RoastOS

Autonomous Roasting Operating System

An autonomous control system for sensors, heat, airflow, safety limits, manual takeover, and roast profile tracking.

Overview

Problem

High-end roasting requires more than telemetry. The system must follow a lagging physical process, coordinate hardware, support immediate manual takeover, and preserve each session for review.

Implementation

Built system

ESP32 edge controller with PT100, environment, RPM, heater PWM, blower, chaff-fan, and drum-control surfaces
Modbus telemetry and Python / FastAPI control brain operating on a 10 Hz loop
Sensor Trust Layer, Safety Supervisor, command limiting, and manual / hybrid takeover
SQLite telemetry and event backbone with predictive profile-tracking concepts

Operator view

Roast state made inspectable

A grounded representative view connects roast phase, material color, telemetry, safety state, and the operator's control context without presenting a cinematic simulation.

System screens

RoastOS

Select a screen to review full size

How it works

How the system works

ESP32 edge controller for temperature sensing, environment context, heater control, fan commands, RPM, PID, Modbus TCP, and OTA
Python/FastAPI control service with a 10 Hz Modbus loop, live telemetry, hardware commands, and state endpoints
Sensor Trust Layer and Safety Supervisor for range, jump, stale-signal, cross-check, failsafe, and command limits
Thermal estimation, rate-of-rise context, predictive braking, and estimated weight loss
Live roast curve, automatic first-crack marker, development tracking, RGB bean-color monitoring, and target-profile matching
SQLite backbone for beans, roasts, telemetry, events, markers, calibration, learning feedback, and future NFC/inventory workflows

Key details

Scope and boundaries

Recovered source material was used only to describe capabilities. Public reconstructions exclude credentials, network values, private paths, and identifiers. Safety behavior remains under Alpha / QA-before-production review.

Capabilities

System capabilities

Autonomous execution

Tracks target profiles through sensor state, predictive control, and coordinated heater, blower, agitator, and chaff-fan commands.

Visual roast intelligence

Connects live curves, first-crack marking, RGB color tracking, digital-twin state, and profile-objective matching.

Bean and learning ecosystem

Links profiles, sessions, bean identity, inventory, expected yield, final results, notes, and refinement suggestions.

Outcome

Result

A four-view lab system showing how physical control, operator experience, bean identity, and profile engineering can make a complex roast process understandable and controllable.

Why it matters