CEDX Functions · Developer & low-code
Functions is the estate's compute runtime: 80 functions across 6 regions, every invocation timed, every cold start costed in dollars, and an error budget that says how burned it is — not just that it is.
The most expensive line on the cold-cost board reads fn-slot-book · $173.92. When a slow function costs real money, the bill should name it. This one does.
Runs on demo data. The cold-cost card is labelled EST. on the screen itself.
What it is
The Invocations table is a 320-row sample with kind, status, milliseconds, init time and region on every row. fn-ledger-post sits at a 500 with a 13,707ms init — that is a cold-start failure you can point at, not a mystery. Chips cut the view to cold, warm, 5xx, slow or throttled.
us-west-2 carries 7.17M invocations a day at 6% cold; eu-central-1 runs colder at 8.6% with a 0.42 error rate; the global edge tier runs 0.5% cold at a 414ms p99. Each region's share of spend is on the same row, so placement is an economic decision, not a habit.
280 lines sampled with level, function and message: fn-cart-merge fails with “OOM heap limit exceeded”, fn-tax-calc with “Throttled concurrency cap”, and fn-catalog-search logs “Unhandled error” three rows running. Timeout and OOM get their own filter chips because they deserve them.
Product tour
Not a mockup and not a concept deck. This is what opens at /app/functions.
01 — Overview
Invocations at 24.9M a day, error rate 0.2%, p99 at 380ms, cold starts at 10.6% of the sample — and then the cost cards: $1.7K a day in cold-start spend with 45 budgets on watch. Alerts arrive root-caused: a cold spike on billing-settle names init p99 1.8s and a 72%-burned error budget.
02 — Invocations
320 invocations with the failures as legible as the successes: 80 server failures, 91 slow or throttled, 34 cold. fn-report-export 404s with an 1,100ms init; fn-image-resize 503s at 81ms. The stream badge says the sample is live.
03 — Regions
Six regions with invocations, cold %, error %, p99 and spend share. ap-southeast-1 holds 10 functions at 4.44M invocations a day and the best error rate in the fleet — 0.13. The eu-west-1 cold tax is an alert, not a shrug: 5.8% cold with report-export init at 2.4s.
04 — Logs
From the billing-settle cold-spike alert to the logs in one filter: level, function and message per line, 47 distinct scripts covered. The “Throttled concurrency cap” lines pair with the 429 chip in Invocations — symptom and cause in the same product.
Who runs it
Roles, not references. We have no named customers yet, so nobody in these photographs is quoted, credited or claimed as one.
Owns placement and budgets: which of the 80 functions sit in which region, and why the cold bill is $1.7K a day rather than twice that.
80 functions · 6 regions
Reads the error budget like a gauge: 72% burned on billing-settle means the fix ships this week, not this quarter.
error budget · 72% burned
The callers behind the traffic — scheduled jobs, event handlers, retries. They read invocations by trigger and argue about cost from the same table.
24.9M invocations/dayThe shape of it
Every figure below is legible in the captures above. Nothing here is a projection of your estate — it is the state of the demo data.
How it runs
80 functions sit across 6 regions, us-west-2 primary — each with its own cold, error, p99 and spend profile, so placement is a decision you can revisit.
The overview reads the last 24 hours: invocations, error rate, p99, cold-start percent and cold dollars per day, against a 99.9% availability target.
Alerts arrive root-caused — the billing-settle spike names cold 8.4%, init p99 1.8s and an error budget 72% burned, not just a red tile.
Invocations and logs carry the evidence: fn-ledger-post's 500 with a 13.7-second init is a cold-start problem, priced on the board at $83.07 a day.
One record
Functions is not a separate platform with a connector. It is the runtime the estate's routes, schedules and deployments already share.
The route table in API names the fn-* functions backing each path — blast radius and all.
CEDX ScriptsScheduled automation runs on the same metered model: duration billed in minutes, failures logged by name.
CEDX DeployA function ships through the same risk-scored deploy pipeline as everything else — canary, window, rollback.
CEDX SandboxA cloned environment is where a cold-start fix gets rehearsed before it touches the production region split.
Limits
Finding this out on the third call is worse for you than reading it here, and worse for us.
Start
Open it right now
Your first functions
Functions with the rest of it
Questions
Yes. Every screenshot is a capture of the running build and you can open the same build at /app/functions. It runs on demo data from the Northline Systems demo workspace.
Real money, on screen. The cold-cost board prices each function's daily cold-start spend — fn-slot-book at $173.92, fn-ledger-post at $83.07 — so a slow init is an engineering task with a dollar figure attached, not a folklore complaint.
Like for like on one table: invocations per day, cold %, error %, p99 and share of spend. The demo fleet runs us-west-2 primary at 28.8% of spend, with a global edge tier at 0.5% cold and a 414ms p99.
It shows up twice, on purpose: a 429 chip and count in Invocations, and the “Throttled concurrency cap” line in Logs. The payments-path alert pairs the two — elevated burn with 429s under concurrency.
No certification has been issued. What we can evidence about hosting, encryption, tenant isolation and production access is written up on the security page.
Live build, demo data, no card. Then find the function you would most want priced per cold start.