[andreasthell]

Follow-up to #7: fixes the remaining correctness findings (5–10) from the code review.

Changes

• Swedish dates for revenue and energy totals. Revenue days and the energy totals' year now use Europe/Stockholm, as the spot price already did, instead of the host timezone. HA does not set the process timezone, so a UTC container showed yesterday's revenue between 00:00 and 02:00. The timezone is now shared as API_TZ in const.py.
• Spot price no longer goes stale. A slot is valid until the next one starts; the last one is assumed to be as long as the one before it (15 min by default). An outdated price list now yields unknown instead of its last price forever. Entries without a Value are skipped instead of raising KeyError, which used to fail the whole refresh because slot selection runs every cycle.
• JWT expiry is decoded with urlsafe_b64decode. b64decode drops - and _, so payloads containing them always fell back to the 10-minute default expiry.
• Logbook seeding uses the newest timestamp among the entries. If the newest line had no timestamp, the seed stayed None and logbook events never fired.
• Outdated CM10 diagnostics are reported as unknown. Current in connectionStatus is the CM10's latest report, however old. Reports older than 3 hours (DIAG_MAX_AGE), or missing ones, now clear battery temperature and internet connection instead of showing the last state indefinitely. The limit is generous because a healthy site's report can be over an hour old (see docs/api.md).
• Failed slow updates are retried after 5 minutes instead of waiting their full interval (up to 4 hours for news). The six last-update timestamps are replaced by one next-due map.

The version is bumped to 1.5.2.

Tests

• New tests for each fix: test_coordinator.py (Swedish dates, retry, logbook seeding, diagnostics age), test_spot_price.py (slot validity, hourly slots, missing values) and a new test_api.py (JWT expiry).
• test_after_last_slot asserted the old behaviour (the last price still shown 22 hours later), so it is replaced by tests for the new behaviour.
• All 86 tests pass. The new tests fail against the previous code. ruff check and ruff format --check are clean.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Tt2YacbPwiWKinpfhCss29

---

Generated by Claude Code