- fetch_openai_models: loop on has_more + last_id (?after=) and fall back to next_page_token (?page_token=)
- fetch_anthropic_models: loop on has_more + last_id (?after_id=&limit=1000)
- fetch_google_genai_models: loop on nextPageToken (?pageToken=&pageSize=1000)
- Dedup by id across pages, MAX_PAGES safety cap per provider
- Bump version to 0.4.1
- Usage dashboard: KPIs, model split, daily bars, year heatmap, recent requests
- Sessions management: list/preview/archive/delete with streaming preview
- Double-click daily bar opens per-model breakdown modal
- Paginated recent requests (30/page)
- Hide pi option from nav (kept code, just removed from UI)
- Timezone fixes: today/day_key/heatmap now use local calendar day
- Streaming preview with 20MB byte cap + 500-char frontend clip
- README attribution to kimicode-dashboard (MIT, JochenYang)
Bug fixes:
- Fix config.toml wiped to single provider on switch (now writes all providers)
- Fix managed:kimi-code oauth key corruption (oauth/kimi-code was rewritten
to oauth/managed-kimi-code, breaking the official subscription)
- Fix default_model not switching due to React 18 deferred state update
(updateConfig now sets configRef synchronously before setConfig)
- Strip KimiSwitch-private default_model from config.toml provider output
Architecture:
- config.toml is now the authoritative source for Kimi Code provider data
- SQLite stores only metadata (notes, official URLs, remembered models)
- Load reconciles config.toml with SQLite for external /provider changes
- Activate is a no-op for KimiCode (save already writes config.toml)
Features:
- Selected provider moves to top of list on switch
- Add models-dev context size lookup
- Add tauri-plugin-single-instance
Tests:
- 3 new regression tests: oauth key round-trip, all-providers write,
private default_model strip
- Replace text '+' icon with centered SVG plus icon for the add-provider button
- Add system tray: minimize-to-tray and quit-from-tray menu (minimize hides window, close exits app)
- Remove unused model role field (no functional effect, only stored/displayed)
- Fix multiple providers showing as active in Kimi Code mode: only the default_model's provider is marked active
- Copy /reload to clipboard after provider switch as a convenience hint
- Bump version to 0.2.0
KimiSwitch Dev