Connect an Opentrons robot
The agent captures terminal run records from a Robot Server accessible inside the lab. This is result capture: it does not upload protocols, start or stop runs, or send worklists.
Set up the connection
- Register the robot as an instrument and assign an agent on a computer that can reach it.
- Choose Opentrons and Inbound in the connector form.
- Set the Robot Server URL, for example
http://robot.local:31950. - Set the polling interval, normally 30 seconds.
- Use the guided API-version, request-timeout, and discovery-page controls.
- Select a parser/profile compatible with the captured JSON wrapper.
A loopback URL refers to the agent host itself, not automatically to the robot. The robot does not need to be exposed publicly.
What is captured
The collector requests Robot Server health, a discovery page of runs, details for each unseen terminal run, and one page of command summaries. It captures succeeded, failed, and stopped runs; active runs are skipped.
The artifact is JSON containing vendor, robot_health, run, commands, and captured_unix. Source metadata includes robot run ID/status and connector/instrument context. Deduplication is based on connector identity and run ID/status; unchanged-status detail edits are not automatically recaptured.
The live artifact is not CSV. generic.csv.long.v1 cannot parse it. Confirm that your deployment has a parser for this wrapper, or use a defined transformation. The virtual instrument simulator's CSV output is a separate demo path.
Verify
flasktrack-agent doctor --config /etc/flasktrack-agent/agent.json
flasktrack-agent once --config /etc/flasktrack-agent/agent.json
Create a known terminal run, confirm its ingestion and raw artifact, then verify parsed values and mappings. Doctor only establishes source reachability.
Current limits
- Run discovery reads one page, bounded by
opentrons_page_length(1–200). - Commands are requested as one page of up to 1,000 summaries.
- This is not a complete historical archive or a full command-detail export.
- Vendor authentication settings and outbound robot controls are not implemented in this native collector.
Increasing the page length does not enable pagination. See the connector reference for defaults and Troubleshooting for source and parser checks.