Ships 2437506 (the restore setting was a no-op with the default quit
clear), the first-use loader for extensions with Aegis 0.29.0 bundled
(~0.7 s less main-thread work at launch, ETH/SOL bridge fixed), the
language picker as an overlay, Pithos 0.3.4 and the Ariadne status row.
"Open previous windows and tabs" and "Clear history on quit" both default
to on, and the quit clear deleted session.json along with the history, so
every launch started from the start page and the restore setting did
nothing. The open tabs are what the user asked to reopen, not history:
while restore is on, the quit clear keeps them and still drops back/forward
and address-bar history. With restore off, the tab list is deleted as
before.
WalletKeys.entry() built a bech32 p2wpkh address whatever the account
path's purpose, so picking Legacy (D...), Wrapped SegWit (S...) or
Taproot (dgb1p...) in Settings showed a dgb1q address from the BIP44/
49/86 key tree, one no other wallet restoring that path would find.
Each family now derives its own address and script, and spending
supplies what its inputs need (previous tx for P2PKH, redeem script for
P2SH-P2WPKH, tap-tweaked key for Taproot, which is active on DigiByte),
with per-family fee sizes. Unknown purposes are refused instead of
falling back to BIP84.
Also: inputs were signed in selection order but the PSBT orders them by
BIP69, so a spend from two addresses tried to sign each input with the
other's key. They are now signed in the PSBT's order.
approvalRequest passed approve/reject keys the host overlay does not
know, so it showed a lone "OK" button whose id never equalled
"approve": every WizardConnect signing request was refused, and the
HTML body was shown as literal markup. It now passes a Sign action and
plain rows: wallet, input count, each output decoded to a cashaddr on
the wallet's network (or OP_RETURN / raw script), total, and who
broadcasts.
lib/dgb/deps.js is ESM, but in a dev checkout the nearest package.json
is TheseusNavigator's, which says "type": "commonjs"; the import threw
and DigiByte was silently unavailable whenever Theseus ran from the
repo. Shipped installs have no package.json above the add-on, so they
were unaffected. lib/dgb/core and lib/dgb/psbt already carry the same
marker.
The isolated-world relay accepted any postMessage carrying the fixed
tag "aegis-aegis", including one from a cross-origin iframe (an ad,
an embed), and attributed it to the top-level origin and its grants.
The tag now carries a per-load random nonce that only the injected
main-world bridge knows, and both listeners drop events whose source
is not this window. The document_start install path is unchanged.
The overlay for tronWeb-built transactions was built from the dapp's
raw_data JSON, which need not match raw_data_hex: a site could show
"1 TRX to X" and get a signature over anything. lib/tron-decode.js
decodes Transaction.raw from raw_data_hex (contract type, owner, to,
amount, TRC-20 transfer/approve calldata, fee limit, memo); the txID
must match the bytes, every contract's owner must be this wallet, and
unlimited approvals or permission/resource delegation get a danger
action.
sendRawTransaction relayed any signed transaction a site handed it;
it now broadcasts only txids Aegis itself signed.
- signAndSendTransaction signed String(Uint8Array) ("1,2,3,..."), so
every dapp transaction got an invalid signature. Adapters gain
signBytes(), which signs the exact message bytes.
- v0 (VersionedTransaction) messages were parsed with the version byte
as the header; the shared parser handles legacy and v0.
- window.solana.signTransaction went through signMessage and its
"moves no SOL" overlay. It now has its own handler and overlay, and
signMessage refuses bytes that parse as a transaction (Phantom's rule),
since such a signature is a valid transaction signature.
- Overlays decode System transfers and SPL transfer / approve /
set-authority, and list everything else as not decoded.
- eth_sendTransaction dropped the calldata, gas and fee fields, so an
ERC-20 transfer went out as a 0-value send to the token contract and
any contract call was broadcast as something else. plan() now carries
data/gas/fee caps/nonce, estimates gas for calls, and the overlay
decodes transfer/approve/permit/setApprovalForAll, flags unlimited
approvals and calldata to a non-contract, and shows estimated vs max
fee.
- personal_sign signed the hex string ethers/viem send as literal text;
it now signs the decoded bytes.
- wallet_switchEthereumChain flipped the global selected wallet, so any
site could move every connected dapp to another chain. Chain is now
per origin; the sidebar selection no longer redirects a dapp.
