Gzuuus

Yes, what I mean is the where and how of the implementation. I'll share my thoughts on how I was planning to do it and how it could work for this use case. My perspective shifted when I created the transparent OAI proxy https://github.com/gzuuus/transparent-oai-api . With that proxy, I could connect to PPQ, intercept all requests and responses, mainly for debugging purposes. However, that concept could also be used to integrate a cashu wallet into the proxy, enabling automatic payments when necessary. A local piece of software that expose the same API of the provider, acting as a smart proxy to manage payments and top-ups using NWC. This way, the user centralizes their connection with providers and interacts with them using the localhost endpoint.
I have more ideas about this, but simply the system is already pluggable to Goose. It would be as easy as configuring the endpoint in its configuration. Its also more portable. You can also write it in any language

2025-05-31 12:19:32 UTC

- reply

For you too 🍸

2025-05-31 12:18:27 UTC

- reply

There is a party in cyberspace

2025-05-31 12:17:47 UTC

- reply

Thats imminent. I want to also structure the packages better to use it as a library. The commons pkg is about as that but could be even more useful.

2025-05-31 10:17:12 UTC

GM nostr! We are celebrating today 🧡

2025-05-31 08:44:28 UTC

- reply

Hay works good right? 👀🔥🔥
How was the experience of implementing it?

2025-05-30 18:32:45 UTC

- reply

Signallamallamadingdong

2025-05-29 07:08:21 UTC

Gf*M 🌞🔥

2025-05-27 14:18:53 UTC

- reply

Why? Is an android limitation or something?

2025-05-27 14:10:37 UTC

- reply

You mean from keeping the subs open?

2025-05-27 10:53:56 UTC

- reply

Was looking forward for these sub features 🔥🔥

2025-05-24 15:56:29 UTC

- reply

Portability and anonymity I think are the most obvious improvements

2025-05-22 06:13:26 UTC

Woke up, pizza dreams, pizza breakfast, pizza day 🍕 GM

2025-05-20 15:05:07 UTC

- reply

Yes, I think so, Zen is a Firefox fork

2025-05-19 20:32:54 UTC

Stay focused, don't get distracted. There are loads of people/entities competing for your attention. Don't give it away for free; use it instead. Stay focused 🧘

2025-05-15 08:16:33 UTC

- reply

I mean to eat, I saw them flying around.

2025-05-14 17:56:44 UTC

- reply

And poncha 🧡
#nevent1q…9j7f

2025-05-10 11:40:42 UTC

- reply

I think coinos is down 🙏

2025-05-10 11:21:27 UTC

- reply

Gm cameri 🖖

2025-05-09 09:01:28 UTC

GmGm

2025-05-07 14:27:23 UTC

I had to touch a windows laptop today. I still feeling dirty

2025-05-04 09:29:17 UTC

- reply

5300 is a DVM kind, when you interact with an algo feed DVM you follow the same pattern of request response than with other dvms, if it behaves like a DVM is a DVM xD and yes you can solve the same use case in different ways

2025-05-03 17:11:33 UTC

- reply

I would have to study a bit more to give you an answer to that XD

2025-05-03 12:44:50 UTC

- reply

Not really afaik, but shouldn't be pretty difficult

2025-05-02 14:30:50 UTC

- reply

A bot can have any kind of arbitrary behavior, such as listening for and responding to kind 1 notes, or listening for and responding to any other type of event in they way they want. DVMS describes and tries to standardize how this interaction is done by defining requests, responses, and job feedback. Of course, you could do the same thing and call it a bot, but that is not the point, bots can use dvms to solve some of their features.
The differences between an http enpoint and a dvm is nostr

2025-05-02 13:59:42 UTC

- reply

DVMs describes a pattern for clients to interact with some computation or functionality, and bots are some sort of systems that have a defined behaviour and do stuff

2025-05-02 13:44:38 UTC

Due to the recent and ongoing discussions about DVMs and the @nprofile…dlnm intention to reset it, I wrote a proposal rewriting the nip, This is a starting point and, hopefully, it will make sense to others. If you are interested in DVM stuff, please read it and provide feedback https://github.com/nostr-protocol/nips/pull/1904

2025-05-02 13:39:25 UTC

- reply

Totally, what about you?

