feat: interval schedules for functions - #14231

What does this PR do?

Adds interval schedules for functions, the first phase of #14170. A function can now run every N minutes instead of on a cron expression, and Appwrite picks when inside each interval it runs.

The scheduler builds Shifted(new Interval($minutes * 60), crc32($functionId) % ($minutes * 60)) for these rows, the same pattern the stats scheduler already uses. Each function gets a fixed slot derived from its ID, so a thousand hourly functions spread across the hour instead of all firing at :00. The slot is computed rather than stored, so it survives restarts and redeploys.

API

• New optional interval param on POST /v1/functions and PUT /v1/functions/:functionId: whole minutes between runs. 0 (the default) disables it. Any value from 1 up to the integer column limit is accepted, so 45 or 90 work as well as 60.
• schedule and interval are mutually exclusive: sending both returns 400 general_argument_invalid.
• interval is returned on the Function and Schedule models as an integer.
• The console-only POST /v1/projects/:projectId/schedules accepts interval for function schedules. schedule is now optional there, and exactly one of the two is required.

Minutes rather than seconds because one minute is also the finest cron resolution, and both kinds of schedule share the same lower bound.

SDK constants such as Interval.Hour * 6 need integer enums, which the SDK generator does not emit yet: every enum template writes string values (StatusCode.Found = '302'). The wire type is already an integer, so those constants can be added later without changing the API.

Storage and scheduling

• interval integer attribute (default 0) on project functions and platform schedules, added to existing installs by V25.
• A schedule is active when the function has a deployment and either a cron or an interval. Updated everywhere active is computed: function create/update, deployment activation, and the jobs worker.
• Rows with a negative or non-integer interval (written outside the API) are reported inactive by Source\Functions::runnable(), so the scheduler drops any previously loaded entry instead of keeping the old cadence.
• New schedule.interval span attribute next to schedule.cron.

Not in this PR

• Cloud columns: appwrite-labs/cloud#6399 adds functions.interval and schedules.interval on Cloud and runs before this ships there.
• Cloud plan gate (free plan: no new crons, interval of at least 60 minutes): parked in appwrite-labs/cloud#6383. The initial rollout leaves cron unchanged on every plan.
• Console interval picker with the Recommended tag.
• Integer enum support in the SDK generator, for Interval.Minute / Interval.Hour / Interval.Day constants.
• Carrying interval through utopia-php/migration exports and the CLI's appwrite.config.json.

SDK usage

Node.js

const sdk = require('node-appwrite');

const functions = new sdk.Functions(client);

// Run once an hour; Appwrite picks the minute
await functions.create({
    functionId: sdk.ID.unique(),
    name: 'Report',
    runtime: sdk.Runtime.Node22,
    interval: 60,
});

// Switch to every 90 minutes
await functions.update({
    functionId: '<FUNCTION_ID>',
    name: 'Report',
    interval: 90,
});

// Back to a fixed cron time (interval defaults to 0, which clears it)
await functions.update({
    functionId: '<FUNCTION_ID>',
    name: 'Report',
    schedule: '0 9 * * *',
});

Python

from appwrite.services.functions import Functions
from appwrite.enums import Runtime
from appwrite.id import ID

functions = Functions(client)

functions.create(
    function_id = ID.unique(),
    name = 'Report',
    runtime = Runtime.NODE_22,
    interval = 6 * 60,  # every 6 hours
)

function = functions.get(function_id = '<FUNCTION_ID>')
print(function.interval)  # 360
print(function.schedule)  # ''

REST

POST /v1/functions
Content-Type: application/json
X-Appwrite-Project: <PROJECT_ID>
X-Appwrite-Key: <API_KEY>

{ "functionId": "unique()", "name": "Report", "runtime": "node-22", "interval": 60 }

Rejected requests:

{ "schedule": "0 * * * *", "interval": 60 }  // 400: set either "schedule" or "interval", not both
{ "interval": -1 }                            // 400: must be between 0 and 2147483647
{ "interval": 1.5 }                           // 400: whole minutes only

Test Plan

• Unit, only for what e2e cannot observe: Schedule/Source/FunctionsTest pins the hourly phase to crc32(id) % 3600 and checks that rows with an invalid interval (written outside the API) are reported inactive. The ScheduleDatabase fake moved to its own file so both source tests share it.
• E2E FunctionsSchedule: interval validation (negative, fractional, text, above the column limit), create/get, null means no interval, switching cron ↔ interval on update (including an unlisted value, 45), both-set rejected, and a real run: interval: 1 with a deployment produces an execution with trigger: schedule.
• E2E Projects/Schedules: interval-only schedule without a schedule field, interval rejected for executions, with cron, and when negative; neither field is still a 400.
• Regression: FunctionsCustomServerTest create/get/list/update/deployment, GraphQL/FunctionsServerTest, and migrate on an existing install adds the interval columns. These ran against the previous string-enum revision; CI runs them for the minutes revision.
• PHPStan (host, phpstan.neon), Pint and Rector are clean on the changed files.

Related PRs and Issues

• Function interval schedules (and restrict cron to Pro+) #14170
• appwrite-labs/cloud#6383
• appwrite-labs/cloud#6399

Related: #14170