gardening-watering-tuning

What problem does it solve?

This Skill resolves persistent watering prediction bias by tuning the per-plant weighting parameters that adjust how cadence forecasts respond to base scheduling and climate factors.

Core Features & Use Cases

• Per-plant weight management: Reads and updates only the four editable weights stored in plants/<id>/tuning.yml (base, climate_temp, climate_rh, moss_pole).
• Nightly proposal of deltas from prediction vs reality: Compares recent prediction cadence to actual watering intervals and generates attribute-scoped adjustments without directly writing them.
• Strict guardrails and auditability: Clamps per-adjust deltas (±0.10), enforces declared bounds for each weight, requires enough fresh watering cycles before proposing, and appends an adjustment_log for user-visible history.

Quick Start

Ask Hermes to run the nightly reflect step to propose safe per-attribute tuning updates across your plants, then apply the accepted proposals through the adjustment command.