DeepSeek Peak-Hours Warning + Local Pricing

A pi extension that:

• warns you when you start a session during DeepSeek peak billing hours (2x pricing), so you don't accidentally burn tokens at double rate
• overrides the local DeepSeek model catalog via pi.registerProvider() so pi's cost estimates use the currently active peak or off-peak rates

Billing rules implemented

• Peak hours (2x pricing): 01:00–04:00 UTC and 06:00–10:00 UTC
• Off-peak all other hours
• Weekend off-peak (effective 2026-08-23 00:00 Beijing): all day on Saturdays and Sundays, Beijing time — handled via an explicit UTC+8 calendar-day check, with the rule date enforced (weekends before that date are normal days)

What it does

On extension load, at every session_start, and at each new turn:

• Updates DeepSeek model costs to the currently active rate set:
  • deepseek-v4-flash / deepseek-v4-flash-vision-exp: peak 0.44 / 1.32 / 0.014, off-peak 0.22 / 0.66 / 0.007 (input / output / cacheRead, per 1M tokens)
  • deepseek-v4-pro: peak 1.32 / 3.96 / 0.044, off-peak 0.66 / 1.98 / 0.022 (input / output / cacheRead, per 1M tokens)
• ⚠️ During peak: warning notify — "DeepSeek PEAK hours — 2x pricing until HH:MM UTC (HH:MM local)" (end is exact, on the hour)
• ✅ Off-peak: quiet info notify — "DeepSeek off-peak rates active (Beijing weekend). Next peak: Mon, Aug 24, 01:00 UTC (02:00 local) in 30m" (exact start time with weekday + date, so a weekend or cross-timezone gap can't be misread)

Both messages include your device-local time and timezone label so you can verify the check matches your clock.

Install

pi install npm:deepseek-peak-hours-warning
# or locally:
pi install ./deepseek-peak-hours-warning

Take effect on next pi launch.

Configuration

None needed — the extension reads the device clock and only overrides the built-in DeepSeek provider's model metadata. Make sure the device clock is synced; getUTCHours() is used for the UTC windows and UTC+8 for the Beijing weekend rule, so it is timezone-agnostic by design.

License

MIT