- wallet_addEthereumChain silently persisted a connection for chains
Aegis already had, handing the address to any site. Adding a chain
no longer grants anything, and RPC URLs must be https.
- eth_sign (blind hash signing) is disabled, as in MetaMask.
Connecting without ticking "Always allow" stored nothing, so the
site's very next call (personal_sign, signTransaction, ...) failed with
"not connected". Plain Connect now grants the origin in memory until
Theseus restarts; "Always allow" still persists. Every bridge (BCH,
Tron, EVM, Solana) checks grants the same way, revoke clears both, and
the connected-sites list shows EVM/Solana grants and which ones are
session-only.
The 2026-09-13 audit of the dapp bridges found real signing bugs whose
fixes were never committed; 0.30.0 carries them, ported onto the current
add-on.
They had a row of their own with nothing else in it, which read as an
empty card with a stray button. Same layout as the Plug-ins list row now:
buttons beside the on/off switch.
The globe-chip menu was a native Electron menu — square corners, system
font, no theming beyond the OS's own context-menu paint. Replaces it with
a floating overlay WebContentsView (lang-picker.html + preload),
following the same pattern the engine picker and the popover already
use: rounded 12px surface, acid-tint accents on the current pin, soft
shadow, dark + light scheme, flush under the chip's bottom-right.
The content is organised around the user's intent — translate first,
pick a language second. The "Translate this page" row sits at the top
when it is actionable (web tab + supported source + supported target),
with the detected source and the target under the label so the user can
tell what the backend will do before they click. Once a page is
translated, that row flips to "Show original (<source>)". Below the
action strip is Automatic + the six supported languages, each with a
two-letter code chip in the left slot and a ✓ on the current pin.
Unsupported languages are hidden by default inside a collapsible "More
languages (translator coming later)" group — click to expand, click to
collapse; a pinned-unsupported auto-expands so its ✓ stays visible.
Plumbing matches engine-picker: deferred-load WebContentsView,
closeOnClickAway, setBounds anchored under the chip, picker renderer
reports its own content height after each render so the overlay
contracts and expands with the "More languages" toggle. State pushes
come from emitTranslateState (so the Translate / Show-original row
updates when a page finishes auto-translating with the picker open)
and from broadcastSettings (so a Settings-side language change
re-paints the ✓).
Verified end-to-end: picker loads, shows Automatic + 6 supported in the
default list and 18 greyed in the "More" group, repaints to "⟲ Show
original (Spanish)" after an auto-translate completes.
The chip menu showed every entry in WEBSITE_LANGUAGE_QUICK (24 rows, most
greyed with "— translator coming later") and the picker read as a wall
of coming-soon noise. The top strip now carries only the languages the
translator actually handles — Automatic plus the six supported ones
(en, es, fr, de, el, ru) — and everything else moves into a "More
languages (translator coming later)" submenu where the greyed rows live
without crowding the main menu. If the user's current pin is one of the
unsupported ones, it stays visible at the top so the ✓ reads at a
glance, not two levels deep.
Translation didn't follow a language change reliably:
- A page with no `<html lang>` left pageLang empty, which the auto-
translate hook took as "no translation needed" and skipped the entire
page. Now an empty source is still translated (the backend auto-detects
the real language), and the hook only skips when pageLang is known AND
matches the user's target.
- Changing the pin via Settings or the chip reloaded the active tab but
did not re-translate — the hook needs auto-offer on, and even then
a `pending` latch set by setWebsiteLanguage was wiped by the reload's
did-start-navigation reset. The `pending` bit now survives that reset,
so the did-finish-load hook translates unconditionally for an explicit
language switch (the user ASKING for a new language IS the request to
translate the current page too). A translated page is reverted before
the reload so the fresh HTML lands on original DOM, not a mix of old
translated nodes and new content. Verified end-to-end: an es→en auto-
translate followed by a pin to French takes the page to French in one
shot without the chip being touched.
Aegis was most of what was left of launch cost: its crypto and
WizardConnect deps block the main thread for ~0.6 s right after the first
frame. It now declares its panel and "activation": "on-demand"; the
dapp bridges are still on every page from the start, and the first page
call, panel open or wiz:// link starts it.
activate() returns a promise the host waits on, so the call that woke
Aegis finds WizardConnect and the mounted wallets. With a locked vault it
does not wait for the mount (derive blocks until unlock), so the page gets
the usual locked answer in ~0.7 s instead of after the host's 5 s limit.
