Memory Logger solves a friction problem: during a trip, you never take the time to write. The journal stays empty. Memories fade. Memory Logger reduces friction to zero — a voice or text message is all it takes, the rest is automatic.
How it works
You send a message to a Telegram bot — voice, photo, or text — from anywhere. The n8n workflow takes over:
- Transcription — Voice messages go through OpenAI Whisper
- Structuring — GPT-4o identifies the location, date, people mentioned, emotions
- Writing — GPT-4o generates a first-person narrative with highlights from the moment
- Archiving — Everything is pushed to Notion: journal entry, locations database, trip summary
By the end of your stay, you have a complete journal without having written a single line yourself.
Architecture
Telegram Bot — Single entry point
- Accepts text, audio, and photos
- No commands to learn — send as if you were texting a friend
n8n — Central orchestrator
- Workflow triggered on every Telegram message
- Branches by message type (voice/text/photo)
- Sequential calls to Whisper then GPT-4o
- Writes to Notion via API
Notion — Memory database
- One page per trip (created automatically)
- One entry per message with date, location, structured text
- Auto-generated trip summary at the end of the stay
Why n8n and not custom code?
n8n lets you modify the workflow visually without touching code. Adding a new source (WhatsApp, email), switching the GPT model, adjusting the prompt — all done in minutes. For a personal tool that evolves, this is a pragmatic decision.
Key features
- Multimodal capture: voice, text, photo
- Automatic voice transcription (Whisper)
- Structuring and writing by GPT-4o
- Organized archiving in Notion (by trip, by date)
- Automatic trip summary at the end of the stay
- Zero friction: no app to open, just Telegram
Built by Thomas Silliard — in use since January 2026 during every trip.