For the complete documentation index, see llms.txt. This page is also available as Markdown.

Services

Optional edge container services a Deployment Manager device can run alongside the inference server.

Alongside the inference server, a managed device can run optional Roboflow services. Add and configure them from the device's Configuration tab, described in Update Device Configuration.

These services are available exclusively for Enterprise customers. Contact the Roboflow sales team to learn more.

Service
What it does
API reference

Stores inference events on the device, with automatic retention and optional cloud backup.

Reads and writes PLC tags over Allen-Bradley, Modbus TCP, or Siemens S7.

Publishes data as OPC UA tags for PLCs and SCADA systems.

Streams an uploaded video file as an RTSP source for testing.

Using the APIs

Each service runs as a container on the device and serves its own HTTP API on the device's address, not through api.roboflow.com. Use them when code on or near the device needs to read inference events, drive a PLC, or control a test stream directly, without a round trip to Roboflow.

The base URL host is the device's IP address, shown on the device page in Deployment Manager. Do not derive it from the inference server's server_url, which can point at a proxy or tunnel where the service ports are not reachable.

Service
Base URL
Interactive docs

Event Store

http://<device-ip>:8001

/docs

PLC Relay

http://<device-ip>:8007

/docs

RTSP Simulator

http://<device-ip>:8080

/docs

OPC UA Server

http://<device-ip>:8092/api

/api/docs

The interactive docs each device serves reflect the exact service version running there. When a device disagrees with these pages, trust the device.

Two error conventions to know before writing a client:

  • The Event Store, PLC Relay, and RTSP Simulator return errors under a detail key that is a string for rejected requests but an array for 422 validation failures, so parse it by type. The OPC UA Server uses an error key instead.

  • Several endpoints report downstream failure inside a 200 response, because the request reached the service but the PLC, OPC UA server, or datastore behind it did not answer. Read the body, not just the status code. Each service's API reference calls out where this applies.

Last updated

Was this helpful?