Invoke an edge function
POST/v1/functions/{project_id}/{name}/invoke
Triggers a deployed handler. The body of the request is NOT transmitted as is: it describes the request to be replayed (method, path, headers, body).
Headers#
| Header | Required | Value |
|---|---|---|
apikey | yes | Project API key |
Authorization | no | Bearer <JWT> |
Content-Type | yes | application/json |
x-aurabase-invoke-mode | no | raw returns the function's raw response (real status, content-type and bytes) instead of the { data, meta } envelope |
Bodysuit#
| Parameter | Type | Required | Description |
|---|---|---|---|
method | string | no | HTTP method seen by the handler. Default POST. |
path | string | no | Path seen by the handler. Default /. |
headers | object | no | Headers passed to the handler (string → string pairs). |
body | any | no | Payload passed to the handler. |
These four fields describe the request that your function receives. On the Deno runtime, req.method, new URL(req.url).pathname, req.headers and await req.json() render exactly what you sent — body is the application body, not the shell. GET and HEAD leave without a body (Fetch spec).
Idempotence#
Idempotency-Key is not read on this route: no deduplication is done, two identical invocations are executed twice. If you need idempotence, handle it in the function code.
Execution Limits#
The timeout is set by function: default 30 s, hard ceiling 300 s (5 min) — it does not depend on the plan. Memory ranges from 16 to 1024 MB. Invocations are limited to their own throughput profile, and each call consumes the project's functions.invocations quota.
Asynchronous execution#
For long work, prefer the job queue: POST /v1/functions/{project_id}/jobs, followed by GET …/jobs/{job_id}, with restart (…/retry) and scrap queue (…/jobs/dlq).