[BlackForestBoy]

Oliver here from the Memex team.

1. Yes the index is updated every time you visit. It appends the new terms, and keeps all old ones

2. That is a bit more difficult and not as reliable unfortunately. If there is a lazy load on the page it often fails to capture the content, because it starts indexing the page after the initial page load is finished/successful. These are improvements we want to work on a bit later.

3. For about a year's worth of history(~20k pages), without also capturing the screenshots, it needs about 400mb of storage. Indexing performance is still good with 20-25k pages but querying gets slower. So you won't feel the performance on your system with a reasonably fast computer (recommended 8gb of ram, and a dual core with at least 2GHz) We are about to work on performance improvements to make it fast and scalable much beyond that amount and with less resources.

4. Unfortunately we didn't get around to optimise the indexing for CKJ characters, but all latin characters should work fine.

[ComodoHacker]

Thank you.

On 2, I mean these (feeds, topic lists etc.) are probably not worth indexing at all, esp. since you keep all old content in the index.

Is your code all home-built, or you're using some FT engine compiled to Wasm?

[BlackForestBoy]

> On 2, I mean these (feeds, topic lists etc.) are probably not worth indexing at all, esp. since you keep all old content in the index.

Yeah indeed. A better implementation for now would be to let people save single posts, or sync with likes, shares, tweets and retweets, so people can search with those facettes.

> Is your code all home-built, or you're using some FT engine compiled to Wasm?

For the search we are using dexie.js, the rest is home-built. Our storage engine is: https://github.com/worldbrain/storex