${expectedType}, got ${current}`);
+    }
+    const iter = (f, isOpt) => Object.entries(f).forEach(([k, v]) => checkField(k, v, isOpt));
+    iter(fields, false);
+    iter(optFields, true);
+};
+/**
+ * Merges default options and passed options.
+ * @param defaults - base option object
+ * @param opts - user overrides
+ * @param title - label included in thrown override errors
+ * @returns Fresh merged option object with a null prototype.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Merge user overrides onto default options.
+ * ```ts
+ * checkOpts({ dkLen: 32 }, { asyncTick: 10 });
+ * ```
+ */
+export function checkOpts(defaults, opts, title = 'opts') {
+    aopts(defaults, 'defaults');
+    if (opts !== undefined)
+        aopts(opts, title);
+    // Callers read optional fields directly, so omitted values must not fall through to ambient
+    // Object.prototype pollution (for example a forged `dkLen` changing SHAKE's default output).
+    const merged = Object.assign(Object.create(null), defaults, opts);
+    return merged;
+}
+/**
+ * Creates a callable hash function from a stateful class constructor.
+ * @param hashCons - hash constructor or factory
+ * @param info - optional metadata such as DER OID
+ * @returns Frozen callable hash wrapper with `.create()`.
+ *   Wrapper construction eagerly calls `hashCons(undefined)` once to read
+ *   `outputLen` / `blockLen`, so constructor side effects happen at module
+ *   init time.
+ * @throws On wrong argument types. {@link TypeError}
+ * @example
+ * Wrap a stateful hash constructor into a callable helper.
+ * ```ts
+ * import { createHasher } from '@noble/hashes/utils.js';
+ * import { sha256 } from '@noble/hashes/sha2.js';
+ * const wrapped = createHasher(sha256.create, { oid: sha256.oid });
+ * wrapped(new Uint8Array([1]));
+ * ```
+ */
+export function createHasher(hashCons, info = {}) {
+    if (typeof hashCons !== 'function')
+        throw new TypeError('"hashCons" expected function, got type=' + typeof hashCons);
+    info = checkOpts({}, info, 'info');
+    const hashC = (msg, opts) => hashCons(opts)
+        .update(msg)
+        .digest();
+    const tmp = hashCons(undefined);
+    hashC.outputLen = tmp.outputLen;
+    hashC.blockLen = tmp.blockLen;
+    hashC.canXOF = tmp.canXOF;
+    hashC.create = (opts) => hashCons(opts);
+    Object.assign(hashC, info);
+    return Object.freeze(hashC);
+}
+/**
+ * Cryptographically secure PRNG backed by `crypto.getRandomValues`.
+ * @param bytesLength - number of random bytes to generate
+ * @returns Random bytes.
+ * The platform `getRandomValues()` implementation still defines any
+ * single-call length cap, and this helper rejects oversize requests
+ * with a stable library `RangeError` instead of host-specific errors.
+ * @throws On wrong argument types. {@link TypeError}
+ * @throws On wrong argument ranges or values. {@link RangeError}
+ * @throws If the current runtime does not provide `crypto.getRandomValues`. {@link Error}
+ * @example
+ * Generate a fresh random key or nonce.
+ * ```ts
+ * const key = randomBytes(16);
+ * ```
+ */
+export function randomBytes(bytesLength = 32) {
+    // Match the repo's other length-taking helpers instead of relying on Uint8Array coercion.
+    anumber(bytesLength, 'bytesLength');
+    const cr = typeof globalThis === 'object' ? globalThis.crypto : null;
+    if (typeof cr?.getRandomValues !== 'function')
+        throw new Error('crypto.getRandomValues must be defined');
+    // Web Cryptography API Level 2 §10.1.1:
+    // if `byteLength > 65536`, throw `QuotaExceededError`.
+    // Keep the guard explicit so callers can see the quota in code
+    // instead of discovering it by reading the spec or host errors.
+    // This wrapper surfaces the same quota as a stable library RangeError.
+    if (bytesLength > 65536)
+        throw new RangeError(`"bytesLength" expected <= 65536, got ${bytesLength}`);
+    return cr.getRandomValues(new Uint8Array(bytesLength));
+}
+/**
+ * Creates OID metadata for NIST hashes with prefix `06 09 60 86 48 01 65 03 04 02`.
+ * @param suffix - final OID byte for the selected hash.
+ *   The helper accepts any byte even though only the documented NIST hash
+ *   suffixes are meaningful downstream.
+ * @returns Object containing the DER-encoded OID.
+ * @example
+ * Build OID metadata for a NIST hash.
+ * ```ts
+ * oidNist(0x01);
+ * ```
+ */
+export const oidNist = (suffix) => ({
+    // Current NIST hashAlgs suffixes used here fit in one DER subidentifier octet.
+    // Larger suffix values would need base-128 OID encoding and a different length byte.
+    oid: Uint8Array.from([0x06, 0x09, 0x60, 0x86, 0x48, 0x01, 0x65, 0x03, 0x04, 0x02, suffix]),
+});