2025-05-02 11:24:17 UTC

- reply

They are different stuff, cannot compare oranges with apples

2025-05-02 09:14:35 UTC

- reply

Loving LNVPS 🧡
#nevent1q…duak

2025-05-02 08:53:56 UTC

- reply

Yes, I was thinking about that thoughtfully when drafting dvmcp spec and my article about dvms and the reflection pattern.

2025-05-02 08:52:32 UTC

- reply

Yea, there are some valid points there

2025-05-01 17:02:23 UTC

- reply

I would like to hear constructive feedback from you. That afirmation is for my recent article about dvms? I agree there is loads of room for improvement

2025-05-01 15:33:52 UTC

- reply

U're a troll, sir

2025-05-01 15:32:21 UTC

- reply

The promise of job chaining and result reuse never materialised

2025-05-01 13:58:17 UTC

- reply

Totally, I moved to ephemeral events for dvmcp

2025-05-01 12:30:01 UTC

- reply

I think we are not yet aware of how powerful this is. It allows a social mechanism to keep open source projects running, thanks to people zapping the machines where they are hosted. This is how we win
#nevent1q…pyct

2025-04-30 12:50:57 UTC

- reply

CC @nprofile…vu4p

2025-04-30 12:41:07 UTC

- reply

There is a point here xD
#nevent1q…8qu0

2025-04-30 12:38:16 UTC

The new DVMCP specification is nearly finalized after some review ( thanks @npub1ye5…knpr ). I won't rush the merge, so there may be more feedback. However, if there isn't any additional feedback, I will merge this tomorrow and begin refactoring the DVMCP packages to implement the updated specification. LG!🚀
https://github.com/gzuuus/dvmcp/blob/docs/spec-rev/docs/dvmcp-spec-2025-03-26.md

2025-04-30 08:45:27 UTC

- reply

Making them pay more

2025-04-29 22:12:57 UTC

- reply

You don't have to support everything specially for specific client APIs

2025-04-29 22:09:34 UTC

Follow packs yeah! At least! people have been using them in a primitive way since nip51 was a mess, it was time they got more attention 💪

2025-04-29 17:36:06 UTC

- reply

To be honest, that sounds more like a specific client UX issue than a feature for a decentralized computing framework. However, there is a point in what you are saying. If there is a defined list of kinds for job type, it might make sense to specify a minimal required input schema for each job type. This way, all DVMs associated with a job type should have that required input parameters. Client developers who want to rely on any DVM out there to solve a specific functionality for their client can understand that there is a specific required parameter that must be fulfilled for the DVM to work. For example, in a translation DVM, the required inputs would be the text and the target language. And you can solve that in your ui/ux as you prefer.
Each DVM might also have optional parameters to enhance or provide additional functionality, and the most effective way to manage this is through the reflection pattern. Currently, how can you implement any translation DVM or any other DVM? It is not possible because the operational details of the DVM are unclear. As a result, client developers can only hardcode specific DVMs that they understand and can work around.
In any case, I believe that reflection is the way forward for DVMs, and even if there is a specification for required parameters for each job type, it must be clearer and more lightweight than it is now

2025-04-29 15:48:52 UTC

- reply

Maaaan! Congrats! ❤️‍🔥❤️‍🔥❤️‍🔥

2025-04-29 14:14:28 UTC

I am currently testing Qwen3 (1B and 4B) locally with Ollama. It is the small model I was looking for to call MCP tools. While it still a bit dumb in its responses, it successfully calls the tools perfectly. No other model of this size can call tools as effectively.

2025-04-29 10:05:27 UTC

- reply

2025-04-29 10:05:04 UTC

- reply

I think this is quite possible, I already do this in dvmcp and dvmcp.fun as you have defined data types and everything is a compliant json schema where the inputs are defined, you can easily build a form with some defined input methods for each data type, string is just a text input, boolean is just a checkbox, etc. There are already libs that can help you with this and do json schema to form

2025-04-29 07:09:51 UTC

What happened in Spain and Portugal yesterday was pre-apocalyptic. People really understood how fragile our comfort, communications, and security is. Hopefully they will keep that in mind and start rebuilding. GM🌞

