Posts
- What Eryxon Flow's Supabase architecture means on the shop floor
A practical look at the Supabase foundation under Eryxon Flow, from work-order records and permissions to reconnecting after a network interruption.
- Buzz's event memory: When does an old conversation help?
Buzz puts conversations, reviews and Git events in one searchable workspace. That makes history easier to find, but does not turn every past claim into verified knowledge.
- What changes when UMH Core runs at the edge of a production line?
A single-container edge gateway can simplify the first machine-data deployment, but network boundaries, persistence and recovery still need deliberate design.
- OpenRig can start an agent team. Who trusts the workspace?
Starting an OpenRig team can also affect provider settings and workspace trust. A small coding trial needs an access boundary as well as a team definition.
- Turning a machine MQTT event into an API record with Node-RED
When a machine publishes MQTT messages but the production app expects HTTP, a Node-RED bridge needs explicit mapping, failure handling and a testable data contract.
- Cloudflare Agents and the handoff between agents
Cloudflare Agents SDK can host several persistent agents, but the handoff record, authority boundaries, and ownership of the work still need design.
- Where a Postgres data model for a factory app should begin
Using Supabase Postgres for a custom production app means deciding how jobs, shop-floor events, access rules and ERP identities fit together before adding dashboards.
- Scheduled work in Hermes: Did the agent run, or did the job finish?
Hermes Agent's scheduled tasks, profiles and memory through the practical problem of verifying delivery and handling repeat runs.
- The Grafana screen is on, but nobody uses it during the shift
A live Grafana screen only helps production when the team can read it, trust its timing, and know what to check next. A practical look at shift-facing displays.
- Buzz and the difference between a recorded decision and an authorized one
Buzz puts people, agents, and project events in one signed log. That record improves traceability, but it does not by itself enforce who can approve a production release.
- Carbon's API and MCP: where reading a job ends and changing it begins
Carbon exposes manufacturing operations through an API and MCP. A shop floor scenario shows why access to job data is not the same as authority to change a plan.
- OpenRig can assign an agent a role. Who owns the result?
OpenRig gives coding agents stable team roles and a work queue. That is useful, but it does not replace review evidence, access boundaries or release ownership.