Closed MikeHuntington closed 3 hours ago
No that’s correct - the engine only looks back as far as the firehose backfill goes (3 days) on first launch - but will store history going forward (up to a limit for some feeds, specified in https://github.com/Bossett/bsky-feeds/blob/ae5dae7d38bf816b897c071e1c0af8702718bb6a/src/algos/cats.ts#L47 e.g.)
@Bossett thanks so much for the answer, that's makes sense.
However, after the backfill is read the first time, Is there a reason the feed still only updates only showing posts from 3 days ago even after a few hours of being published?
Here's an example of the feed I published: Cats Feed
Oh Wait, I think I understand, you mean it will just take some time for all of the feeds (from 3 days ago) to reach present posts?
yes that's right - just make sure that it's able to write to the sub_state collection (https://github.com/Bossett/bsky-feeds/blob/ae5dae7d38bf816b897c071e1c0af8702718bb6a/src/db/dbClient.ts#L163) which lets it keep track of where it's up to (you can use mongo's compass tool to check & create if necessary)
Awesome, I think that solves my issue, but one last observation... with the delay it takes for the posts to be processed for the feed, it seems like it would never catch up to current posts.
Is there a way for me to make that processing happen a bit faster? (As you are probably aware, traffic on BSky is high and the number of posts coming through is crazy high)
It'll move as fast as your bandwidth/cpu allows and it skips all non-posts so it will generally be 'fast enough'. It is inefficient though since it was built before tools like jetstream (https://github.com/bluesky-social/jetstream) were available.
I run it in production now though - and it is keeping up at ~25mbit - noting that right now there a big US east coast outage unrelated to bsky that's breaking things
Your help is much appreciated! Thank you!
When running the sample algos like "auspol" or "cats", The feeds only show posts from 3 days ago. Is there a configuration that I could be setting incorrectly to cause that?