offline geoip timezone proxy

Offline timezone resolution from a proxy exit IP

The wrapper resolves the browser timezone offline: it reads the proxy exit IP once, over the proxy it was already using, and maps that IP to an IANA timezone against a locally cached geolocation database, with no per-launch call to a geolocation service. That keeps the browser's timezone in agreement with its exit without adding a request a real browser never makes.

The problem this solves is a simple one. If your browser reports a timezone that does not match the country its traffic comes out of, that disagreement is a cheap, decisive flag. The obvious fix is to look up the exit IP against a geolocation service at launch and set the timezone to whatever it returns. That fix quietly creates a second problem: the lookup itself is a request that does not look like browsing.

This page is about resolving the zone from the exit without phoning a geolocation API on every launch, why that matters, and the failure behaviour that decides whether a bad lookup breaks your run or silently ships a mismatch.

Why the online lookup is itself a tell

A real browser, on a real machine, does not contact a geolocation endpoint before it loads a page. Its timezone is already set, by the operating system, from a value the user configured once. So a session that fires an outbound request to a public "what is my location" service at the exact moment it starts is doing something no ordinary browser does, and it is doing it from the same address it is about to browse from.

That request is visible. If it goes out over the same proxy, it lands in the exit's logs next to your real traffic, at a predictable moment, to a category of host that free scripts use and people do not. If it goes out over the host network instead, it defeats the whole point: it reads the host's public IP, not the exit's, so the timezone you set is the timezone of the wrong location.

The development harness for this project did exactly the first thing for a long time: it called an online geolocation API on every launch to resolve the zone. That is fine for a harness. It is not something you want in front of a real session, and the shipped wrapper does not do it.

What the shipped wrapper does instead

The wrapper resolves the zone offline. The only network step it needs is one it was going to make anyway: reading the exit IP, over the proxy, so it knows which address the session actually presents. That IP is then mapped to an IANA timezone using a geolocation database kept in a local cache, with no call to any geolocation API.

The result is that a session behind a proxy comes up with a timezone that agrees with its exit, and the only thing an observer at the exit sees is the browsing itself. The zone is derived, not announced.

Two behaviours make this practical rather than fragile, and they are the rest of this page: the database keeps itself current without a per-launch API call, and a failed resolution degrades in a direction that does not silently ship a timezone that disagrees with the proxy.

Deriving the zone from the exit

The whole mechanism sits behind the default. timezone="auto" is not an opt-in; it is what you get if you say nothing, and it resolves from the proxy exit when a proxy is set.

from invisible_playwright import InvisiblePlaywright

proxy = {
    "server": "socks5://gate.example.com:1080",
    "username": "u",
    "password": "p",
}

# timezone defaults to "auto": the zone is derived from the proxy exit,
# offline, with no per-launch call to a geolocation service.
with InvisiblePlaywright(seed=42, proxy=proxy) as browser:
    page = browser.new_page()
    page.goto("https://example.com")

    tz = page.evaluate("Intl.DateTimeFormat().resolvedOptions().timeZone")
    print("browser timezone:", tz)   # e.g. America/Chicago for a US exit

The browser object is a real Playwright Browser, so every method works as documented upstream. Nothing about the timezone path changes the API surface.

If you already know the zone you want, pass an explicit IANA identifier and the resolution step is skipped entirely. This is the right choice when the exit is fixed and you would rather not spend even the one IP-reading request on it:

with InvisiblePlaywright(seed=42, proxy=proxy, timezone="America/Chicago") as browser:
    page = browser.new_page()
    page.goto("https://example.com")

An explicit IANA string is the only way to override the derived value. Everything else, including the empty string, means "resolve from the exit". The full set of accepted values is in Configuration.

Keeping the database current without an API call

The offline database stays current without an API call. On each launch the wrapper makes one cheap, unmetered check against a permalink that always points at the newest published build, and downloads only when the local copy is behind. There is no metered version check and no fixed multi-day recheck window.

An offline database has an obvious failure mode: it goes stale, IP ranges get reassigned, and eventually the exit maps to the wrong country. The naive fix is a version check against a package index or a release API on every launch, which puts a metered request back into the launch path and, on a busy host, can hit a rate limit.

The wrapper avoids the metered request. On each launch it makes one cheap, unmetered check against a permalink that always points at the newest published build of the database. That check is a lightweight HEAD whose redirect carries the current version in its target, so a single response tells the wrapper whether the local copy is behind, with no API call and no rate limit to hit. Only when the permalink names a newer build than the cache does it download; otherwise it uses what it has. Old versions are pruned so the cache does not grow.

The upstream database is rebuilt on its own weekly cadence. The per-launch check is not a weekly timer on our side: it runs every launch and simply does nothing on the common path where the cache is already current. That is deliberate. A timer that only rechecks every N days is a window in which a freshly published fix does not reach you, and the HEAD is cheap enough that there is no reason to open that window.

What happens when the lookup fails

When resolution fails, the launch fails, with a proxy or without one. The interesting part of any resolution step is what it does when it cannot resolve, so both cases are set out below.

