analyze_shot is what makes the rest useful: it turns a binary shot log into labelled indicators, so the assistant can say why a shot ran sour instead of reciting numbers at you. Every write is fenced — no shot control, clamped temperature and pressure, and deletes limited to the profiles it made.
A server for a Gaggimate-modded espresso machine: it reads shot telemetry, manages profiles on the device, records tasting feedback, and keeps local notes on coffees, grind settings and what has worked.
- manage_profile — create, view, update, delete and list brewing profiles on the device, with partial updates so temperature or phases can change without respecifying the rest
- analyze_shot — the raw shot log parsed into puck resistance, channeling risk, temperature deviation, pressure and flow stability and profile compliance
- Every number in that analysis carrying a plain-language band label, so the assistant can read it without knowing the thresholds
- Three depths on that same call — a summary for triage, a per-phase breakdown, or the full time series
- manage_shot_notes — ratings, tasting notes, grind setting and dose recorded against a shot and synced back to the machine
- list_recent_shots — shot history with ids, timestamps, profile names and any ratings you have given
- diagnose_connection — automated checks for reachability and common misconfigurations
- manage_coffee, manage_user_setup, manage_grind_map and manage_brewing_insights — local records of each coffee, your equipment, the grind settings that worked, and patterns noticed across coffees
- Readable resources for the bundled espresso knowledge files, coffee records, setup, grind map and accumulated insights
A Gaggimate-modded espresso machine reachable on the same network, and Python with uv from a checkout. No account, no key. The server cannot start, stop or trigger a shot; temperatures and pressures written into profiles are clamped to safe ranges, and it can only delete profiles it created itself.