🚀 New Chart Types

✨ Three new chart modes: Box Plot, Waterfall & Histogram

📦 Box Plot — chart_mode: box

Min / Q1 / median / Q3 / max boxes per time bucket, computed from the dense source samples (independent of points_per_hour). Bucket width follows group_by, or is picked automatically from the window.

Use it when the question is "what's typical, and how much does it spread?"

• 🌡️ Room temperature or humidity — daily spread over a month; instantly shows unstable rooms and day/night swing
• 🖥️ CPU / RAM / load sensors — hourly distribution profiles instead of a noisy line
• 📶 Ping / latency / Wi‑Fi signal — consistency at a glance: a tall box means jitter
• ⚡ EPEX / Nordpool price sensors — the daily price range and where the median sits
• ☀️ Solar production per day/week — spread across the period, not just the total

Box Plot vs Candlestick — they look like cousins, but the body means something different. A candlestick body is open → close of the bucket: direction matters, color tells you up/down, and the order of values is essential — it's a momentum chart. A box plot body is Q1 → Q3 with the median inside: order is irrelevant, there is no open/close and no direction. "Where did it go?" → candlestick. "What was typical, and how much did it spread?" → box plot.

🪜 Waterfall — chart_mode: waterfall

Each entity is one +/− step of a running total with an automatic Total bar (hide with waterfall_total: false) and dashed connectors — the classic bridge chart. The step value is the entity's aggregate over the visible window; the default aggregate is sum in this mode.

Use it when the question is "how do the parts add up to the total?"

• 🔌 Energy balance — grid import + solar production − grid export = real house consumption (the killer use case; see README example)
• 💶 Monthly budget — income steps up, rent/groceries/utilities step down, Total = what's left
• 🔋 Battery balance — charged vs discharged energy over the window
• 🏠 Per-room / per-circuit consumption summing to the house total
• 💧 Water or gas: inflows vs usages

For total_increasing counters (energy/water/gas meters) set aggregate_func: diff — a plain sum would add the meter reading repeatedly instead of the consumption. Use invert: true for the subtracting steps.

📊 Histogram — chart_mode: histogram

The value-frequency distribution over the visible window: X = value ranges (bins), Y = how often the value landed there. Time is discarded entirely. Bin count is automatic (Freedman–Diaconis) or fixed via histogram_bins; multiple entities share the same bins side by side, and X labels sit centered under the bars showing each bin's midpoint.

Use it when the question is "what value does it usually sit at, and how often?"

• 🌡️ Thermostat reality check — how much time the room actually spends inside the comfort band
• 🖥️ CPU / RAM / load sensors — hourly distribution profiles instead of a noisy line
• ⚡ Electricity price — how often the price lands above/below your automation threshold
• 🔌 Appliance power draw — idle vs active shows up as a two-humped (bimodal) distribution
• 📶 Ping / signal strength — one tight hump = healthy, a long tail = intermittent problems
• 🚗 EV state of charge — the SOC ranges the car actually lives in

Histogram vs Bar — a bar chart's X axis is time: each bar is the value of one time slot ("CPU was 62% at 14:00"). A histogram's X axis is value ranges and Y is the count: "CPU spent most of this month in the 55–65% band". Bar tells you what happened when; histogram tells you what's typical and how often. Technically the bar chart draws the points_per_hour display series, while the histogram counts the dense source samples.

📝 Docs

• README: new Box Plot / Waterfall / Histogram sections with anatomy, YAML examples, an option reference (waterfall_total, histogram_bins), "Box Plot vs Candlestick" and "Histogram vs Bar" explainers, and three new columns in the Chart Mode Compatibility table.