Two things only work while Aegis runs: a live WizardConnect pairing (no one
else listens on its relays) and "stay unlocked" (which also opens
Settings > Passwords). While either is on, Aegis asks the host to start it
at launch, and withdraws the request when both are off. Pairings left
behind by a removed wallet do not count.
Boot trace, same profile, 3 warm runs: main thread blocked in the first
6 s 970-1040 ms -> 300-320 ms, longest block 605-667 ms -> 227-244 ms,
toolbar 1.34-1.61 s -> 0.83-0.87 s.
The main-world bridge is appended at document_start, which often runs
before the page has an <html> element. The append threw on null, and the
catch marked the origin as Trusted-Types-blocked in localStorage, so every
later visit skipped window.ethereum, window.solana and the EIP-6963
announcement on that site. On example.com it failed on 6 of 6 loads.
Wait for <html> with a MutationObserver (it fires at the microtask
checkpoint before the first parser-inserted script, so the bridge is still
first), remember only real Trusted Types refusals, and use a new key so the
origins wrongly marked by the old one get the bridge back.
The left-edge panels landed meanwhile, so a panel record now carries both
its side and replace-by-id; a manifest-declared panel may say side:left too,
or a dormant add-on would show its panel on the wrong edge until it starts.
Ships 08b101f (a quick-unlock PIN for the vault, shared with extensions
through api.vault.requestUnlock), 3fd0fe3 (extensions can open from the
left edge, beside the quick links), Pithos 0.3.0 bundled (c2df6e3: guided
setup, PIN gate, recovery phrase from the vault, on the left) and e1c5d91
(one language chip with "Translate this page" and automatic translation).
Extensions could only put panels in the right sidebar. An add-on can now
register a panel with side: "left": its icon joins the quick-links strip
(above the web-app links), and it opens in the strip's panel slot in its
own view with the sidebar preload, so the add-on bridge, events and the
widen/narrow controls work as on the right. A left panel and a quick-link
web app never share the slot; reopening keeps the panel's page and state.
Left panels drop out of the right sidebar and its dock. With the strip
turned off they fall back to the right sidebar so they stay reachable.
Pithos is the first: bundled copy rebuilt with side: "left".
The vault re-locks on every restart and only the master password opened
it, so every extension that needs it (Aegis, now Pithos) either asked for
the master password itself or grew its own PIN. Theseus now owns one:
- Settings > Passwords sets, changes or removes a 6-digit PIN. The PIN
wraps the master password (PBKDF2-SHA256, 600k iterations, AES-256-GCM)
and the result is sealed with the OS keystore (safeStorage: DPAPI /
Keychain / libsecret), so a copied vault-pin.json cannot be brute-forced
elsewhere. Every unlock still ends at the master password.
- Three wrong PINs in a row require the master password. The strike count
lives in the same file, so a restart does not reset it; a successful
master-password unlock does. A PIN whose password no longer opens the
vault (password changed) is dropped.
- unlock.html is Theseus's own prompt, over the whole window: PIN pad, or
the master password. Extensions call api.vault.requestUnlock({ reason })
(vault-derive capability) and get { ok } back; what the user typed never
reaches them. Settings' locked screen offers "Unlock with PIN" through
the same prompt.
Two chips carried the same word in two shapes — a globe (Accept-Language)
and a translate chip (chip lights when page lang differs) — both labelled
"RU" at the same time for a Russian user. The chip for translation is
gone. The globe menu now covers both: a "Translate this page from X to Y"
item appears at the top when the loaded page is in another supported
language, flipping to "Show original" while a translation is on screen.
The chip's own code still shows the user's language (EN, RU, …); its
tooltip switches to "Translated to <X>. Menu: Show original." when a
translation is up, so the one chip reads the whole state.
With "Translate automatically" on, Theseus translates in place on
did-finish-load the first time it sees a supported source + target
mismatch for the active tab — no chip-click needed. A `_tr.autoTried`
latch keeps it to one attempt per document (a failing backend doesn't
retry on every reflow), and the latch resets on did-start-navigation so
the next page gets a fresh shot. The setting copy in Settings › Language
now says "Translate automatically" instead of "Offer to translate", so
the switch's label matches the behaviour.
