Skip to main content
POST
get endpoint workers

Authorizations

Authorization
string
header
required

API key must be provided in the Authorization header

Body

application/json
id
integer
default:12345
required

ID of the endpoint to monitor

Required range: x >= 1
Example:

12345

Response

200 - application/json

Successfully retrieved endpoint workers.

id
integer

Worker instance ID

Example:

67890

gpu_name
string

GPU model of the worker's machine

Example:

"RTX 4090"

num_gpus
integer

Number of GPUs allocated to the worker

Example:

1

perf
number

Performance (load per second) the serverless engine expects from this worker when planning capacity. Unavailable and Offline workers count for a fraction of their full performance, and workers that are in Error or being destroyed count for 0.

Example:

812.4

measured_perf
number

Performance (load per second) measured on this worker

Example:

830.1

reliability
number

Share of routed requests that the worker has completed, from 0 to 1. Reset to 1 each time the worker starts.

Example:

0.98

dlperf
number

DLPerf score of the worker's machine

Example:

105.6

cur_perf
number

Current throughput of the worker, in load per second

Example:

640.2

new_load
number

Load assigned to the worker since its last status report

Example:

0

cur_load
number

Load currently assigned to the worker that has not completed yet

Example:

1500

cur_load_rolling_avg
number

Smoothed load per second handled by the worker

Example:

580.7

reqs_working
integer

Number of requests currently assigned to the worker

Example:

3

loaded_at
number

Unix timestamp (seconds) when the worker became ready after its most recent start, or 0 if it is not ready yet

Example:

1760000000

ready_ever
boolean

Whether the worker has been ready at least once since it was created or since its last template update

Example:

true

started_at
number

Unix timestamp (seconds) when the worker was last started

Example:

1759999700

disk_usage
number

Disk space used by the worker's model, as reported by the worker

Example:

16.2

status
enum<string>

Current worker state. See Worker States.

Available options:
pending,
creating,
created,
loading,
model_loading,
idle,
starting,
stopping,
stopped,
rebooting,
unavail,
offline,
error,
destroying,
update_ready,
update_rebooting,
update_starting,
unknown
Example:

"idle"

available
boolean

Whether the worker's machine can currently run it. An inactive worker whose GPUs are rented by someone else reports false.

Example:

true

outdated
boolean

Whether the worker is running an older PyWorker version than the latest stable release

Example:

false

hidden
boolean

Whether the worker is a replacement that is still being prepared. Hidden workers do not receive requests.

Example:

false

contract_end_date
number

Unix timestamp (seconds) when the worker's rental contract ends, or 0 if it has no end date

Example:

0

replaces_worker_id
integer

ID of the worker this worker is replacing before its contract ends, or 0 if it is not a replacement

Example:

0