2025-04-28 07:33:40 UTC

GMGMGoodMonday 🌞

2025-04-28 07:33:01 UTC

- reply

Not if you copy it 🫟 (writing from an android) #nevent1q…7s9a

2025-04-28 07:01:31 UTC

- reply

Agreed

2025-04-26 17:41:15 UTC

- reply

Hmm... I think the most logical flow would be per kind, reusing the req/res kinds of each job type, so in this way you will request the input schema from a dvm using its 5xxx req kind and it will respond in its 6xxx, the request can contain a standardized method to request the input schema so clients will know how to request this schema on any dvm. This is also more private than creating a new kind just for req res input schema, since this will look like a normal operation.
And yes, the dvm can choose not to announce itself publicly, so no nip89, this is something I am also introducing for the revised dvmcp spec.
wdyt? Someone should draft this to potentially upgrade nip90

2025-04-26 10:57:10 UTC

- reply

BTW, I am not a mathematician XD

2025-04-26 09:53:45 UTC

- reply

Please @nprofile…xnfk @nprofile…6rpa @nprofile…7n8d can you validate this? 🙏

2025-04-26 09:52:44 UTC

- reply

Thanks to @nprofile…4umc as we discovered this together

2025-04-26 09:52:03 UTC

The GFY framework
Abstract:
In this note, we explore a whimsical yet insightful relationship between two vectors representing opposing life forces: Good For You (GFY1) and Go Fuck Yourself (GFY2). We formalize the idea that the more something is good for you, the less go fuck yourself is, and vice versa.
Introduction
Life often presents us with choices and experiences that can be broadly categorized as either beneficial or detrimental. To capture this duality mathematically, we define two vectors in an abstract vector space:
GFY1: The Good For You vector, representing positive influences, benefits, and well-being.
GFY2: The Go Fuck Yourself vector, representing negative influences, detriments, and frustration.
Defining the Relationship
Intuitively, these two vectors are inversely related, an increase in one corresponds to a decrease in the other. We express this relationship as:
GFY2 = -GFY1
And consequently:
GFY1 + GFY2 = 0
These equations state that the Go Fuck Yourself vector is the additive inverse of the Good For You vector, and their sum is always zero. This is a profound insight for your life.
Interpretation
When GFY1 points strongly in a positive direction (high benefit), GFY2 points equally strongly in the negative direction (low detriment). Conversely, if GFY2 increases (more frustration), GFY1 decreases (less benefit). The sum of these opposing forces is always zero: GFY1 + GFY2 = 0, which symbolizes the balance (or tension) between good and bad.
This apparently simple equation offers a powerful framework for evaluating life choices. When faced with a decision, we cannot truly assess its GFY1 value (how beneficial it is) without simultaneously contemplating its GFY2 value (how frustrating or detrimental it is), as they inherently influence and define each other. To understand how good something is for you requires understanding how much "go fuck yourself" is contained within it, and vice versa. This interdependent relationship helps us make more holistic assessments, allowing us to prioritize choices that maximize GFY1 while naturally minimizing GFY2 in our lives.
Extensions and Applications
This model can be applied to real-life decision-making, where each choice or event can be represented as a vector in the GFY space. By consciously measuring the magnitude and direction of these vectors, we can make more informed decisions. The GFY framework provides a simple yet powerful tool for navigating complex situations, helping us distinguish between what truly serves our interests and what merely frustrates our progress.
Practical applications might include evaluating job opportunities, relationships, or daily habits. By assessing whether something increases your GFY1 or GFY2 vector, you gain clarity about its true value in your life. This mindful approach to decision-making can lead to greater overall satisfaction and reduced stress.
The next time you face a dilemma, just remember: The more GFY, the less GFY, and vice versa. This perspective can guide you toward healthier choices and a more balanced, fulfilling life.

2025-04-26 09:13:26 UTC

- reply

I think I'll write a note expanding on this, it might be something important
#nevent1q…hmrk

2025-04-26 08:37:15 UTC

- reply

It's a mathematical expression I'm working on, indeed is more like GFY₂=-GFY₁ where GFY₁ is the *Good For You* vector, and GFY₂ is the *Go F*** Yourself* and GFY₁+GFY₂ = 0

2025-04-26 07:59:25 UTC

- reply

Did you know that GFY != GFY?