Situation	On a failed resolution	Why
Proxy set	Fails early at launch	A foreign exit paired with the host timezone is a mismatch worth stopping for
No proxy	Fails early at launch	The host's timezone can disagree with the host's own address, which is why it is resolved at all

With a proxy, a failed resolution fails early. The reason is precise: if resolution fell back to the host's own timezone while the traffic exits through a foreign proxy, the session would come up with exactly the browser-versus-exit mismatch this whole path exists to prevent, and it would do so silently. A foreign exit paired with the host timezone is a timezone_mismatch waiting to be scored, so the wrapper refuses rather than ship it. You find out at launch, not from a block three requests later.

Without a proxy, the session presents the host's own address, and the timezone is still resolved from that address rather than read from the host. The reason is the case this section used to wave through: a server kept in UTC whose public address is in another country. Until invisible-core 47.41.0 a failed resolution here fell back to the host timezone and the session went on, which handed exactly that server a timezone contradicting its own IP, with nothing said. Now it fails at launch and names the cause. To launch with no network at all, pass an explicit IANA zone, which asks the network nothing.

So the rule is a single one: a resolution that cannot be honoured is an error, never a quiet guess. This matters most when you rotate proxies across sessions or run a different exit per browser context, because each new exit is a fresh resolution, and each one either agrees with its browser or refuses to start.

Measuring it: one launch, zero geo-API calls

The claim worth checking is not "the timezone is set" but "the timezone was set without an online geolocation call". You can confirm both from inside the session, using two standard browser APIs: Intl.DateTimeFormat().resolvedOptions() for the zone, and Date.prototype.getTimezoneOffset() for the offset it must agree with.

from invisible_playwright import InvisiblePlaywright

proxy = {"server": "socks5://gate.example.com:1080", "username": "u", "password": "p"}

with InvisiblePlaywright(seed=7, proxy=proxy) as browser:
    page = browser.new_page()

    # 1. The zone the page sees, derived from the exit:
    page.goto("https://example.com")
    tz = page.evaluate("Intl.DateTimeFormat().resolvedOptions().timeZone")

    # 2. The offset a script would read, which must agree with the zone:
    offset = page.evaluate("new Date().getTimezoneOffset()")

    print("zone:", tz, "offset(min):", offset)

For the second half of the claim, watch the process's own outbound connections at launch with whatever packet or connection tool you already use. Behind a proxy you will see the session's traffic and a single request to read the exit IP; you will not see a request to a public geolocation host on every launch, because the IP-to-zone mapping happens against the local database. That absence is the point. The dev harness produced one such call per launch; the shipped wrapper produces none.

Because every field is seed-derived, seed=7 gives the same identity each run, so you can repeat the measurement and compare it against a stock browser on the same exit. The zone should match the exit's country and the offset should match the zone. A session that comes up with the host zone behind a foreign exit is the failure this path is built to turn into a launch-time error instead of a silent tell. For everything else that has to line up once the exit and the browser tell the same story, see the WebRTC surface behind a proxy.

Conclusion

Matching the browser timezone to the proxy exit is table stakes. Doing it without adding a per-launch request to a geolocation service is the part that keeps the fix from creating a new tell, because that request is something a real browser never makes and an exit operator can see. Deriving the zone from an exit IP the session was already going to read, mapping it against a local database that keeps itself current with an unmetered check, and failing early only when a quiet fallback would ship a mismatch, is how the shipped wrapper does it, and all of it lives behind the default timezone="auto".

Short answers to the questions that lead here

How do I set the Playwright timezone to match my proxy? You do not have to. Leave the default timezone="auto" and the zone is derived from the proxy exit at launch. Pass an explicit IANA identifier only when you want to override that.

Does resolving the timezone call an online geolocation API? No. It reads the exit IP once, over the proxy, and maps it against a local database. There is no per-launch call to a geolocation service.

Will my session break if the timezone lookup fails? Yes, on purpose, with a proxy or without one: a timezone that contradicts the exit is a mismatch worth stopping for, and the host's own timezone can contradict the host's own address. Pass an explicit IANA zone to launch without the lookup.

How does the offline database stay current? Each launch makes one cheap, unmetered check against a permalink to the newest build and downloads only when the cache is behind. There is no metered API call and no fixed multi-day recheck window.

What if the machine is offline? A cached database is reused, so a launch that cannot reach the update permalink still resolves from the last copy it downloaded. Only a cold cache with no network leaves the zone unresolved.

Why not just set the timezone to my own machine's? Because behind a proxy your machine and your exit are in different places, and a browser whose timezone is the host's while its traffic leaves a foreign exit is the exact disagreement detectors look for.

Sources

• This project's runtime geolocation path, which resolves timezone="auto" from the proxy exit against a locally cached database, and the fail-early-with-a-proxy behaviour described above.
• The development harness's earlier online-lookup approach, kept for the harness and deliberately not shipped in the wrapper.
• IANA, Time Zone Database, retrieved 2026-08-28
• MDN, Intl.DateTimeFormat.prototype.resolvedOptions(), retrieved 2026-08-28
• MDN, Date.prototype.getTimezoneOffset(), retrieved 2026-08-28

See also: when the timezone does not match the proxy for everything that has to agree, and rotating proxies across sessions where each new exit is a fresh resolution.

---

Written while maintaining invisible_playwright, a Firefox patched at the C++ level driven by stock Playwright. The offline path exists because the online one, convenient in a harness, is a request a real browser never makes.