Major features:
- Training plan editor: edit preset days/duration per plan (in-place
override via custom_plans storage; preset ids preserved so existing
records stay valid)
- Voice prompts at 4 fixed points during training (halfway / 30s /
10s / done) via Tencent Cloud TTS (cloudfunctions/tts + utils/voice.js)
- Free-training button: switched from outline (transparent) to ghost
variant (theme-tinted) for visual weight
Bug fixes:
- 4 functional pages: SVG data URIs failed to render in WeChat because
'\#' in colors was parsed as a data-URI fragment delimiter. encode the
whole SVG via encodeURIComponent in utils/icons.js build().
- Old records (saved before id field was added) could not be deleted
(data-id was empty, triggered defensive guard). Backfill stable
legacy-<month>-<index>-<duration> ids in getRecords() and persist.
- settings.js plan-row editor button event was bubbling up to the row's
bindtap (which also fired onSelectPlan). Wrapped in catch:tap.
UI:
- Settings page: 3 hardcoded plans replaced with dynamic buildPlansList
that overlays custom_plans on top of presets
- Plan editor: bottom-sheet modal in settings page (regular view, not
ui-modal — WeChat custom component root element drops position:fixed
in this runtime)
- Free-training button: rounded pill style, full-width, 24rpx gap from
primary action; bottom sheet uses max-height: 88vh + internal scroll
- Version bumped to v1.5, last updated 2026-06-10
Removed:
- Daily reminder section (dailyReminder / reminderTime) — replaced by
the 4 voice prompts which cover the same user need without requiring
long-term scheduling that WeChat mini-programs can't actually do
Misc:
- utils/plan.js refactored to formula-driven: presets declare
totalDays/startTarget/increment/cycleDays, days[] is generated. Same
formula applies to custom plans.
- Timer _remind() guards each prompt on minimum duration so short
free-mode sessions don't fire 'last30' at the start
- Cloud storage cloud_plans field added to data push payload; restored
on first install via _restoreFromCloud
- .gitignore added for local AI tool caches (.reasonix/, reasonix.toml,
.codegraph/daemon.pid)
Before: add-first-then-update caused duplicate docs on each push
because add() always succeeds once collection exists.
Now: query {_openid} first → update if found, add if not.
Each user always has exactly one document in plank_data.
Also removed debug console.log — only production-scale errors remain.
The collection 'plank_data' no longer needs to be created manually.
db.collection().add() auto-creates the collection on first write.
Changed push strategy from 'query-then-add' to 'add-first':
- add() succeeds → collection created, data stored
- add() fails (duplicate) → fall back to query + update
- .get() on non-existent collection now caught gracefully
Also switched updatedAt to db.serverDate() for consistent timestamps.
Problem: All data stored in wx.StorageSync is permanently lost when
user deletes the mini-program or clears cache.
Solution: Sync data to WeChat CloudBase (wx.cloud.database).
- Graceful degradation: if cloud env isn't configured, all local
storage works exactly as before.
- On app launch: if local records are empty, pull from cloud.
- On writes: push to cloud (debounced 2s) after saveRecord,
deleteRecord, saveSettings, updateStreak, setTheme.
- Clear data in settings also clears cloud.
Files:
- utils/cloud.js: cloud sync wrapper (init, pullAll, pushAll, clearAll)
- app.js: cloud.init() + restore on fresh install
- utils/storage.js: cloud.pushAll() after every mutation
- utils/theme.js: cloud.pushAll() after setTheme
- pages/settings/settings.js: cloud.clearAll() when clearing data
- project.config.json: cloudfunctionRoot added
Setup required (one-time in WeChat DevTools):
1. Open 'Cloud Development' panel → enable CloudBase
2. Create environment → copy env ID
3. In app.js, change wx.cloud.init() → wx.cloud.init({ env: 'YOUR-ENV-ID' })
4. Create 'plank_data' collection in cloud database
5. Set collection permission to 'Only creator can read/write'
lc