* main:
Fix add card not working in the cards screen
Submit and refresh results as you type terms
Fix: When expanding filters, submitting filters should keep the filters expanded
Invalidate cache when filter changes
Switch to gpt-5-chat
Handle filter toggles with JS
Prevent glitch where you see the filter bar expanded when you clear it
Remove unused templates
Fix: all collections link not working after moving to turbo frame
Restore cache I deleted by mistake
Speed up filter expansion with turbo stream
Fix: filtering collections not working after moving menu to turbo frame
Respect params when loading the menu or they won't be reflected!
Add HTTP cache for the menu
Load main menu lazily with a turbo frame
Inline :has(&) as it's REALLY slow in Safari
Extract helper method
Remove transition that is making Safari choke when navigating to Fizzy menus
Reduce page size to reduce load for processing local times on initial page load
Avoid having to parse strings
# Conflicts:
# app/views/cards/index/_add_card_button.html.erb
# test/vcr_cassettes/period_highlight_test-test_don_t_generate_highlights_for_existing_periods.yml
# test/vcr_cassettes/period_highlight_test-test_generate_period_highlights.yml
# test/vcr_cassettes/user/highlights_test-test_don_t_generate_highlights_for_existing_periods.yml
# test/vcr_cassettes/user/highlights_test-test_generate_weekly_highlights.yml
Also, remove the dates and duration from the highlights records. Those are not relevant, only the events-derived
key is. If two users with different timezones have different activities, they should see different summaries.
We don't use to_prompt to generate AI responses anymore, only for events. Grabbing
all the comments don't make as much sense, and I'm sure it is quite expensive
Also, downgrade LLM for commands to mini. It's much cheaper and I only upgraded because of the
interpretation complexities that the insight command introduced.
We are implementing a much better approach with Ask.