Search Every Conversation You Have Ever Had

An assistant you have used for a year is also an archive of a year of your thinking. MyBenAI indexes every message so that archive stays searchable — on the phone, in the app's own database, with nothing sent anywhere to make it work.

How it works

Messages are written to SQLite and mirrored into an FTS5 full-text index as they arrive. Search is a local query: instant, offline, and unaffected by whether you are online or how many conversations you have accumulated.

Organising rather than hoarding

Conversations can be renamed, and grouped into folders so a project's threads stay together. A folder can also scope which documents a conversation may draw on, which is what stops a question about your thesis being answered out of your tax paperwork.

Search finds messages; memory recalls facts

These are different tools for different jobs and it is worth keeping them straight. Search is for "where did I discuss that" — it finds the conversation. Long-term memory is for "the assistant should know this about me" — a small set of facts you explicitly approved, which the model can recall mid-answer without you having to find anything.

And deleting

Individual messages, whole conversations, or everything at once. Deleting removes the row and its index entry, and because there was never a copy anywhere else, deleted means deleted — see where your data lives.