Logika writes to you rarely, and only when something changed: new issues in an environment you follow, and what your agent did about them. Each email is short, and everything in it is a link to the page that tells the rest. There are no reminders, no digests of things you've already been told, and no tracking.
The new-issues email
When a pass finds new issues (or a fixed issue comes back) in an environment, everyone subscribed to it gets one email, once. It's never repeated, however often the issues recur; if you ignore it, they wait in the console.
Subject: shop (prod): 3 new issues
3 new issues in shop (prod):
- [Nov 3 12:04 UTC] checkout timed out ×2
- [Nov 3 12:04 UTC] checkout.captured: amount must be > 0 ×5
- [Nov 3 12:09 UTC] refund failed with an error; again, after it was fixed
Wake my agent
Each issue is a line: a pill with when it was raised, its title as a link, and how many times it happened so far when that's more than once. The title is the ledger speaking: the event's name and what it broke, or a rule's or an invariant's own description ("An order is charged once"). "Again, after it was fixed" marks a regression: an issue your agent closed that came back on code shipped after the fix.
Below the list, one line says what your agent is doing about it, depending on the environment's wake setting and the state of your monitor:
| You read | Because |
|---|---|
| Wake my agent | the environment waits for someone to ask (on_request, the default), and your Logika Monitor is connected |
Run logika login in the repository to start Logika Monitor and ask your agent to fix the issues. | the same, but your monitor isn't connected, so a button couldn't work |
| Your agent is on it; its report will follow here. | your agent took the issues: the environment is auto and your monitor took them, or your agent is already working there |
| Your agent couldn't start: “…”. Start it and ask it to fix the issues. | your monitor tried, and the agent didn't start; its own words, when it gave any, are quoted |
| Your agent hit its $50 limit for November; raise the limit in Settings. | you reached your monthly spending limit in the project |
| nothing | the environment is set to never, or a teammate's agent is on it |
You're never told about a teammate's agent: if theirs couldn't start, your email reads as if nobody had taken the issues. Wake my agent is offered only where it can work: while your monitor isn't connected, the email says to start it, in one line, instead.
Also open, Also changed
Under that, two short lists may ride along. Also open holds what's still waiting in the project: other open issues, instrumentation feedback, and account notices (your quota at 80%, a ledger deployed with warnings, an address that bounced). Also changed holds what Logika noticed that no release is to blame for: traffic rising across several events, a funnel's conversion falling from 4.1% to 3.2%. Each is told once, on the next new-issues email; the console's What changed in Analytics has all of them, with a Why? button that asks your agent. See Analytics.
Neither list ever sends an email on its own. Account notices that had nothing to ride along on for a week are sent together, in one email.
The report, in the same thread
What your agent does comes back as a reply in the thread of the email that announced the issues, so an issue's whole story is one thread in your inbox:
Subject: Re: shop (prod): 3 new issues
Your agent finished: 1 fixed in pull request #7; 1 still open in pull request #7; 1 dismissed.
- Fixed checkout timed out: No retry around the capture call, which the provider answers 503 under load
- Still open payout.transfer failed with an error: needs the owner
- Dismissed refund failed with an error: an intended decline
The first line says why the agent stopped (it finished, or "rounded up at its 2h limit" when the run passed its environment's time limit) and what became of the issues, with the delivery linked, leaving out a count of none. When it fixed and dismissed nothing, it says so: "Your agent couldn't fix these issues; 2 are still open in pull request #7:". Each issue then gets a bullet led by a coloured pill (Fixed in green, Still open in yellow, Dismissed in grey) with the agent's reason: the root cause for a fix, why it's not a problem for a dismissal, what's in the way for one left open; and its title links to its page, as in the first email.
When the run went wrong, the reply says that instead:
- "Your agent stopped: “…”. Start it and ask it to fix the issues.", when what it said reads as signed out: nothing to wake, since a woken agent would stop the same way; you start it on your computer.
- "Your agent stopped before reporting back: “…”.", with a Wake my agent again button.
- "Your agent was stopped: Logika Monitor stopped responding on your computer.", with the same button.
- "Your agent finished after another agent took over; its work is at pull request #7: review or close it."
Each of these lists the issues the run had taken, open again. Where the button couldn't work yet (your monitor isn't connected), one line says to start Logika Monitor instead. The report goes to you alone: nobody is emailed about a teammate's agent, and what a person does by hand, in the console or a terminal, is never emailed: the thread is the agent's.
Other emails
- Your spending limit: once a month at most, when your agent's runs in the project, across its environments, reach the soft limit you set: it finishes its current work and starts nothing more there this month, unless you raise the limit.
- Your agent couldn't start: you pressed Wake my agent, but Logika Monitor wasn't answering on your computer for 5m, so nothing started; the email says to start it with
logika loginin the repository and ask your agent to fix the issues. - Account notices: those that waited a week for a ride, together.
- And the ones every service sends: an invitation, an order's confirmation, a project's notice reminders after a missed payment.
What every link does
- An issue's title opens its page in the console: what happened, when, what's being done, and the evidence your agent reads.
- Wake my agent opens a page that starts your agent on the project's open issues, an environment at a time, once your Logika Monitor is connected; if it isn't by then, the page says so and waits, starting the agent the moment it connects. The link is yours alone, works once, and expires after 7d; it runs no code by itself.
- The pull request opens it, wherever your repository hosts it.
- Settings, in the spending limit's line, opens the project's settings, where your limit is, beside your name among the people, and billing.
- Logika, at the bottom left of every email, opens the console.
- Unsubscribe, at the bottom right of every email, stops all of Logika's email to that address in one click, with no sign-in; to receive fewer emails, change which environments you follow in the project's settings instead.
What we don't do
Every email has an HTML part, in the same paper look as the console, and a plain-text part that says the same thing, links written out. There is no tracking pixel, and no link is rewritten to count your clicks. Anything your application logged, or your agent wrote, is shown as data: grey and quoted in HTML, between ⟦ and ⟧ in text, so nothing a user of your app types can pose as an instruction. Owners can see every email the project sent in the last 90d, and whether it was delivered, in Settings' Emails we sent.