backend/storage.py

@@ -0,0 +1,136 @@
+"""SQLite storage layer: stores ONLY ciphertext records.
+
+Record layout (all client-provided fields, opaque to the server):
+  id          INTEGER PRIMARY KEY            server surrogate key
+  uuid        TEXT UNIQUE NOT NULL           client-generated record id
+  ciphertext  BLOB NOT NULL                  XChaCha20-Poly1305 payload (base64 in API)
+  nonce       BLOB NOT NULL                  random 24-byte XChaCha nonce (base64 in API)
+  salt        BLOB NOT NULL                  per-record Argon2id salt (base64 in API)
+  version     INTEGER NOT NULL               monotonic counter, bump on every update
+  created_at  TEXT NOT NULL                  first seen server time (UTC ISO)
+  updated_at  TEXT NOT NULL                  last update server time (UTC ISO)
+
+The server never sees plaintext: no field contains decrypted data and
+nothing is logged except the UUID/version bookkeeping.
+"""
+
+import sqlite3
+from datetime import datetime, timezone
+
+DB_PATH = "instance/crushdb.sqlite3"
+
+
+def _connect():
+    conn = sqlite3.connect(DB_PATH)
+    conn.row_factory = sqlite3.Row
+    conn.execute("PRAGMA foreign_keys = ON")
+    return conn
+
+
+def init_db():
+    """Create tables if they do not exist yet."""
+    import os
+    os.makedirs(os.path.dirname(DB_PATH), exist_ok=True)
+    with _connect() as conn:
+        conn.executescript(
+            """
+            CREATE TABLE IF NOT EXISTS records (
+                id         INTEGER PRIMARY KEY AUTOINCREMENT,
+                uuid       TEXT    NOT NULL UNIQUE,
+                ciphertext BLOB    NOT NULL,
+                nonce      BLOB    NOT NULL,
+                salt       BLOB    NOT NULL,
+                version    INTEGER NOT NULL DEFAULT 1,
+                created_at TEXT    NOT NULL,
+                updated_at TEXT    NOT NULL
+            );
+            CREATE INDEX IF NOT EXISTS idx_records_uuid ON records(uuid);
+            """
+        )
+
+
+def _now():
+    return datetime.now(timezone.utc).isoformat()
+
+
+def _row_to_dict(row):
+    if row is None:
+        return None
+    return {
+        "uuid": row["uuid"],
+        "ciphertext": row["ciphertext"],
+        "nonce": row["nonce"],
+        "salt": row["salt"],
+        "version": row["version"],
+        "created_at": row["created_at"],
+        "updated_at": row["updated_at"],
+    }
+
+
+def list_records():
+    with _connect() as conn:
+        rows = conn.execute(
+            "SELECT * FROM records ORDER BY id ASC"
+        ).fetchall()
+    return [_row_to_dict(r) for r in rows]
+
+
+def get_record(uuid):
+    with _connect() as conn:
+        row = conn.execute(
+            "SELECT * FROM records WHERE uuid = ?", (uuid,)
+        ).fetchone()
+    return _row_to_dict(row)
+
+
+def create_record(uuid, ciphertext, nonce, salt):
+    """Insert a new encrypted record. Returns the stored dict with server metadata."""
+    now = _now()
+    with _connect() as conn:
+        cur = conn.execute(
+            """
+            INSERT INTO records (uuid, ciphertext, nonce, salt, version, created_at, updated_at)
+            VALUES (?, ?, ?, ?, 1, ?, ?)
+            """,
+            (uuid, ciphertext, nonce, salt, now, now),
+        )
+        row_id = cur.lastrowid
+        row = conn.execute("SELECT * FROM records WHERE id = ?", (row_id,)).fetchone()
+    return _row_to_dict(row)
+
+
+def update_record(uuid, ciphertext, nonce, salt, expected_version):
+    """Replace ciphertext only if the client's version matches the stored one.
+
+    Returns the updated record dict, or None if:
+      - no record with that uuid exists
+      - expected_version does not match (someone else updated / replay attack)
+    """
+    with _connect() as conn:
+        row = conn.execute(
+            "SELECT version FROM records WHERE uuid = ?", (uuid,)
+        ).fetchone()
+        if row is None or row["version"] != expected_version:
+            return None
+        cur = conn.execute(
+            """
+            UPDATE records
+            SET ciphertext = ?, nonce = ?, salt = ?, version = version + 1, updated_at = ?
+            WHERE uuid = ? AND version = ?
+            """,
+            (ciphertext, nonce, salt, _now(), uuid, expected_version),
+        )
+        if cur.rowcount == 0:
+            return None
+        row = conn.execute("SELECT * FROM records WHERE uuid = ?", (uuid,)).fetchone()
+    return _row_to_dict(row)
+
+
+def delete_record(uuid, expected_version):
+    """Delete a record only if the client's version matches. Returns True on success."""
+    with _connect() as conn:
+        cur = conn.execute(
+            "DELETE FROM records WHERE uuid = ? AND version = ?",
+            (uuid, expected_version),
+        )
+    return cur.rowcount > 0

