The biggest LLMnesia update yet — Memory, and search by meaning
0.2.0 is the largest release since launch. Two headline features change how you work across assistants: a portable Memory that carries your context into any AI, and semantic search that finds chats by meaning rather than exact words. Around them sit summarise-and-continue, starred chats, a built-in shortcuts sheet, and a wave of capture fixes — all still running entirely on your device.
Memory — carry your context to any AI
Import what your assistants already know about you, then drop it into any new chat with one tap. Move from a maxed-out ChatGPT to DeepSeek, Kimi, or anywhere else and they'll already know who you are and what you're working on. Each platform's memory is kept separately, and everything stays on your device.
Search by meaning, always on
Search now finds conversations by what they're about, not just the words in them. It builds itself quietly in the background as you use the extension — no button to press, no waiting — resumes across browser restarts, and runs fully on-device.
Summarise a long chat, then continue it elsewhere
A local summariser condenses a 40-turn conversation down to its key points before you carry it into another assistant, so you don't spend the new chat's context on a giant transcript. You can save the summary as reusable Memory too.
Star the chats you keep coming back to
Star any result straight from search, filter to just those with the new “Starred only” switch, and know they're protected from cleanup so they're never removed to make room. (This replaces the old, hard-to-find Pinned list.)
A keyboard shortcuts cheat sheet, built in
Press ? in the search box to see every shortcut at a glance — open the palette, move through results, add your context, open a result in a background tab — without ever leaving your chat.
Also in this release
- One-click Import from Grok on the welcome screen, alongside ChatGPT, Claude, Gemini, and Perplexity.
- A redesigned, roughly half-height filter panel, plus empty states that tell you when a filter — not a lack of matches — is hiding your results, with one-click Clear filters.
- Simpler Memory controls (a single Off · On · Auto switch per memory) and a cleaner Settings page with colour-coded sections.
- Reliability fixes across the board: Claude and Perplexity answers are captured again after their layouts changed, “Continue in…” now carries the whole chat (including the opening that frames everything) with formatting intact, and continuing into DeepSeek works on the first try.