pippellia
Pip the social graph guy

One can easily use #rely to:
- maintain a whitelist of npubs
- ask the client to auth when connecting/when posting/ when requesting chats...
so yeah, basically a community, right?

2025-04-07 21:56:03

- reply

teach me this power master jedi

2025-04-07 07:39:01

- reply

well, I found many race conditions, and one weird bug where many clients unregistering all at once caused the blocking of all threads

2025-04-07 07:37:24

- reply

yes. random requests, random events, random filters, random disconnects...
I've found so many weird bugs, like the unregistration queue filling up causing all threads to be blocked

2025-04-06 22:29:01

- reply

noice. Btw recently I've learnt about random testing and I'm pretty proud of this
https://github.com/pippellia-btc/rely/tree/main/tests

2025-04-06 22:26:20

- reply

hey guys, maybe you can improve your processes by using Vertex and npub.world

2025-04-06 21:52:14

Search and pay is the ultimate paymente UX.
#nevent1q…pxfv

2025-04-04 21:29:35

- reply

yeah go is amazing, and one can really appreciate how much thought went into it by trying to build stuff and realised bit by bit why things are the way they are

2025-04-04 19:23:30

- reply

mmm, you gave me some things to think about. Thanks!

2025-04-04 18:38:36

- reply

yestr, collaboration and pollination of ideas is how we win

2025-04-04 16:11:07

- reply

I think you'll find it very simple :)
Let me know any feedback sir

2025-04-04 16:01:00

- reply

@nprofile…fngn I think you are conflating khatru and rely. When I say that rely took inspiration from khatru, I mean that I copied the ease of use (users supply their rely functions), but the inner working is completely different. completely.

2025-04-04 15:58:49

- reply

> i don't even understand how khatru based relays are still functioning the excessive amount of threading they use
yeah I know, I've been running Vertex on khatru for few months. Rely uses only 2 go-routines per websocket, one for reading, and one for writing. Then, one goroutine for dealing with client registration, unregistration and requests.
> also, yeah, not to forget, i totally flattened the concurrency of the event store, so, it also is a solution for that, excessive amounts of threads in each query were a component of the resource leak problem i found with khatru/eventstore design
Oh yeah, that part is gone in my design. Users supply their query method and that's it.
In fact, all your cool DB stuff can be used in the rely.

2025-04-04 15:10:03

- reply

thanks! Well, for now REQs have multiple filters, so I'll keep it the way it is until the PR gets merged

2025-04-04 15:02:49

- reply

> relayer has better architecture and realy was forked from that base
can I ask why it's better?
> i don't like go-nostr at all, though i'm not saying that realy's libraries are exactly perfect
Yeah, I feel similarly about go-nostr. I am considering switching. Why don't you extract the library from realy, making its own repo? Other people might use it in that case
I'll take a closer look at realy sir. The first time I was just scared away by how big the repo was.

2025-04-04 14:33:54

- reply

@nprofile…fngn would love to know what you think

2025-04-04 14:25:07

💫 Introducing rely
A framework for building super custom relays you can rely on. Designed to be simple and stable. About 1000 lines of code.
https://m.primal.net/QECM.mp4

2025-04-01 08:58:34

- reply

yes exactly, but the community already exists in the social graph. 99% of people will be invited to communities, won't be their founders. Founders will need to acquire reputation themselves, but these are the type of people that aren't afraid of doing the work.

2025-04-01 08:33:42

Try using npub.world. 99% spammers free
#nevent1q…lv2m

2025-03-30 10:14:16

others interested? Might code it up
#nevent1q…du3c

2025-03-30 09:45:23

- reply

grazie Daniele :)

2025-03-30 09:44:42

- reply

no idea, sorry

2025-03-30 09:30:10

- reply

We don't discriminat non-human npubs 🤪
If people follow it, then they probably find it valuable

2025-03-29 08:23:29

- reply

pretty brutal yes. There are ~300k reputable pubkeys in total

2025-03-28 18:45:42

- reply

If you are interested @nprofile…5czt I can point to you where that is in the code

2025-03-28 18:44:57

- reply

This algorithm.
https://en.wikipedia.org/wiki/PageRank
We use a slight variation but it's basically identical.

2025-03-28 18:44:08

- reply

An interesting idea to improve it is to add recommendation DVMs (like this one https://vertexlab.io/docs/nips/search-profiles-dvm/), so people can start with a high-quality feed.