2025-04-26 07:56:56 UTC

- reply

GM its a good day to read specs and provide feedback 🤙
#nevent1q…uyrg

2025-04-26 07:29:21 UTC

- reply

That's why #soveng @nprofile…3ra5
#nevent1q…e30c

2025-04-25 12:21:25 UTC

- reply

If you are interested in dvmcp read this, we are reviewing the spec and looking for feedback
#nevent1q…k7kk

2025-04-25 06:24:34 UTC

- reply

Fried brains with beans

2025-04-25 06:13:13 UTC

- reply

Aaa I wish I could assist, sounds super fun

2025-04-24 15:35:21 UTC

- reply

Just wrote this article based in what I'm learning from building dvmcp and the ongoing discussion about dvms
#naddr1qq…fgn4

2025-04-24 07:20:08 UTC

- reply

Wow, the David Lynch tribute page is lovely ❤️‍🔥

2025-04-24 06:30:18 UTC

- reply

Ask if you have any doubt 👌

2025-04-23 21:40:45 UTC

- reply

Yes, for this you can use the nostr beating heart mcp, it has a rag db behind it and can also search for code snippets. You can use dvmcp to use my instance without having to run it locally, it has all nostrbook.dev and nips repo embedded. It will provide all the context to cline about nostr protocol related stuff

2025-04-23 21:36:07 UTC

- reply

Looking forward to catch up also with the new url sharing kinds for nostree 🫡

2025-04-23 21:30:26 UTC

- reply

#nevent1q…56ka

2025-04-23 21:30:03 UTC

- reply

Yes, I would recommend visiting https://modelcontextprotocol.com to address some initial questions. After that, you can choose one of the sdks. If you prefer a different programming language that isn’t available, you can check gh for libraries in other languages. You can also create your own implementation based on the base protocol... if you want xD Once you pick your poison, writing a server is quite straightforward, for example, I wrote this some time ago in ts https://github.com/gzuuus/mcp-echo/blob/master/src/mcp-server.ts You can also choose from one of the mcp servers available on mcp.so or other repositories. Then if you want to test it out, use the inspector https://modelcontextprotocol.io/docs/tools/inspector which will allow you to call tools and debug the server. After that, you can integrate it with an llm if you wish. Additionally, there is a workshop that utilizes also the dvmcp packages https://dvmcp.fun/blog/dvmcp-workshop

2025-04-23 13:41:58 UTC

- reply

Have you guys seen the movie Snowpiercer?
#nevent1q…9p6l

2025-04-23 11:53:37 UTC

- reply

Yes 👌

2025-04-23 11:45:08 UTC

- reply

I like goose

2025-04-23 10:54:24 UTC

Showtime!🔥
I've been busy working on the @npub1dvm…3jdm stuff, focusing on the spec and packages, and I forgot to show some of the servers I'm already running that are free and open to everyone, these are like apps you can plug into your llm or agents to give them new capabilities. Of course, there's no guarantee of uptime, but they've been running well for a few weeks now with no noticeable downtime. I wanted to put together some useful tools that the majority of people might find useful.
- DuckDuckGo Search
https://dvmcp.fun/dvm/dvm-announcement-duckduckgo
his is an MCP server that you can use to perform searches with DuckDuckGo and retrieve the results. This is useful for improving your search experience by collecting real-time data, performing research, and anything that involves fetching fresh data from the Internet and making it available to your llm.
- Nostr Beating Heart
https://dvmcp.fun/dvm/dvm-announcement-nostr-beating-heart
This MCP server is something I'm experimenting with. Behind it is a RAG DB that contains all the knowledge about the Nostr protocol, from the NIPs repo to all the content from nostrbook.dev (cc: @npub1q3s…d26p). The experimental part involves providing pluggable RAG DBs into your LLMs, allowing you to have these kind of buckets of context to retrieve accurate and concise data from a knowledge base, which is especially good for avoiding bloating the context window of your LLM. It has been working pretty well since nostrbook.dev was indexed. It also includes a tool to search code snippets, making this MCP server quite effective for quickly retrieving data about the protocol, NIPs, kinds, code examples etc.
- Nostr Daily News
https://dvmcp.fun/dvm/dvm-announcement-nostr-daily-news
This is another experimental MCP server where you can find tools to retrieve data from configured sources, including Nostr and RSS. The cool part is that it is collaborative, meaning you can add some RSS sources to Nostr relays, could be like a colaborative bookmark of news sources. It also has some preconfigured relays to fetch news from Nostr, Stacker News, Hacker News, and custom sources like The Rage, TechCrunch, Hugging Face Blog, and more. You can simply prompt your LLM with "give me an overview of the news from all your sources," and it will gather the latest posts from each source, providing you with a comprehensive overview, pretty cool for a morning news digest.
- Nostr MCP Server
https://dvmcp.fun/dvm/dvm-announcement-nostr-mcp-server
This is the MCP server from @npub1s9e…lxzl. It has some interesting tools for using Nostr, including getting profiles, retrieving the latest notes from someone, zaps, long-form content, searching the NIPs repo, and even sending anonymous zaps, very powerful.
---
These are the tools I'm running, but there are also other interesting tools that others are running, which you can plug directly into your LLM or simply visit dvmcp.fun to use them without needing an LLM. If you want to use this with your LLM, I recommend using Goose, but you can use any other client with MCP capabilities. Once you have it set up, using these tools is as easy as configuring the commands or settings you can find in the 'Install' tab of each tool on dvmcp.fun. For example, by setting this command as the MCP server:
`npx @dvmcp/discovery --server #naddr1qv…f7mt`, you'll be plugging into the Nostr Beating Heart. You don't have to run the whole RAG system locally; you can leverage the instance that I'm running on a cheap $2 VPS from @npub1lnv…7vjj
Hope you find this useful! Let's keep building! If you have any doubts, do not hesitate to ask :)