The picker (both in Settings and in the globe menu) still lists every
language in WEBSITE_LANGUAGE_QUICK, but entries whose base code isn't
on the translator backend (en, es, fr, de, el, ru today) are shown
greyed out with "— translator coming later", and "Other… (Accept-Language
only, no translation)" is explicit about what free-form tags buy you.
The menu is a roadmap, not a lie: a user picking one of the greyed
entries sets Accept-Language and nothing else surprises them.
Rebuilt from Pithos/ with scripts/build-theseus-addon.mjs. Pithos moves from a toolbar menu that opened a loopback tab to a left-sidebar panel. Existing installs get the same version over the extension update channel.
0.3.72 shipped without the Settings › General › Updates copy update:
"Theseus already checks the release manifest at boot and every 6h" is
accurate for a successful first check, but silent about the retry
backoff (8s, 30s, 2min, 10min, 30min) that fires when the startup
attempt is offline — on a slow or captive-portal connection the user
would see several checks in the first 43 minutes and the copy made
that look like a bug. The new wording owns the retry.
Ships 70b7325 (Pithos, the s3d control panel, as a bundled extension:
open it from the dock to run your own S3 gateway on Sia), dae6b9a
(Settings gets its own Language page), f3efae3 (translator served from
silentmode.st/libre with libre.x / lingua.x) and 604a990 (BNS names read
from Ariadne's indexer when it is installed, Theseus's own as standby).
Website language, offer-to-translate and the translator peers list used
to sit as three sub-sections inside General, and Privacy › Anti-
fingerprinting duplicated the language picker on top of them — three
places to edit the one languageMode/languageValue setting. Settings
grows its own Language entry in the sidebar now; everything about
language — the preferred-language picker, the auto-offer toggle, the
peer list, the API key — lives there, and the Privacy duplicate is
gone. The chip in the URL bar still edits the same setting, so the
toolbar surface is unchanged.
The copy for the picker ("Your language") now says what the setting
actually drives: it's sent as Accept-Language and it's the translator's
target. Legacy "hide"/"spoof" language modes migrate to Automatic on
first open of Settings (the picker only has Automatic / a tag / Other),
so a profile that still carries one of those from an older release
lands on a valid state the first time Settings opens.
Greek (el-GR) was already in the picker; this release's silentmode.st/
libre backend added `el` to --load-only so the translator now has real
en↔el support — the picker and the backend are in step.
Ships Pithos (the s3d control panel) as a built-in extension: the dock
menu opens it in a tab served from a loopback port, and "Stop s3d"
shuts the gateway down. Generated by Pithos/scripts/build-theseus-addon.mjs.
yaml is vendored under vendor/yaml/lib rather than node_modules/ or
dist/, because Theseus git-ignores both and a clean-worktree release
build would otherwise ship the add-on without its only dependency.
Migration step 3 (DESIGN-bns-indexer-service.md). Ariadne's Thread now owns
BNS indexing on the machine, so Theseus no longer runs a second electrum
indexer beside it.
bns-indexer.js keeps its process and its messages to main.js, but inside
it is now an index host on the shared source chain:
- Ariadne's indexer over its pipe, trusted only after ariadne-helper.exe
has checked the server process on that connection (found through
Ariadne's uninstall key), then pushes;
- the local copies: Ariadne's files for both scopes, Theseus's own raw
copy, the bundled one. The richest wins.
- Theseus's own index copy, written from the pipe data.
The shared core runs as Theseus's own indexer only while Ariadne is
unhealthy. That means: no pipe 4 s after launch, a pipe that fails the
check, a pipe that went silent, or an index not confirmed for 10 min while
Ariadne is not paused. The own indexer warm-starts from Ariadne's
snapshot, so there is no download and no cold sync. It hands back after
90 s of health, so a flapping service does not start and stop it. Economy
is not a failure and never triggers a takeover. With no checkable Ariadne
(portable, not installed, older than the pipe) the own indexer starts at
once, as in 0.3.70. The one thing Theseus does on Ariadne's side is run
the indexer's task at launch when "Launch at start" is off.
main.js passes the shared module paths (packaged as .mjs, which is why the
shared modules no longer import each other) and keeps the host's status.
