Privacy Architecture: Your Audio Never Leaves Your Device

When you record a conversation with dijin, the audio never touches a server. This is not marketing — it is a fundamental architectural decision.

🔑

Your audio never leaves your device. This is not a setting you can toggle — it is how dijin is built.

The On-Device Stack

Component	Purpose	Runs Where
On-device speech engine	Speech-to-text transcription	100% on-device
on-device language model	Insights, memory, local reasoning	100% on-device
Semantic search model	Semantic search embeddings	100% on-device

What Gets Synced vs What Stays

✗ Synced (encrypted)

✗Text transcripts

✗Insights and action items

✗Speaker labels

✗Usage metrics (anonymous)

✓ Never Leaves Device

✓Audio recordings

✓Voice embeddings

✓Raw audio buffers

✓Voiceprint data

ℹ

Cloud sync is entirely optional. You can use dijin completely offline — every feature works without internet.

Data Flow

Privacy Data Flow

Microphone │ ──▶ Audio buffer (RAM only)

speech engine │ ──▶ Text segments (on-device)

┌───────────────────┐

Audio buffer

DISCARDED

(never saved)

└───────────────────┘

Local DB │ ──▶ Local database (encrypted)

(optional)

└────────────┘```

GDPR Data Deletion

One-click

Deletion

All tables

Scope

30 days

Timeline

Transcripts, insights, labels, usage, account

Coverage

One-click deletion removes all your data from our servers. Cascades across all tables. Within 30 days, everything is permanently gone.

Privacy Mode

🔒

When Privacy Mode is enabled, dijin operates with zero network egress. No analytics, no sync, no cloud. Everything local. Available to Free and Pro users.

We built dijin this way because your conversations are among the most sensitive data you produce. They deserve infrastructure that treats privacy as a requirement, not a feature.