2025-04-22 14:47:51 UTC

- reply

Love

2025-04-22 14:47:30 UTC

- reply

Just saw it today, thinking in spin it up as dvmcp. Also https://dvmcp.fun/dvm/dvm-announcement-nostr-beating-heart goes in the same direction

2025-04-22 08:13:36 UTC

- reply

Fuk, this one looks even worse🙃
#nevent1q…muh3

2025-04-21 16:45:28 UTC

- reply

https://github.com/nickfarrow/blind-sig-demo

2025-04-21 15:51:52 UTC

- reply

Tool calling are just structured outputs, but seems that small models sometimes struggle to follow the structure, or get confused. I think this will be solved in the tiny models of the future because standards and over feeding

2025-04-21 15:33:23 UTC

- reply

Men that's hot ❤️‍🔥

2025-04-21 14:59:23 UTC

- reply

You mean the new qat models? I'll be running and testing 12b asap. It's the biggest one my laptop can afford, no racks for this pleb yet 🙃

2025-04-19 14:17:45 UTC

- reply

For the curious cats, this was at the beginning #internetisforcats #catsareforinternet
https://video.nostr.build/a94f00782dad853b6501eca62fcc25b2faae2c2973361d8479d25b0533ee807e.mp4
#nevent1q…mphx

2025-04-19 08:23:36 UTC

GM!🌞🐈 new tiny creature have appeared, just born ~5 w ago
https://image.nostr.build/e31bc17d1fd8dd134ac2b494f72b7229417a94326fed04c05e59913c328ed6c9.jpg

2025-04-16 07:36:29 UTC

GM 🌞

2025-04-11 11:03:27 UTC

- reply

Goose + dvmcp discovery

2025-04-11 07:27:40 UTC

- reply

Nostr will have the best UX in the future, others will just have an authentication hell. GM🌞
#nevent1q…h0f5

2025-04-10 09:10:51 UTC

We are so early for some things and so late for others

2025-04-02 16:44:04 UTC

Random Access Melancholy

2025-04-01 21:21:46 UTC

- reply

I agree, I'm just saying that things are changing, and now it's like you can have some sort of chef next to you teaching you how to do the soufflé, I've never said that you shouldn't touch any code 🤷 anyways just opinions about what I can observe around

2025-04-01 09:45:57 UTC

GM I'm going to leave nostr and bitcoin and get a fiat job, then get a mortgage and die happy as a normie, it was good to meet you all 🫡

2025-04-01 09:40:14 UTC

- reply

I've been thinking lately that learning how to code is not the right approach anymore, now when a friend asks me how to learn development and code I give more importance to learning software and system design, architecture principles and so on. Code is the mechanical part of software development, wdyt?

2025-03-30 08:33:10 UTC

GSM 🌞

2025-03-26 08:50:50 UTC

- reply

Gut Monster xD