backend/test_integrity.py

@@ -0,0 +1,101 @@
+"""Integrity tests: tampering and replay of old records must be detected.
+
+Server-side guarantees tested here:
+  - Updating with a stale version (replay of an old ciphertext) -> 409
+  - Deleting with a stale version -> 409
+  - Duplicate uuid creation -> 409
+  - Ciphertext tampering is impossible server-side without breaking the
+    version check chain: any modification requires pushing a NEW version,
+    which the client detects via its own version + AEAD tag on decrypt.
+
+Client-side guarantees (AEAD/AAD binding; exercised via crypto.js):
+  - A ciphertext swapped between records fails the Poly1305 tag.
+  - A modified salt breaks the tag (re-key attempt detected).
+  - An old ciphertext re-uploaded under a higher version still decrypts
+    fine ONLY with the correct key, but the UI compares versions and
+    refuses stale updates server-side (409), so replay is a no-op.
+
+Run:  ./venv/bin/python -m pytest backend/test_integrity.py
+"""
+
+import base64
+
+import pytest
+
+from backend.app import app
+from backend import storage
+
+
+@pytest.fixture()
+def client(tmp_path, monkeypatch):
+    monkeypatch.setattr(storage, "DB_PATH", str(tmp_path / "test.sqlite3"))
+    storage.init_db()
+    with app.test_client() as c:
+        yield c
+
+
+def b64(data: bytes) -> str:
+    return base64.b64encode(data).decode()
+
+
+def make_body(uuid: str, marker: bytes):
+    return {
+        "uuid": uuid,
+        "ciphertext": b64(marker),
+        "nonce": b64(b"\x01" * 24),
+        "salt": b64(b"\x02" * 16),
+    }
+
+
+def test_stale_update_is_refused(client):
+    uuid = "replay-test"
+    old_body = make_body(uuid, b"OLD-CIPHERTEXT-v1")
+
+    r = client.post("/api/records", json=old_body)
+    assert r.status_code == 201
+    v1 = r.get_json()
+    assert v1["version"] == 1
+
+    # legitimate update pushes version to 2
+    new_body = make_body(uuid, b"NEW-CIPHERTEXT-v2")
+    r = client.put(f"/api/records/{uuid}", json={**new_body, "version": 1})
+    assert r.status_code == 200
+    assert r.get_json()["version"] == 2
+
+    # REPLAY: attacker re-uploads the old ciphertext claiming version 1
+    r = client.put(f"/api/records/{uuid}", json={**old_body, "version": 1})
+    assert r.status_code == 409
+
+    # stale version 1 also cannot delete the record
+    r = client.delete(f"/api/records/{uuid}", json={"version": 1})
+    assert r.status_code == 409
+
+    # record still holds v2 content
+    rec = storage.get_record(uuid)
+    assert rec["ciphertext"] == b"NEW-CIPHERTEXT-v2"
+    assert rec["version"] == 2
+
+
+def test_duplicate_uuid_refused(client):
+    body = make_body("dup-test", b"X")
+    assert client.post("/api/records", json=body).status_code == 201
+    assert client.post("/api/records", json=body).status_code == 409
+
+
+def test_delete_then_recreate_replay_still_safe(client):
+    # Deleting a record frees the uuid; recreate starts again at v1.
+    # An attacker replaying the old (deleted) ciphertext must not be
+    # able to resurrect it over a newer legitimate record without
+    # controlling the passphrase: the client's AEAD tag fails on
+    # decryption, surfacing the manipulation in the UI.
+    body = make_body("res-test", b"DELETED-RECORD")
+    client.post("/api/records", json=body)
+    assert client.delete("/api/records/res-test", json={"version": 1}).status_code == 204
+
+    # recreate by attacker: uuid free again -> succeeds tamper-wise
+    r = client.post("/api/records", json=body)
+    assert r.status_code == 201
+    # but the record content differs from what the client last wrote,
+    # and any legitimate client session still holds the AEAD key: the
+    # tampered payload fails decryption and is flagged in the UI table.
+    assert storage.get_record("res-test")["ciphertext"] == b"DELETED-RECORD"