2025-03-28 18:38:10

- reply

Hey, thanks miljian!
For context I'll add here a response I just wrote. It's a bit convoluted I know, I should probably make a video explainer.
nevent1qvzqqqqqqypzpa5rapcrtaadfazwpwvvl0v4xlskg4df9nfcem7yevcaka2h7hhjqyt8wumn8ghj7mtvv44h2tnjv4skc7fwd3hkctcpzamhxue69uhhyetvv9ujumn0wd68ytnzv9hxgtcqyqscjdyme00uaezmgs3d4cg6p2eezfj365xm8qqeuy7gfvkm9jhcyupgfpd
> how exactly does somebody make the cut of being reputable? It has to be followed by at least one reputable
> how many real humans do you think you are missing in your reputable follower counts?
No idea tbh, but these users haven't gained a single reputable follower, which probably means they haven't posted anything / met anyone in meatspace that is already on Nostr.
One can find these people by their npubs, but not using the search DVM. This is a very opinionated decision that I think will make more and more sense in the future: I choose to crawl what people find interesting/valuable. and nothing else.

2025-03-28 18:22:07

- reply

> What’s a reputable follow?
great question!
some definitions related to our graph database:
1. a reputable node is one that passed our internal filters, and thus was added to the graph (point 4.)
2. an active node is one which contributes to the global Pagerank
3. a node get's promoted from inactive to active if it's Pagerank score surpasses a certain threshold.
4. to be added to the graph as an inactive node, one has to be followed by at least one active node
At the moment there are about 100k active nodes, and 300k reputable nodes.

2025-03-28 08:41:58

- reply

Looking forward to seeing it.

2025-03-28 08:13:38

- reply

> there is no point in doing sliding window spidering on events once you have the history, from then on just open a subscriptiion
I am not following sorry. My crawler does this:
- Firehose: subscribes for new kind3s and kind0s
When a pubkey that is in the database is "promoted" (has acquired enough reputation)
- Query: query directly for its own kind3s and kind0s.
The query starts when one of these two conditions is met:
- One minute has passed since the last (with at least one pubkey)
- There are 100 pubkeys that have been promoted
Normally the crawler in production promoted ~100 pubkeys/day.
However, when I do tests locally I re-do the entire crawl which obviously is much more intensive, but I'll make sure I'm not using your relay for that sir.

2025-03-26 15:03:06

Nostr is home to a lot of spam,
but now there is an effective way to deal with it.
https://m.primal.net/PtrC.mp4

2025-03-21 11:16:37

Follower counts that make sense, live at npub.world 🌍
https://m.primal.net/PniX.mp4

2025-03-14 09:45:46

I must give a massive shout out to @nprofile…pvsl and @nprofile…rp63 for designing and implementing npub.world's new design.
https://m.primal.net/PfSy.mp4

2025-03-13 11:53:25

- reply

yeah, and for my math blog

2025-03-13 11:08:17

- reply

love there is latex

2025-03-11 14:21:00

- reply

it can be improved a lot by not downloading a bunch of events in the first place.
By using DVMs this burden can be moved to service providers, which benefit from better infra and economies of scale.
An example is using Vertex vs downloading 1000 kind:3s to compute WoT things.

2025-03-10 15:15:17

- reply

is that your real voice? If so, how didn't you start a podcast before 😄

2025-03-06 08:37:02

- reply

lol we live in a simulation
https://m.primal.net/PVZy.jpg

2025-02-28 18:53:28

- reply

stunning

2025-02-28 18:52:23

- reply

yeah true. But on the other side, asking the user for approval and making him pay can be taking away from the experience. Maybe clients can pre-pay for new users, or all users, and monetizing somewhere else if they wish so.

2025-02-28 18:08:02

- reply

wooooow. GG guys

2025-02-28 18:07:22

- reply

we can provide the same service in other ways, but a DVM is very straightforward to integrate and it has the benefits of being a native nostr event

2025-02-28 18:06:22

- reply

this is a very interesting question. @nprofile…rp63 and I were so busy building that we haven't though out the payment aspect yet 😅

2025-02-28 18:05:08

- reply

Thanks mate!

2025-02-28 18:02:32

- reply

as Fran pointed our already, we already have a suite of services using our algos in multiple ways!
https://vertexlab.io/docs/nips/
Perhaps we can in the future start indexing and analyzing e-cash mints rating.

2025-02-28 17:59:00

- reply

I think coracle performs search client side, right @nprofile…92wl ?
So comparing it in speed and "breath" would be unfair.
https://m.primal.net/POth.mp4

2025-02-28 16:29:57

- reply

yeah I know, our Firehose is inspired by yours. Not the same because we want to aggressively filter spammers, and we use internal metrics for that.

