A private home for family memory

Keep the stories
only you would know.

Send familiar the voice notes, photos, recipes, and everyday details you never want to lose. Ask for them again whenever the moment comes back to you.

One Telegram account. One isolated memory space.Read our privacy policy.

An open family photo album, old photographs, a wristwatch, pressed flower, and voice recorder on a sunlit linen table

Kept closeText · voice · photos

How it works

Memory keeping without a new habit.

There is no family group to configure and no archive to organise. Open the bot one-to-one and share things in the moment.

01

Share it naturally

Send a message, photo, or voice note in your private Telegram chat.

02

familiar keeps the thread

Gemini understands the moment. D1 stores it while Turbopuffer indexes your separate search space.

03

Ask when it matters

Return months later, ask in your own words, and get an answer grounded in what you saved.

Designed around one person

The same question can mean something different to you.

familiar searches only the memory space attached to your Telegram identity. Someone else can ask the same words and get a different answer from what they chose to remember.

Separate D1 recordsSeparate vector namespaceCited memory answers

Built during the hackathon

When the harness did not fit the promise, we built the boundary.

familiar needed Gemini to reason conversationally without letting the model choose whose history it could load, what counted as trusted context, or when a failed reply became durable. Rach Pradhan extracted that boundary into standardharness, a small model-agnostic npm package used around familiar's Gemini loop.

  • 01 Owner checked before context
  • 02 Recent history marked untrusted
  • 03 User turn saved before inference
  • 04 Only a real reply is committed
A person speaks into a bounded conversation chamber; one guarded path reaches a private memory archive and returns a grounded answer
one person · bounded context · one search gate · grounded return

Begin with one story

Your archive already exists.
It sounds like conversation.

Send the first detail you want your future self to find.

Open familiar in Telegram