charge-recommendations

Charge Recommendations

When to use

Activate this skill when the user asks anything along the lines of:

• "Any advice on how I charge?"
• "How can I lower my charging cost?"
• "Am I charging the right way for battery health?"
• "Give me recommendations on my charges."
• "What should I change in my charging routine?"

For a full cost summary or a deep dive on a specific session, prefer the charge_review MCP prompt. For battery degradation specifically, use battery_health_report.

Prerequisites

You need a TeslaMate car_id. If unknown:

1. Call teslamate_get_cars to list available cars.
2. If exactly one is returned, use it silently.
3. If multiple, ask the user which one before continuing.

Workflow

1. Pull the dataset (current + previous period)

Call teslamate_get_car_charges twice:

• Current period — default: 30 days ago → now (RFC3339). If the user specifies a different period, use that instead.
• Previous period — same duration, immediately preceding the current window. For a 30-day current window, that means days -60 → -30.

Always pass both start_date and end_date. Use fetch_all=True on any window longer than ~1 month — the TeslaMate API pages at 100 entries.

Skip the previous-period call only if the user explicitly says "just this month" / "no comparison".

2. Classify each session

For each charge:

• Location: home / supercharger / other. Use address, geofence, or the fast-charging flag if present. If unclear, treat unnamed sessions above ~50 kW peak as fast-charging.
• kWh added (charge_energy_added or equivalent).
• Start/end SoC (start_battery_level, end_battery_level).
• Cost if exposed (cost, currency).
• Time bucket from start_date: night (22h–06h), off-peak day, peak day. Adjust if the user told you their utility's windows.
• Peak power if available (kW).
• Skip sessions with kwh_added <= 0 or duration < 5 min.

3. Compute the signals that matter