HTTP API

The HTTP API lets an MES, a SCADA system or your own software work with the camera directly over the network, with or without a PLC. From the API you can:

• Trigger an inspection and get its result in the same call
• Read past results, or follow them live as they finish, with their images, heatmaps and segmentation masks
• List recipes and switch the active one
• Switch between Setup and run mode
• Read and reset the pass/fail counters

Go to Integration → HTTP API:

Area	What it gives you
Connection	The Base URL every call starts with, for example http://10.250.0.100/api/integration/v1, the API version, and whether calls need a username and password
Download OpenAPI spec	A machine-readable description of every endpoint, with its parameters and responses. Give it to your developers, or load it into Postman or any OpenAPI tool
Ready to trigger?	A live checklist of what must be true before POST /trigger will work
Endpoints and Examples	Every endpoint, grouped by what it does, and copy-ready example calls

Before the first trigger

The Ready to trigger? panel checks three things. POST /trigger is refused until all three pass, and the error names the one that failed.

1. Run mode. Switch the camera out of Setup from the top bar, or call POST /mode/run.
2. Active recipe. A recipe must be active. Activate one from All Recipes, or call POST /recipes/{id}/activate.
3. Accepts HTTP triggers. In the active recipe, open Step 2: Triggering, choose External, and tick HTTP under Enabled sources.

HTTP is one source among several

You can tick more than one source. A recipe can accept a hardware trigger on the line and an HTTP trigger from your software at the same time. A source that is not ticked is ignored, even if something is sending to it.

Trigger an inspection

curl -X POST http://10.250.0.100/api/integration/v1/trigger \
  -H "Content-Type: application/json" \
  -d '{"wait_ms": 5000}'

The call waits up to wait_ms milliseconds (5000 by default, up to 30000) for the inspection to finish:

Response	Meaning	What to do
200	The finished result	Read judgment: pass, fail or not_inspected
202	Accepted, but not finished within wait_ms	Poll GET /results?trigger_id= with the trigger_id it returned
409	The camera is not ready	Fix what code names: run mode, active recipe, or the HTTP source
503	A temporary condition, such as an inspection still running	Retry

Send one trigger at a time

The camera inspects one trigger at a time. A trigger sent while the previous one is still running is refused with 503, unless the recipe's Image trigger buffer allows it to wait. Send the next trigger after the previous response arrives.

Treat not_inspected as a reject

not_inspected means every inspection tool was skipped, usually because alignment did not find the part. Nothing was checked, so handle it as a reject, never as a pass.

Read results

Call	Returns
GET /results/latest	The most recent result
GET /results	Results, newest first, filterable by recipe or trigger
GET /results/{id}	One result, with each tool's verdict
GET /results/{id}/image	The inspected image, as a JPEG
GET /results/{id}/tools/{tool_id}/heatmap	The heatmap of one region, as a PNG
GET /results/{id}/tools/{tool_id}/mask	The segmentation mask of one region, as a PNG
GET /results/stream	A live stream of results as they finish, from every trigger source

GET /results/stream uses Server-Sent Events: keep one connection open and each finished inspection arrives as it happens, whether it was triggered over HTTP, by a PLC or by a sensor. That is usually simpler than polling.

Each tool reports ok, ng, indeterminate or not_applicable, and the result as a whole reports pass, fail or not_inspected.

Other calls

Group	Calls
Device	GET /health, GET /device, GET /status, GET /whoami, GET /openapi.yaml
Recipes	GET /recipes, GET /recipes/active, POST /recipes/{id}/activate
Statistics	GET /statistics, POST /statistics/reset, GET /statistics/stream
Control	POST /mode/run, POST /mode/setup
Fieldbus	GET /fieldbus, GET /fieldbus/descriptor (the EDS or GSDML file for the active protocol)

A recipe is identified by its number, the same PLC recipe ID a PLC uses to select it, set on Step 5: Outputs.

Wait for the mode switch to finish

After POST /mode/run, poll GET /status until mode is run and mode_changing is false before you trigger. Switching mode takes a moment while the recipe loads.

Sign-in and security

While the camera has no Admin user, the API needs no credentials. Once an Admin exists, every call except /health, /whoami and /openapi.yaml needs the username and password of a camera account, sent as HTTP Basic authentication:

curl -u line3-mes:PASSWORD http://10.250.0.100/api/integration/v1/results/latest

Give each connected system its own account

Create one account per system in Users and Permissions, so the Audit Logs show which system did what. Give it the lowest role that does the job; GET /whoami with that account's credentials lists exactly what it is allowed to call. Turn on HTTPS in System Settings so passwords do not cross the network in clear text.

The API only grows within /api/integration/v1: new endpoints and fields can appear, but existing ones never change meaning or disappear. Code written against it today keeps working after a firmware update.

Where to go next

• Step 2: Triggering, to enable the HTTP source
• Users and Permissions, to create an account for each system
• Industrial Ethernet, if a PLC is involved too
• Remote Storage, to send images and results to a file server instead