AirOrchestra
AirOrchestra · a Torspan product
AirOrchestra — SupportNothing we could break stops your crew.
AirOrchestra runs on the computer in your office. If our end of it went dark this morning, your men would still send, the hours would still record, your documents would still be on your disk and the day would still get written down. That is not a service level. It is where the software is, and it is what makes every other line on this page keepable.
There is one level of support and you already have it. Not a queue, not a tier, not a success manager, not a bot that asks you to describe the issue in a box. There is nothing here to upgrade to.
A bad day at our end is not a bad day at yours
A cloud vendor's support promise is load-bearing. If their servers are down, only their people can fix it, so they have to promise around the clock and staff it. On a computer you own, there is nothing of ours in the path of a crew message — which is the only honest reason a support commitment like the one below can be kept.
| What goes wrong at our end | What your crew notices |
|---|---|
| This website is down | Nothing. It is not in the path of anything. |
| Our release server is down | Nothing. Updates are pulled by your machine, in your window. A machine that never updates again keeps working. |
| Our subscription check cannot be reached | Nothing. The check reports; it does not gate. A licence check that can stop a shop's dispatch on a Friday afternoon is not a licence check, it is a hostage. |
| Nobody at Torspan answers all day | Nothing, that day. The machine in your office keeps answering everybody in your building. What you lose is us, not the product. |
| You stop paying us | Updates stop and support stops. The system keeps running and your records stay yours. We do not brick a company's operational record over an invoice. Ever. |
| Torspan stops existing | The software is on your machine, the model is on your disk, the database is on your disk. Nothing switches off, because nothing of ours has to be reachable for it to work. What our systems know about your machine → |
| Your office internet goes down | This one is real, and it is ours to say. Everyone in the building keeps working normally. The field cannot reach the office machine until the line is back, and what your men sent is held on their phones and delivers when it is. Nothing is lost. Some things are late. |
The last row is a genuine step back from software that lives in a data centre, and printing it here rather than letting you find it in month three is the point of the table. The mitigation is cellular failover on your own office router, and it is on the list we hand you before the machine is ordered.
What you actually get
Written to be kept, not to win a comparison chart.
One address, and a person reads it
admin@torspan.com. Every message is read the same day and answered in writing by the end of the next business day. A question, a change, a new man, a new job, a report you want shaped differently — that is all the same lane and it is all included.
There is no form on this site, no chat window waiting to open, no calendar to book and no ticket number. You write a sentence; you get a sentence back, from somebody who can change the code.
A number that reaches the person who set your machine up
Your account has a direct number, and it is on the receipt you sign the day the machine goes in. Call it any day between 6 in the morning and 9 at night, Pacific, and a person calls you back within two hours.
Six to nine, seven days, is chosen against your clock rather than ours. Crews start before six and the office closes the day after three — the two moments you would actually be calling about are the two a nine-to-five window misses.
It is a callback, not a fix
Two hours is a promise that a human being is reachable. It is not a promise about how long a repair takes, and nothing on this page implies one, because how long a repair takes depends entirely on the fault.
Vendors lose trust on the second promise, not the first. So we make the first one and we do not make the second.
The people who answer wrote the software
Sergei Sysoev and Andrei Tcelikov. Whichever of them set your machine up is named on the receipt you sign, and that is who you call first — the same name in month one and in month forty.
Both names, the filing and the standard they are held to: Torspan, the company torspan.com.
Most of the time you will not need us, and that is the design — not a compliment we are paying ourselves.
Four things you will not find here, and the reason for each
Every one of them is normal in this category, and every one of them would make this page look more serious and be worth less.
| Not offered | Why not |
|---|---|
| A service credit when something goes wrong | A credit against a monthly invoice pays you a few dollars for a lost morning and buys us a way to fail on paper. It is a number that makes a contract look strong and does nothing for a crew standing on a floor. What is offered instead is the thing you actually want: the phone gets answered. |
| A priority tier | There is nowhere to buy your way up to, because there is nowhere up. If there were a faster lane, it would mean the ordinary lane had been made slower on purpose so somebody could sell the difference. |
| A success manager between you and the code | The layer exists to protect engineers from customers. When the person you reach cannot change the thing you are complaining about, you are being managed rather than helped. |
| A promise that somebody drives out | The install is remote and support is remote, and it is identical whether you are forty minutes away or two thousand miles. A truck at your door is a promise about a calendar, not about the fault, and it is the first promise a vendor breaks. |
And there is no charge anywhere in here. Support is in the price on the price page, there is one price, and it does not have a support line under it. If you ever get an invoice with one, you do not owe it.
