alelyon.com 0.16.0
- Development only: the site's production modules (src/lib, the Pages Functions and the access proxy in cloudflare/) are TypeScript, type-checked by npm run check; what the site and the Functions do is unchanged.
- The developer console's Monitors panel: make, mute, test, switch off and delete monitors, read their history, and add the chat channels alerts go to; switched on by site.config MONITORS_READY after the identity service serves it.
- The Privacy Notice (2026-10-21.6) says what monitors keep and where alerts go.
- The Privacy Notice (2026-10-23) says what Pages short links keep (each link's count and the last visit's time, nothing about the visitor), and that Alelyon keeps which post a scheduled draft became for 90 days.
- The developer console's analytics say what is behind a spike. An error or latency spike's own calls are counted by answer, route and key or app against how calls usually divide, so At a glance and What changed say "mostly 503 on search" and its steps go to those calls and to what the answer asks. A key or app that only follows its route is not named. A spike no one route or caller carries says so, and Logs names what is behind the failures it lists.
- Fixed: a spike at a quiet hour is one finding, not several, nor lost; and spikes are found for small accounts, and for a single route or key, where calls a minute are few.
- The developer console's analytics notice a key or app that stopped calling, or one that started, against the time before. A scheduled job that stopped or a program that lost its key shows even when the account's calls barely move. Steps are given for it: check that it is still running and that its key still works, and see when it last called. A key you revoked or made in that time is not flagged.
- The developer console's analytics, deeper: What changed (error and latency spikes, quiet stretches, a drop or surge against the time before, each with a closer look and its calls); Apdex and the latency distribution; time steps earlier and later and zoom out; compare with the time before; this month so far with where it ends, and rate-limit headroom; a route or a key or app opened in full; Logs patterns, an answers facet, the calls around a call, a call found by its request id, JSON export and keys; Explore, any measure split by kind, route or key, with its charts pinned to Home; saved views in every view; and Customize on every board.
- The developer console's analytics fit inside the console's page frame: the keys list (?) shows the console's own keys after the view's, and a view's bar stays below the console's header as the page scrolls.
- The developer console's analytics, at a glance: every view, and a route or key opened in full, starts with what its charts show in plain words, a verdict (Healthy, Worth a look, Trouble or Quiet), and what to do, each step linked to where it is done and copied as text with a press; Logs searched as a log explorer is (fields, lists, wildcards, answer and duration ranges, "not" and phrases); rate limits in Usage, the busiest keys minute by minute against a key's limit with the calls refused for it marked; compare with a week before in Performance, Usage and Explore; charts read with the keyboard, and drawn faster; every table saved as CSV with its raw figures; and a link to the very time shown.
- Fixed: a link to a view with another time range, or one pasted into an open console, shows that range (the view kept its old figures); and a copy the browser refuses says so.
- The developer console's analytics, connected: one cursor across a view's charts; what changed marked on the charts it shows on, and lit when pointed at; each chart's points as CSV, and the same question in Explore; routes and keys by what changed most against the time before; Explore compared with the time before; Logs' live tail holding new calls while one reads; and [ and ] to step the time window, - to zoom out.
- Fixed: on a phone the analytics views keep to the screen's width, with tables scrolling in place; and a time window stepped back and forth returns to its preset.
- The developer console's analytics mark what happened around the account's calls: Alelyon's own maintenance, as the Updates page records it, and what was done to the account's keys (made, changed, rolled, revoked, and the moment a rolled key's old secret stopped). What changed names the maintenance a stretch came with, and the key change a rise in refusals followed. A route, or a key or app, opened in full sets its figures and charts against the time before and shows its latency distribution and Apdex. ? lists the keys, and a view prints, or saves as a PDF, as a plain report.
- The developer console's analytics say what moved behind a drop or a surge in calls. The key or app, or route, carrying more than half of the change against the time before is named in At a glance and What changed ("mostly from CI deploy bot, which made no calls"), with steps for it: check that it is still running and that its key still works, and see when it last called; for a rise, whether it is expected and how near its rate limit it comes.
- The developer console's analytics judge a time window that reaches now on now. At a glance opens with the past 15 minutes. A spike or a quiet stretch still going is said so ("Errors at 40% for the past 19 minutes"), and its steps go to the newest calls. One that ended is marked with when it ended, and asks for a look rather than calling it trouble.
- Fixed: a spike at a quiet hour of the day, when no single minute holds enough calls to show it, is found.
- The developer console's analytics find a failure on Alelyon's side even in an account whose own refused calls are many: failures are judged against their own usual share, and named as failures. Over 7 or 30 days, a spike counts as still going only while the newest hour, and the hour so far, still show it, and the newest calls are read from this hour so far. Without telemetry, the time before is counted as the same number of days, so a week is no longer set against two, and a view filtered to one route or key says the keys' log cannot show that over time instead of showing every key's calls.
- The developer console can show its times in UTC: a UTC switch in the analytics bar, kept in this browser, moves every chart's axis and read-out, Logs, the tables, what changed and the custom time range into UTC, and the console's other panels follow it. A chart copies as a picture (its title, legend and series), to paste in an issue or a chat, or saves as a PNG file where copying a picture is not allowed.
- Fixed: a day on a chart is named as the UTC day it counts, so west of UTC it no longer reads as the day before.
- The developer console's analytics, ready for the console to open: a Home board that names what needs a look; Performance (requests, errors, p50/p95/p99, routes and callers, failed and slow requests); Usage (calls against the time before, by outcome or kind, busiest hours, rate limits); a Logs explorer with facets, search, live tail and CSV; one time range across them, kept in the address; each board in the app's dashboard schema for agents. Nothing on the site changes until the console opens them.
- Updates: the accounts update of 22 October afternoon is listed with its minute of downtime.