AI Didn't Remember What I Said

1. Confirm It Really Was Written to Long-Term Memory

• Panel → Memory, search for the keyword.
• If you can't find it, the AI didn't store it in memory — it remembered it for this chat only.
• Fix: tell the AI "please save XXX to long-term memory".

2. Manual Is Most Reliable

If you don't trust the AI's auto-judgement, add it yourself in Memory → New:

• Title — short keyword, e.g. "Company Address".
• Content — full fact, e.g. "HQ is at 100 Songren Rd., Xinyi District, Taipei".

Manual memories load first at the start of every chat.

3. Memory Limit Reached

• Starter: 50, Pro: 500, Enterprise: unlimited.
• At the limit, new memories are silently dropped (no error, but no effect).
• Sort Memory by "least used" and clean up, or upgrade.

4. Wrong Workspace

Long-term memory is bound to the workspace, not the account.

• What you taught in workspace A, workspace B doesn't know.
• If the AI feels "dumber" after switching, check you're in the right workspace first.

5. Use Feedback to Strengthen

• 👍 after a reply that used a memory makes it more likely to be reused.
• Repeated 👎 nudges the AI to stop using that memory.
• See Giving Feedback.

6. Outdated Info Overrides New

• If a memory says "I work at A" but you've moved to B, the AI prefers newer chat context but the old memory is still there and causes confusion.
• Edit or delete stale entries in Memory.