Import Your History Twice. Nothing Duplicates.

There's a small hesitation that stops people from importing their ChatGPT history: “What if I import it now, and then want to import again later — will everything double?”

It's a fair worry, and with most tools it's justified. With Engram it isn't. You can re-import your export as many times as you like and nothing duplicates. Here's exactly how that works — including the one honest limitation, because you should know the whole truth before you trust a tool with your memory.

The fear is reasonable — most tools do double

Plenty of import tools treat every upload as brand-new data. Import a file today, import the same file next week, and you've got two copies of everything — bloated storage, duplicated search results, a mess to clean up. People have been burned by this, so they import once, nervously, and never touch it again. That's the instinct we wanted to remove.

Engram deduplicates by conversation

Every conversation in a ChatGPT or Claude export carries a stable id — the same id every time you export. When you import, Engram records a fingerprint from that id (chatgpt:<id>). On any later import, before it stores a conversation it checks: have I already seen this fingerprint?

• Already imported → it skips the conversation entirely. Nothing is written, no duplicate is created, no storage is used.
• New conversation → it imports normally.

So a re-import of the same file finishes with a summary like “0 imported, 128 skipped (already imported).” Your memory is exactly as it was — not one message heavier.

What this means in practice

The useful consequence: you can treat import as a habit, not a one-shot.

• Re-upload the same exportanytime — to double-check it worked, or because you forgot you'd done it. Harmless.
• Export again months laterand re-import. Every conversation you've had sincethe last import gets added; everything before is skipped. This is the intended “top up my history” flow — your Engram memory stays current with your ChatGPT usage, no manual bookkeeping.
• Hit the free ceiling, upgrade, re-import— the conversations that already fit are skipped, and only the remainder that didn't fit before gets added. You're never re-charged storage for what's already there.

The one honest limitation

Deduplication matches at the conversationlevel, not the individual message. That has one edge worth knowing: if you go back into a conversation you've already imported, add a few new messages to it in ChatGPT, and re-export — those new messages inside the old conversationwon't be picked up. Engram sees the conversation id already exists and skips the whole thing.

We chose that deliberately. The alternative — trying to merge new messages into existing conversations — risks duplication and reordering, and we'd rather never duplicate than maybecatch every mid-conversation edit. New conversations always sync; new messages appended to old ones don't. In practice people re-import to pull in new conversations, which works perfectly — and for capturing things as they happen, connecting Engram live to your AI tools handles the going-forward side without any export at all.

Import as often as you like

The short version: re-importing is safe. Engram deduplicates by conversation, so your history never doubles no matter how many times you upload. Import once to get set up, re-import whenever you want to stay current, and stop worrying about it.