The Ariadne panel takes its state from the indexer task when one exists,
and the sub-page says where Theseus's names come from.
panel.html loaded lib/jsqr.js synchronously: 257 KB and 10,105 lines of
vendored decoder parsed on every panel open, so on every hide/show, for a
feature most sessions never touch. It is now fetched the first time someone
imports a QR image, with concurrent callers sharing one load and a failed
load retryable rather than cached as a permanent rejection.
qr.js and panel.js get `defer`, so the browser can paint the (already dark)
document before executing 300 KB of panel.js. Order is preserved, so qr.js
is still defined before panel.js runs.
Parsed on open: 650 KB -> 399 KB.
This shortens the blank window but does not remove it. The white box itself
is Theseus-side: the sidebar panel's view is built as
new WebContentsView({ webPreferences: { preload: "sidebar-preload.js" } })
with no backgroundColor, and Electron defaults a view's background to white —
so white shows from view-attach until the page paints. Every other view in
the app passes a colour, and registerSidebarPanel takes only
{id, title, icon, page}, so an add-on cannot declare one. Not fixed here:
main.js belongs to the Theseus work in flight.
The translator client now ships with the right defaults for the actual
deployment: silentmode.st/libre (ICANN, via the main cert and no new
subdomain) is the primary peer; libre.x and lingua.x are registered on
BNS with `p` records that reverse-proxy back to the same backend; the
public LibreTranslate.com key-gated tier stays as the last-resort entry.
Two wiring fixes make the BNS fallback actually usable from Theseus:
1. translatorPostOnce rewrites the request URL through targetUrlFor
before fetching, so a peer whose host is a BNS name (libre.x) is
dispatched via the in-process bns:// handler — Chromium's net stack
has no way to resolve `.x` by itself.
2. serveBns's p-record branch now forwards the method, headers and body
of the original request to the upstream, not just a GET. Without
that, a POST /translate against libre.x arrived at the backend as
a GET with no body and 400'd — now the proxy is actually a reverse-
proxy, as the record type's name promises.
Verified end-to-end against the live silentmode.st/libre instance from a
fresh Theseus profile with a Spanish test page: both the direct
silentmode.st/libre peer and the libre.x -> bns:// -> serveP -> upstream
path translate the page and the revert path restores the originals.
Records where the Ariadne 0.2.0 implementation differs from the design
and why: the protected index\ subfolder, the helper-relayed pipe check,
ariadne-run.exe as the restarter because Task Scheduler does not restart
a program that exits with an error, the resolver exiting in Theseus-only
mode, no owners on the HTTP API, and setup's own scope page. It also
records what the Theseus client needs from the pipe. Economy produces no
"fresh" pushes, so a missing heartbeat while paused must not trigger a
takeover. Last, the plan for bundling Ariadne's setup into Theseus's
installer, not yet wired into the build.
Second half of the Theseus-lag investigation. 0.27.1 removed the 7.4 MB
store; this removes the two things that produced it and the latency that
came with it.
Measured on this profile's own cache: 6,981 transactions, 7.38 MB, and one
consolidation with 400 inputs (median 2).
- loadHistory() awaited getTx() per displayed transaction and then again per
INPUT, strictly one at a time. The 400-input row alone cost 400 serial
round trips. Both waves are now prefetched with bounded parallelism
(getTxMany, 12 at a time). Against a simulated 10 ms link the same 473
fetches take 771 ms instead of ~4,730 ms; on a real 30-50 ms link the
serial version was 15-25 seconds per refresh, per wallet.
- Parent transactions were persisted forever. They exist only to compute a
delta for the 25 rows on screen, and keeping every one ever seen is what
grew the file. Only the displayed window is written now — 25 entries,
16.4 KB in the harness — while parents stay in a process-lifetime map
bounded at 20,000, seeded from the window so a restart does not refetch
what is already visible. A second refresh issues zero fetches.
TX_CACHE_VERSION 4 discards v3 caches. The v3 cap of 400 was also exactly
the wrong number for this data: a 400-input transaction needs 401 entries,
so it would have evicted and refetched on every single refresh.
