Shipped
Already in the 1.0 build.
- Encrypted-at-rest vault: Argon2id + XChaCha20-Poly1305 + HKDF-SHA256, per-entry AAD-bound seal (title, body, tags).
- SQLite repository of per-entry ciphertext, full-text search with BM25 ranking, encrypted attachments.
- Local AI tier router with on-device candle backend across four model tiers (0.5B / 1.5B / Qwen 3 4B / 7B, picked to fit the device): multi-turn reflection, AI summaries, and ask-your-journal Q&A with cited entries + semantic 'find by meaning' — all on-device.
- Voice journaling with on-device Whisper transcription — tap or hold to talk, transcribed and answered without leaving the device.
- Model server you run yourself (Ollama / LM Studio over your own network), off until you set it up and surfaced by the network indicator whenever it is in use.
- End-to-end encrypted multi-device sync through a folder you already sync — sealed files only, exchanged on demand. In beta.
- OpenAI + Anthropic streaming cloud adapters (in the core; bring-your-own-key cloud routing is not yet enabled in the app — today AI runs on-device unless you point it at a model server you run yourself).
- Insights over a week or month window, mined on-device from your own entries, plus an automatic weekly read you can open into a conversation.
- Mood inference from what you wrote — suggested, never assumed, and always yours to change.
- Guided CBT programs (offline, on-device), free on every tier.
- Encrypted .ajbackup format with cross-vault restore.
- Per-tag AI access controls (incl. an absolute #no-ai floor), decoy vault, panic-wipe.
- Shamir secret-sharing for Legacy Mode key recovery (in the core; not yet surfaced in the app).
- Markdown export to Obsidian-compatible vault folder.
- Lifetime Pass entitlement, offline Ed25519 redemption-code path.
- Marketing site: Lighthouse 100/100/100/100, JSON-LD across the board, no third-party JS.