n8n agents
Four workflows handle dynamic pricing, consumer purchases, restocking, and sellout logging. Live store activity directly influences agent behavior.
n8n
Webhooks
03 / Pipeline
Owl Park.
Live data.
Four n8n agents feed Supabase, Fabric models the data, and Power BI turns events into decisions — in real time.
Pipeline flow
Supabase source
Captures transactions, weather impact, events, products, visit dates, and warehouse activity as the raw source-of-truth layer.
Supabase
Postgres
Fabric medallion
Fabric pulls on a real-time schedule, shapes data into dimensional and fact models, then writes to Warehouse (live) and Lake (daily backup).
Fabric
Lake / Warehouse
Power BI + AI
Power BI dashboards drive schedule optimization. Claude + PBI MCP suggest pricing-agent tunings inside critic-enforced fairness boundaries.
Power BI
Claude
MCP
Want the deep-dive? The full architecture infographic has every box and arrow.
Open full infographic ↗Notes per phase
Why these choicesWhy n8n for ingest?
n8n made it cheap to run four agents in parallel without orchestration overhead — webhooks in, webhooks out, every workflow self-contained.
Why Supabase as source?
Postgres-native means real SQL for downstream consumers. Free tier covers the demo; row-level security keeps it safe to show off.
Why Fabric for modeling?
OneLake + Warehouse gives a live serving layer (sub-minute) AND a Lake backup (daily) without copying data twice — Fabric handles the duplication semantically.
Why Claude + PBI MCP?
The MCP server lets Claude propose model tweaks against the Power BI semantic layer. A critic loop checks for fairness before any pricing-agent change ships.
Up next
Swipe →
04 / Shop
See the agents in action
Watch pricing, restocking, and fair-buying agents respond to live purchases.
Open →
05 / Resume
Resume + ATS
PDF, downloadable, plus an ATS plain-text version.
Open →
02 / Work
Back to the catalog
Featured Power BI build, fraud detection, churn modeling, and certs.
Open →