Ships 057629d — add-on stores kept in memory instead of re-read and re-parsed
on every get (a 7.5 MB Aegis store held the window in Not Responding for
16 s) — plus the in-page translator (4fbfc9e, f54ebd6, b43b180).
An add-on's storage.get read and parsed its whole store file on every call,
and storage.set read, parsed and rewrote it — synchronously, on the main
thread. Traced on a real profile (installed 0.3.70): with a 7.5 MB Aegis
store, 30 of the first 35 s of main-thread time went to storage.get, the
window sat in "Not Responding" from 3 s to 19 s, and the first page showed at
19 s. One get cost ~73 ms; Aegis does dozens per state update.
lib/addon-store.cjs keeps one in-memory copy per store, shared by the
add-on's api.storage (addons-host.js) and its pages (addon-storage-* IPC in
main.js). After a one-time load a get costs microseconds; values are copied
in and out (structuredClone), so callers keep the old semantics. Writes are
coalesced (100 ms) and land as temp-file + rename, and are flushed on quit;
a store that doesn't parse is moved aside instead of being replaced by {}.
Measured on copies of the same profile, dev build:
first page 16.9-17.6 s -> 1.5-1.7 s; main thread blocked 24.7-25.8 s of
30 -> 1.0-1.1 s; longest freeze 13.7-15.0 s -> 0.6 s.
The Aegis side (capping its unbounded txCache) ships separately through
Aegis's own update channel. The boot tracer gains total/longest block columns.
Aegis is most of the remaining launch cost: ~165 ms of synchronous
activation, then ~610 ms of main-thread work loading its crypto and
WizardConnect deps. It was left on startup because another session
owns bundled-addons/aegis. NOTE-aegis-on-demand.md says exactly what it
needs, measured against 0.27.1:
- addon.json: "activation": "on-demand", plus panels [{id:"main",
title:"Wallet", page:"panel.html"}].
- activate() must return a promise that resolves once deps are loaded
and, only if the vault is already unlocked, wallets are mounted.
vault.derive never rejects on a locked vault, so awaiting
mountAllWallets there would stall every first call for the full 5 s
host wait.
- Persisted WizardConnect pairings (wc/<walletId>/uris) need a relay
listener: call api.startAtLaunch(true) while any exist, false when
none are left.
- Bridge calls, panel open, wiz:// links and Settings calls already wake
it. The dapp globals are injected from the manifest before it runs.
GOTCHAS: an enabled add-on may not be running, so code in main reaches
add-ons through dispatch() rather than checking isActive()/hasHandler().
Every enabled add-on used to be activated synchronously in initAddons(),
during app.whenReady and before the window exists. That is the largest
launch cost left (boot tracer, 2026-10-03). An add-on can now say
"activation": "on-demand" in addon.json. It is then listed at launch but
not started. Its declared surfaces stay live: "panels" (new: sidebar
panels declared up front), toolbar-menu, context-menu-items and the
page-inject bridge, whose source is read on the first matching page.
activate() runs on first real use: a panel opened, a menu or context
item picked, a message from its panel, tab or page bridge, a Settings
addon-invoke, a wiz:// link (for Aegis). All of those go through
AddonHost.dispatch(), which starts the add-on and waits for it, so no
call is dropped. Concurrent callers share one activation, and
activations run one at a time.
Startup stays the default: the host cannot tell what an older add-on
does in activate(). request-filter add-ons and add-ons that declare no
surface are forced to startup. If activate() returns a promise, calls
wait for it (at most 5 s). api.startAtLaunch(bool) lets an on-demand
add-on ask to be started at launch again (for live relay sessions).
Converted: notepad, screenshot, translate, docx-editor, pdf-editor, vpn
(none has launch-time work: no file association, no auto-connect, and
add-on file tabs are not part of the saved session). Shield and Cookie
Pop-ups stay startup: Shield owns the request filter and must see the
first request; Cookie Pop-ups costs ~6 ms and acts unasked on every
page. Aegis stays startup and untouched: another session owns it. See
NOTE-aegis-on-demand.md (next commit).
Settings › Performance › Startup:
- "Start extensions when first used" (default on). Off = all at launch;
switching it off starts the waiting add-ons immediately.
- "Start the wallet at launch". Shown disabled with a hint until the
installed Aegis manifest allows on-demand. It applies with no Settings
change once Aegis opts in.