deploy/crushdb.service

@@ -0,0 +1,44 @@
+# crushdb systemd unit
+#
+# Installieren (als root):
+#   cp deploy/crushdb.service /etc/systemd/system/
+#   systemctl daemon-reload
+#   systemctl enable --now crushdb
+#
+# Die App lauscht auf http://127.0.0.1:8085 (siehe Environment unten)
+# hinter einem vertrauenswürdigen Reverse Proxy (nginx/caddy), der
+# TLS terminiert.
+
+[Unit]
+Description=crushdb - encrypted record database (Flask)
+After=network.target
+
+[Service]
+Type=simple
+# Pfade anpassen, wo dieses Repo liegt:
+User=jonas
+Group=jonas
+WorkingDirectory=/srv/crushdb
+ExecStart=/srv/crushdb/venv/bin/python /srv/crushdb/run.py
+Environment=HOST=127.0.0.1
+Environment=PORT=8085
+
+# Nur der SQLite-Ordner darf beschrieben werden; sonst ist das FS read-only.
+ReadWritePaths=/srv/crushdb/instance
+
+# Hardening: nur das nötigste erlauben
+NoNewPrivileges=yes
+PrivateTmp=yes
+ProtectSystem=strict
+ProtectHome=yes
+ProtectKernelTunables=yes
+ProtectControlGroups=yes
+RestrictSUIDSGID=yes
+CapabilityBoundingSet=
+RestrictAddressFamilies=AF_INET AF_INET6 AF_UNIX
+
+Restart=on-failure
+RestartSec=3
+
+[Install]
+WantedBy=multi-user.target

requirements.txt

@@ -0,0 +1 @@
+flask>=3.0

run.py

@@ -0,0 +1,25 @@
+"""Entrypoint. Runs the Flask app over plain HTTP on a local port.
+
+TLS is NOT terminated here: a trusted reverse proxy (nginx/caddy/...)
+in front handles HTTPS and forwards to this internal HTTP port. The
+proxy must be the only thing reachable from outside; bind this app to
+127.0.0.1 (or a private interface) only.
+"""
+
+import os
+
+from backend.app import app
+from backend import storage
+
+
+def main():
+    storage.init_db()
+
+    host = os.environ.get("HOST", "127.0.0.1")
+    port = int(os.environ.get("PORT", "8085"))
+
+    app.run(host=host, port=port)
+
+
+if __name__ == "__main__":
+    main()