2025-02-28 16:26:15

- reply

and thank you :)

2025-02-28 16:26:01

- reply

we are doing all ourselves.

2025-02-28 16:20:38

- reply

that's coming soon

2025-02-28 15:37:59

- reply

ahhhh, I'll add it so we can have some more commies indexed 😂

2025-02-28 15:37:21

- reply

ah dude, we have a daily limit of 30 free DVM requests per day. We should probably have a chat on how to extend that for olas users. cc @nprofile…rp63

2025-02-28 15:34:01

- reply

not yet, what's that?

2025-02-28 15:33:02

- reply

you get profiles that contain "cia" https://m.primal.net/POmL.png

2025-02-28 15:32:08

- reply

the source parameter can be omitted, not the search parameter.
Here is a working nak command:
nak event --kind 5315 --tag param="search;jack" wss://relay.vertexlab.io

2025-02-28 15:30:06

- reply

yeah, npubs of all kinds. Even npubs from bridges like @nprofile…jkvk

2025-02-28 15:29:00

- reply

kind 5315
Simple spec documented here:
https://vertexlab.io/docs/nips/search-profiles-dvm/

2025-02-28 15:28:09

- reply

yours is a popular name 😅 . Maybe we should add more rows in the results.

2025-02-28 15:27:24

- reply

most likely it's because we haven't been able to find your profile event. It is rare but can happen. Try changing it, we should detect the new event immediately.

2025-02-28 15:26:18

- reply

search terms should have more than 3 letters, since we are using a trigram tokenizer. We should problably tell this in the UI @nprofile…rp63 .
https://m.primal.net/POlt.png

2025-02-28 15:22:37

- reply

thanks!

2025-02-28 14:04:36

https://m.primal.net/POhO.png
#nevent1q…ts2v

2025-02-28 13:53:07

🌍 npub.world vs Primal showdown.
Who has the best search?
https://m.primal.net/POgV.mp4

2025-02-28 09:02:42

- reply

@npub1zaf…26k5 I was not able to find the bridge nopub of this Jay Graber.
To be indexed, an npub has to:
- have a kind:0
- be followed by a reputable noub that is already indexed

2025-02-28 08:29:01

"Sacrifice is the measure of how ambitious you can be without being a fool."
cit. me, just now

2025-02-18 10:52:18

- reply

> What made you pursue your project?
My passion for math first, then the appreciation for the problem, and finally the fact that it can be so useful for others that they might be willing to pay for it.
> What problem does it solve?
Making WoT analysis simple so apps can focus on building great experiences.
> Where did you hear about this problem? How do you know it exists?
I've experienced this problem myself, and all devs on Nostr have to deal with WoT one way or another. I've talked to many many in NostrRiga.

2025-02-18 10:24:25

The analogy between grants and the war on drugs was pretty unexpected, and yet it has merit. I see too many projects that are solutions in search of a problem.
Hopefully mine is not one of these 🤣
https://fountain.fm/episode/j1HmqwNUqCWANccTRjwo
#nevent1q…rjy6

2025-02-14 11:11:51

With just a dozen lines of code, @npub10r8…t2p8 protect its users from malicious apps.
This is only possible thanks to @npub1kpt…nhmz, which makes adding WoT to your project as simple as signing an event.
https://m.primal.net/Ofac.mp4

2025-02-05 13:15:22

- reply

neither am I. I think in time the protocol matures, but now it's the fun stage of fucking around to find out.

2025-02-04 17:00:02

- reply

oh yeah totally. Like the internet. No one cares about TCP/IP

2025-02-04 13:55:49

- reply

Onchan also has generally better security because keys don't have to be hot.
what do u mean by "bitcoin's relevance will slowly fade"? Lightning is Bitcoin

2025-02-04 10:40:10

- reply

bring back classical architecture

2025-02-01 09:29:37

- reply

I am starting to realise it note1ahy9z6uqmmpfgvx5vvve6d3wda2eryl6cmxd2mraua8dnjj5xensw947pu

2025-02-01 09:23:01

- reply

for most kind1 clients I see the benefit of websockets. But when a single request is needed, I think http is just simpler.
Is it true @npub180c…h6w6 that you "don't like" http? If so, what's the rationale?

2025-02-01 08:47:17

I've started programming seriously only 4 months ago, so there are many things that I still don't understand. Here is one:
Why do we always use websocket to connect to relays?
If I have to post or get one single event, why not a single http?
Why do I have to use http, then upgrade to websocket, then send/receive one event and then close the connection?
#asknostr

2025-01-26 10:52:10

- reply

Let me take a closer look. Which npub are u using ?

