Support & FAQ

Open Settings → AI Engine and choose your path: · Apple Intelligence (recommended): available on supported devices with Apple Intelligence enabled — no model download needed. Turn on Auto-categorization and you're set. · Gemma 4 E2B (offline fallback): agree to the Gemma Terms of Use and download the on-device model (~2.4 GB; a stable network is only needed for this one-time download). Then turn on Preload on Launch and Auto-categorization. Capture and search work perfectly without AI — it's entirely optional.

Apple Intelligence uses the system Foundation Models built into iOS — zero download, instant setup on supported devices. Gemma 4 E2B is IdeaNook's full offline fallback: a ~2.4 GB model you download once, available on all compatible iPhones. Both paths provide categorization, summaries, and image understanding — all on-device, no cloud.

Hold to talk in voice mode, release to send. IdeaNook uses Apple's on-device Speech framework to transcribe your recording — the text appears in the message bubble and is fully searchable and categorizable. Language packs are managed by iOS in Settings → General → Keyboard → Dictation Languages.

When you send a photo, on-device AI generates a semantic description of the image. This description is used for auto-categorization, full-text search, and inclusion in digests — all without uploading the image anywhere.

Check your network connection, or retry in Settings → Manage Model Files. The download doesn't affect core capture — you can always record ideas. If you're on a supported device, try Apple Intelligence instead (Settings → AI Engine) for instant AI with no download.

All data is stored locally via SwiftData. Settings include an “Allow iCloud Backup” toggle: on (default), data is encrypted within the iCloud device backup and restores seamlessly; off, it's fully isolated to the current device. A dedicated export feature is coming in a future version.

On-device models have limits — very short text may land in “Other.” You can review it in category browsing, and we keep improving accuracy in future versions.

All of IdeaNook's AI inference runs locally on your iPhone with no cloud calls, so it neither needs nor requests network permission — that's the core of our privacy promise.