We cannot look at your computer unless you open the door
When a contractor asks whether we can read his crew's messages, the answer should not be a paragraph. It should be a screen he can open.
| Way in | How it is agreed | What we get, and for how long |
|---|---|---|
| Health signals | Once, in writing, on the order form, with the whole field list printed underneath it | The list below. Numbers, timestamps, versions and codes. Nothing else. Continuous, and you can switch it off. |
| A support session | Each time, by you, in your own app | Whatever the fault needs, written down on your machine as it happens. Sixty minutes, and it closes itself. |
| Nothing else | — | — |
Let Torspan connect to this computer
Andrei Tcelikov is asking to connect for 60 minutes to look at: the message queue.
While connected, he can see this machine's screen and its files. He cannot take a copy of anything.
This session will appear in your Support Access log with the time, the reason and how long it lasted. You can end it whenever you like by clicking Disconnect.
Four things on that screen are doing work: the person is named, not “Torspan support”; the reason is stated before and is written to the log whether or not it turns out to be the right reason; the clock is short and runs down without anybody remembering to stop it; and the record is written on your side, in a screen you open without asking us.
There is no fourth way in
No standing key. No permanent tunnel. No account of ours on the machine. And no break-glass path — a break-glass path is a back door that has been given a respectable name. If you are down and cannot open a session, the failure mode is that we talk you through it on the phone. That is slower, and it is the price of the sentence at the top of this page.
And the rules inside a session
We look at what the fault requires. We do not read the message archive, browse the photographs, or export anything unless you ask us to. A support session is not a licence to read the business.
Taking a copy of your data away to reproduce a bug is prohibited outright. If it is genuinely the only way, it is a written, time-boxed, customer-signed exception with a deletion receipt at the end of it — a procedure written down in daylight so it is not invented at ten at night with a customer waiting.
The health signals, printed in full
Not summarised, not “such as.” This is the list, and it is the same list printed on the order form you sign:
- An identifier for the machine
- The application version
- The model version, and the fingerprint that proves it is that version
- The macOS version
- How long it has been up
- How much disk is free
- Whether memory is under pressure
- When the last backup ran
- Whether it succeeded
- How many turns stayed on the machine
- How many went out
- Counts of error types
- How much is waiting in the queue
- The time of the last heartbeat
And what is never in it, under any setting
- Message text. Any of it.
- People's names, or your job names.
- Photographs, or anything derived from one.
- Documents, or the text taken out of them.
- Hours, dispatch rows, or any record content.
- Your outside-reading key.
- Any free-text field of any kind.
That last line is the one that matters, and it is enforced by a test that fails the build rather than by our good intentions. A single free-text field added on a Tuesday afternoon for a good reason is how a sentence on a website quietly becomes untrue.
Switching it off, and what it costs you
It is a switch in your app. Turn it off and we stop seeing that the disk is filling, that a backup failed, or that the machine has been down for three hours — so those reach us when a person notices instead of when they happen, and support gets slower. There is no reduced tier, no price change and no nagging banner. A consent that is expensive to withdraw is not consent.
When something breaks, here is what you are told
These sentences are written down in advance, before the morning anybody needs them, because the sentence a tired person invents at twenty to seven is the one that overpromises. This is a page out of the book.
Logic board, disk, or it will not power on
Service or a replacement machine, then a restore from the disk in your own room, then the four proofs — a named day back, a named photo open, a PDF drawn, the app answering. One business day, plus getting hardware.
The building lost power
Nothing is said unless it keeps happening. The battery covered the write, everything restarts on its own and the database recovers. This is why a UPS is on the list rather than a suggestion.
It is thinking too long at half past six
We tighten what it loads and what it holds at once. If it is genuinely the machine, we say so, and the fix is a bigger one — your money, our error in sizing it, and we say that part too.
The office line is down
Your provider fixes the line. Ours is the sentence, and the honest version of it is at the top of this page.
The office can, the phones cannot
Restart the path and re-issue the device credentials. We test this from a phone on cellular data, never from the office wifi, because the office wifi proves nothing.
Under twenty per cent free
Scheduled, not urgent. What gets swept is derived — page images, logs, thumbnails. Never a business record. Not a day plan, not an hours entry, not a photo, not a document.
The health check is red
Repaired in place if it is clean, restored if it is not, and the queue replayed. You are told which of those it was.