2025-01-24 15:55:06

- reply

ofc it does!

2025-01-24 12:50:31

what if the nostr killer app is onlyfans? I mean it might be, as these "models" need censorship resistance

2025-01-23 11:45:28

Is there a service or a test suite for probing a bunch of relays to see which is online, as well as their latency? #asknostr

2025-01-22 09:51:07

- reply

a childish act for sure.

2025-01-22 08:26:30

- reply

> What does PageRank measure?
It measures the probability of reaching a node in a random walk scenario.
The underlying graph determines which "moves" are allowed in this game. In the case of personalized, the walk start from the source node(s).
It's a measure thatclt combines distance (because the walks have a dandling factor alpha < 1), as well as flux, which is the number of path that exists from the source node to the destination node.
It allow to surface and make it easy to understand the underlay data. If that is follows, then what does it surface? Who the hell knows, people follow others for many reasons, work connections, valuable content, and yes popularity. If we are talking about PP, then what you get depends on who you follow, simple as that.

2025-01-21 12:42:06

- reply

> PageRank isn’t the sum of followers. It’s the sum over each follower’s popularity score
Just false. It's the sum of alpha*pagerank(follower)/outdegree(follower), over all followers (plus a constant).
You keep equating Pagerank = popularity as if it's a given, when it's not because, as I said many times, Pagerank on the zaps network can be very different than Pagerank on follows.

2025-01-21 08:14:10

- reply

Personalised Pagerank doesn't measure popularity. I got recommended npubs with less than 500 followers, because they were valued by the people I follow.

2025-01-20 18:24:30

- reply

> How much weight you want to give to each pubkey will depend on the context.
Sure, but then we go back to the data problem. What data do u use? If follows and mutes, you are already narrowing down the context to something loosely related to popularity.
> What about the more complicated case where you want a context specific weight which is proportional to someone’s skill?
I don't see grapevine solving this anytime soon, as I don't see anyone producing high-quality low-ambiguity data about that. Maybe a school can start posting skills-attestations for its students. Maybe. Now no one is doing it, the UX is terrible, and if only a few do it then the rating is completely skewed because there is not enough data.

2025-01-20 18:18:10

- reply

yes I agree that the name is bad. An alternative is "listable". A set is listable if you can enumerate all it's elements in a list (which can be infinite). You can do it for the real numbers, as shown with Cantor's diagonal argument

2025-01-20 17:33:59

- reply

oh boy, this is a nice rabbit hole.
There are countable infinities, means the same cardinality as the natural numbers. Then there are strictly greater infinities, like the real numbers (pi, e, ...).
Aaaaand, given an infinite set, you can always make one that's bigger O.o
https://en.wikipedia.org/wiki/Cantor's_theorem

2025-01-20 17:28:21

- reply

okay, now I understand what you are saying. Still, you can get the same result using a threshold function on PP.
if pp[node] > threshold
then set it to 1.0
or using a logistic function.
That's why the scores aren't that important, what is important is the order. Scores can be rescaled arbitrarily without changing the order.

2025-01-20 17:25:56

- reply

infinity is the cousin of zero, not one

2025-01-20 16:40:01

- reply

> The ratio of K Kardashian’s PageRank score divided by A Einstein’s PageRank score (the example from my prior post) is very high. I suspect it’s unbounded as Einstein’s follower count remains fixed and Kim’s score continues to grow. This is true of global as well as personalized PR.
I am really not getting what you are saying. The ratio of their two GrapeRanks will be unbounded too.

2025-01-20 13:37:20

- reply

> we are all in a workshop now.
we are all in an hackathon imo

2025-01-20 13:13:03

- reply

totally. I've ignored all pubkeys that don't come right after the "p", and I plan to continue to do so unless a solid % of follow-list adopt this new standard.
Why breaking shit like that is beyond me. Add it to the end of the tag, like pet-names that have always been considered optional

2025-01-20 12:57:48

- reply

so they do {"p", "tags", <pubkey>} ?
What the hell

2025-01-20 12:54:38

- reply

what do they put?

2025-01-20 10:50:43

- reply

yes, this decreases the required memory for the graph I am using by a LOT. uint32 vs a long string.

2025-01-20 10:48:58

- reply

what's exactly do I mean by pubkey index?
A map pubkeyID --> pubkey?

2025-01-19 18:27:17

- reply

The average does, but the variance is really high