Rule engine

The engine lives in two places: the pure logic in rules.py (what should this channel be?) and the application loop in the plugin (apply_channel_rules() — compare against reality, rate-limit, switch).

Resolution order (per channel)

1. Manual mode wins. A channel in manual mode always resolves to its remembered manual_on state; automation never touches it — no timeout, no auto-fallback.
2. State matrix. In auto mode, rules[state] decides: on, off, or ignore (= no opinion, the channel keeps whatever state it has).
3. Temperature rule (optional, per channel): activates when the sensor reaches the threshold, releases only below threshold − hysteresis (falling-edge dead band, prevents flapping). While active it contributes its above action; while inactive it has no opinion. A missing sensor value keeps the previous verdict.
4. Combine logic (global): how matrix and temp rule merge — temp_override (temp wins while active), and, or or.

Application loop

apply_channel_rules(reason) runs on every printer state change, temperature update (only when at least one temp rule is enabled), settings save, and reconnect. It:

• skips entirely when automation is disabled or the WS is down,
• resolves each channel's target and compares it to the live state from the Panda,
• enforces the per-channel rate limit (min_switch_interval): a switch that falls inside the window is not dropped but deferred — a coalesced timer re-runs the engine right after the window expires,
• sends the switch and records the timestamp.

Initial sync

On every (re)connect the current state is evaluated and applied, so channels switched externally while the plugin was away come back to a defined state. On the first connect after OctoPrint starts, startup_behaviour runs first: leave (default), all_off, or restore (re-assert manual channels' remembered state).

Failsafe

Each channel has a failsafe (off / on / hold) applied on OctoPrint shutdown — the last moment the plugin can still talk to the Panda. A lost connection cannot be failsafed (there is nothing to send to); that case is covered by the initial sync on reconnect. This is an honest limitation of a network-controlled hub, not a bug.

State mapping

map_octoprint_state() folds OctoPrint's state ids into the five matrix states:

OctoPrint state id	Matrix state
STARTING	prepare
PRINTING, RESUMING, FINISHING	printing
PAUSED, PAUSING	paused
ERROR, CLOSED_WITH_ERROR	error
everything else (incl. no printer)	idle