Limits and timeouts¶

Poly enforces two different kinds of limits. Do not debug a 429 as a timeout, or a Job MAX_EXECUTION_TIME_REACHED as a killed server function.

Usage / rate limits (your tenant’s limit tier): how many functions you may own, and how many API / server / variable / chat calls per day or per minute. Over the cap → HTTP 429.
Execution time / resources: how long one server function run may execute on Knative, plus a separate Job wait budget.

This page does not publish free-tier CPU, memory, or time numbers. For SKUs and what a tier includes, see PolyAPI pricing.

Usage and rate limits¶

Your tenant is assigned a limit tier. Numeric caps that are unset generally mean no check (unlimited) for that field.

Typical caps:

maxFunctions — inventory (API functions, server/client functions, webhooks, and some related resources). Create is rejected when you are over.
Function calls per day
API function calls per minute vs server function calls per minute (separate buckets)
Variable calls per day
Chat questions per day

Exceeding a call cap returns HTTP 429 (FUNCTION_CALLS_LIMIT_REACHED).

Client-only work (no server function) is not billed as a server execution. Other resources you touch (API functions, Vari, Tabi, …) still count against the tier. See pricing rather than inventing meters here.

Server function execution limits¶

Each server function run has a Knative request timeout (seconds):

timeoutSeconds = function.limitTime
                 OR tier.serverFunctionLimitTime
                 OR 60

limitTime is per function, in seconds. The 60-second value is the platform default when neither the function nor the tier sets a time.

CPU and memory limits on the container come from the tenant tier when those tier fields are set (millicores and MiB). This page does not list the numbers.

alwaysOn keeps at least one warm pod (never scale below 1). It does not change how long a single execution may run. You can set both limitTime and alwaysOn. alwaysOn is a separate add-on (pricePerAlwaysOnFunction).

The execute HTTP client also has a 30 minute timeout. That timeout cancels only the waiting HTTP client. The Knative server function may keep running after the client gives up.

Job limits¶

Jobs wrap one or more server function calls. Defaults (instance → tenant → environment overlays):

Field	Default	Unit
`maxExecutionTime`	10	minutes (Job wait budget)
`minimumExecutionInterval`	5	minutes (floor for schedules)

Warning

A Job can show MAX_EXECUTION_TIME_REACHED (default 10 minutes) while the server function keeps running until its own Knative limitTime (and/or until logs stop). The Job timeout is a wait race. It does not cancel the server function.

Tenants cannot raise maxExecutionTime in Canopy themselves. Contact support@polyapi.io.

Who sets what¶

Knob	Who
Limit tier on the tenant	Poly ops
Server function `limitTime`, `alwaysOn`, scale	You, on the function (API / deploy config)
Jobs `maxExecutionTime`	Support (not self-serve in Canopy)

Troubleshooting¶

Symptom	Likely limit
HTTP 429 on execute	Day or per-minute call cap, not a timeout
Cannot create another function	`maxFunctions`
Server function dies around 60 seconds	No `limitTime` / tier time → 60s default
Job failed ~10 min; logs still going	Job wait budget vs server function hard timeout
`timeout of 1800000ms exceeded`	30-minute HTTP client timeout; the function may still be running