Answers on the machine are failing
Restart, then verify it is the exact version it should be. If it is not, it is reinstalled from the copy already sitting on your disk.
The outside-reading account hit its cap
You raise it, or it waits. The machine keeps answering everybody the whole time. How the outside lane works →
The limit you set yourself, not a fault
Ninety seconds. We never present a limit you set yourself as a product failure.
Something we shipped broke
If it never swapped in you hear nothing, because nothing happened to you. If it did: one command back to the version that was working, and to the snapshot taken before anything touched the database. Then a plain note within one business day.
The new one was worse for your work
One command, and no download, because the previous one never left your disk. A large re-download on a shop's connection is not a roll-back.
Last night's run failed
We are told on the first failure, not the third. Then a restore test, because a backup that has been fixed and not restored is still a belief.
Somebody clicked the big upgrade
The native pieces get rebuilt and re-verified. We cannot prevent this on a machine we do not manage, and we do not claim we can — the control is that we see it the same day and call you.
Nobody knows either one
If both copies are genuinely gone, nobody opens that disk — not us, not Apple — and the answer is a wipe and a restore. Which is why two named people hold it from day one, and why it is checked once a year.
It is being asked about paper it has not read
The unread count is a number on a screen, not silence. We fix the class of file and re-queue it. What it's read, and what it hasn't →
How a new version reaches you, and how a bad one goes back
On a machine in your building, two thousand miles from anybody who could fix it by hand, an update that cannot be undone is not an update. It is a gamble with somebody else's morning.
- It arrives in the window you chose. The default is two in the morning and you can move it whenever you like.
- It is laid down beside the version that is running, not on top of it.
- A snapshot of the database is taken and verified before anything touches it.
- The new one is checked against itself and smoke-tested while the old one is still the one answering.
- Passes → it swaps over in one move. Fails → it is thrown away, the old one never stopped, we are told, and you are told nothing, because nothing happened to you.
- The previous version stays on the disk for a month.
A failed update you never noticed is a successful update.
Six things we never do
- Update the model on your machine automatically. It is pinned to an exact version and it changes only in a release.
- Ship a model change and an application change in the same release. One variable at a time, or nobody can tell which one broke it.
- Push a release to every machine at once. It runs on our own machine on real traffic for a week before anybody else sees it.
- Change how it behaves by quietly editing the words we give the model.
- Update outside your window without asking.
- Ship a change that throws data away and a change that adds it in the same release. The destructive half waits a full cycle behind the additive half, in the field, where it can be watched.
Each of those is a shortcut that is available and tempting, which is why it is written down as forbidden rather than left to judgement on a bad week.
What's live today
Where this page and today's arrangement are the same thing, and where they are not. A commitment page without a date on it is worth nothing.
Standing today
- The written lane. admin@torspan.com, read the same day, answered in writing by the end of the next business day, by somebody who can change the code. That is the arrangement on the price page and it is the arrangement now.
- One level of support, included, nothing to upgrade to. No credits, no tiers, no manager in the middle. Those are refusals, and refusals do not need to be built.
- The names. Sergei Sysoev and Andrei Tcelikov. Nobody hands you on to anybody else.
- What we are not building — the standing key, the permanent tunnel, the break-glass path, the account on your machine. None of them exists, which is the only version of that promise worth anything.
Arriving with the machine
- The direct number and the two-hour callback. The line, the rota behind it and the timestamped record that makes it measurable rather than remembered arrive with the machine, and the number is on the receipt you sign. Until then, the way in is the address above.
- The consent button and the Support Access log. The screen pictured above is what a support session will look like from your side.
- The health signals. The field list above is the whole list and the build test that holds it to that list is part of the same work.
- The staged update with the automatic roll-back. Described above; being built with the rest of the machine.
- The disconnection test. Before “nothing we could break stops your crew” is measured rather than argued, we unplug our own end for a full working day against a running machine and show that it kept capturing, answering and recording. Cheap, repeatable, and the strongest thing this company can produce about itself.
This block is not a schedule and there are no dates against the second list, because a published plan gets read as a promise. Ask, and you get an honest guess from a person writing the code, labelled as a guess. The same block for the application is on What it won't do; the one for the machine is on Your computer.
Write a sentence. A person reads it the same day.
No form, no chat window, no calendar, no ticket number. If your question is whether this fits your shop, say how many trucks roll out on a Monday and you will get a written answer.
If our office burned down tonight, your crew would work tomorrow.
That is not loyalty and it is not a service level. It is where the software is. Where it runs →