- "Preload common menus" gates prewarmOverlays().
- "Use lightest" preset.
New keys are plain SETTINGS_DEFAULTS through the existing settings-set.
No new IPC channels.
Measured: boot-trace, fresh profile, --seconds 20 so the 30 s add-on OTA
poll can't swap Aegis mid-series; warm runs 2-3 of two paired series.
- Add-on activation at launch: 121-154 ms -> 104-174 ms. The six
converted add-ons went from 16-20 ms to 0. The rest is Shield (83-148
ms, noisy) and Aegis (15 ms in this tree's 0.9.0).
- Toolbar painted: 1278-1584 ms -> 1282-1481 ms (within noise).
- With "Preload common menus" off: 0 overlays prewarmed, 8 processes
instead of 11, about 50-70 MB less at 15 s.
The bundled add-on versions are not bumped. Existing profiles keep their
old addon.json, and so stay on startup activation, until those add-ons
ship with a higher version (seedBundledAddons only reseeds a strictly
newer bundle).
translate.silentmode.st had "translate" in the subdomain and in the
LibreTranslate path, which read awkwardly on both the chip tooltip and
the Settings list. The shipped defaults rename the primary peers to
libre.silentmode.st and libre.x (plus lingua.x registered server-side
as an alias — same ip record, so it's a URL users can also remember
without being another independent peer in the client's fallback list).
The chip ran against a single endpoint, which is the fastest way to go
dark: libretranslate.com's public tier moved behind an API key in late
2026, and most of the historical public mirrors (libretranslate.de,
argosopentech, lt.vern.cc, translate.terraprint.co) either 502 at any
given time, serve a parked page, or started requiring a key of their
own. One URL in settings meant one of those going down meant the chip
stopped working.
Settings.translateEndpoints is now an ordered list. The translator tries
each peer in order and returns the first non-error answer; a dead peer
is logged and skipped. Order is preserved — the first entry is the
primary. Shipped defaults put Silent Mode's own instances
(translate.silentmode.st, the BNS name translate.x) at the top and keep
libretranslate.com as the last-resort entry; neither Silent Mode URL
answers today, but a user's chip starts working as soon as either goes
live without a browser release.
Settings › General › Translate pages grew a list editor (same shape as
the quick-links one): PRIMARY tag on row 0, add a peer, remove any row;
bns:// URLs are accepted so a BNS translator doesn't have to be fronted
by an https host. The single-URL `translateEndpoint` setting carried
over from the earlier draft is migrated on load — a custom URL goes to
the front of the list, the historical default is dropped.
The Website-language setting only tells servers what the user prefers via
Accept-Language — many static sites (including names on BCDN) serve one
language and ignore it, so e.g. hello.bch loads in English for every user,
in every language. This adds a translator that converts the page's visible
text in place, so a Lithuanian user reads hello.bch in Lithuanian without
asking the server for anything.
The URL-bar grows a translate chip next to the website-language globe. The
chip lights up when the page's declared `<html lang>` differs from the
user's preferred language. Click it once to translate in place; click again
to revert — originals are kept in a renderer-local state slot and swapped
back without a reload. Right-click opens the chip menu (change target /
translator settings).
The engine lives behind a swappable adapter in main — this ships with the
LibreTranslate backend (POST /translate with {q, source, target, format}).
The endpoint defaults to the LibreTranslate public tier but is settable in
Settings › General › Translate pages, so a user with a self-hosted
LibreTranslate (or Silent Mode's own translate.silentmode.st once it is
up) swaps it there without a code change. On-device Bergamot (the WASM
engine Firefox Translations uses) will plug into the same adapter in a
later release — same contract (array of texts in, array of translations
out), the chip and revert path are already engine-agnostic.
The fetch goes through session.defaultSession.fetch so Tor and add-on
proxy rules apply uniformly, chunks the batch at ~3.8 KB per POST so a
large page spreads across several requests, times each one out at 45 s,
and reports a failure to the chip's tooltip so a dead endpoint reads as
such and not as a silent no-op. The injected walker skips SCRIPT / STYLE
/ CODE / PRE / NOSCRIPT / TEXTAREA and contentEditable subtrees, keeps a
reference to each text node and the original text, and reverts by
restoring from that pair.
