Mental model

BTU is one product with four runtime pieces:

flowchart TB
  subgraph desk [Frappe Desk - BTU app]
    Task[BTU Task]
    Schedule[BTU Task Schedule]
    RunLater[BTU Run Later]
    Config[BTU Configuration]
    Log[BTU Task Log]
  end
  subgraph daemon [btu_scheduler_py]
    Cron[Cron evaluator]
  end
  subgraph redis [Redis]
    RQ[RQ queues]
    RPC[Scheduler RPC]
    RLKeys[run_later keys]
  end
  subgraph workers [RQ Workers]
    Runner[run_task_by_id]
  end
  Schedule --> Cron
  Cron -->|HTTP enqueue| Task
  Task --> RQ
  RQ --> Runner
  Runner --> Log
  RunLater -->|poll_for_ready_work| Task
  RLKeys -->|poll_for_ready_work| Task
  Config --> RPC
  RPC --> daemon

Data flow for a scheduled task

1. User creates a BTU Task (function path + arguments) and BTU Task Schedule (cron, timezone).
2. Scheduler daemon reads enabled schedules from the database, computes next run times.
3. At fire time, scheduler calls Frappe to enqueue the task on RQ.
4. RQ worker runs btu.btu_core.task_runner.run_task_by_id, which executes the Python function and writes a BTU Task Log.

Control vs execution

Plane	Component	Responsibility
System of record	Frappe app	Tasks, Schedules, Logs, permissions, Desk
Time engine	Scheduler daemon	Cron evaluation, fire-at-time enqueue
Execution	RQ workers	Run Python callables
Transport	Redis	Job queues + scheduler RPC

Why two GitHub repos?

See Why BTU and Technical design.