Cybrex Queries
I have added an additional GPU to our server and decided to keep LLM queries free for the season of the coursework and diploma projects.

Wish you a good luck with your researches. Join us for a smoke break in #communities; it might help you connect with peers.

💬 Chinese #freetalks

We now have a chat group (just talks about Nexus, but not requesting papers!) for Chinese-speaking users, similar to the one for Russian-speaking users.

我们现在有一个针对中文用户的聊天组（不是用于索取论文！），类似于针对俄罗斯用户的聊天组。

欢迎！https://t.me/+ONfP7KWfymIyZjA0

#freetalks | #communities | #solarmind | #hub | #libstc | #aaron | nexus.stc@proton.me

We are celebrating 4 years today.

The first prototypes of the Telegram bot were coded at the end of 2019. The Nexus Search channel was created on the 3rd of March, 2020.

Thank you for staying with us all this time.

😊️️️️️️ Nexus March Digest 😊️️️️️️

👯‍♂️ The /q command now offers multilingual semantic search based on BGE-M3 embeddings. You now can search semantically using any language and in documents written in any language.

💻 Developers have detailed (join #communities to read)the neural networks utilized in Nexus for search and LLM.

📙 An HTTP API for uploading (join #freetalks to read) papers outside of Telegram is now available for testing to active contributors.

🗺 Nexus has integrated the Worldcat catalog, enabling the upload of any book. Credit for the Worldcat scrape is attributed to Anna's Archive.

🤖 Testing of Cybrex continues, with access to LLM granted to users who have uploaded papers and activated a Gaia subscription in their profile. Others may join #freetalks to take a look on what our LLM is capable to do.

📟 The Iroh protocol is advancing to the next version, incorporating our feedback. Development of software based on Iroh is ongoing to enable real-time replication. Users can join replication via the /seed command to receive freshly uploaded items, though ongoing bug fixes in progress.

🌏 We have conducted early tests of ALMA 13B to make automatic translations of scholar papers into any language possible. We are looking for contributors with GPU-rakes to setup a production cluster for auto-translations in addition to our own GPU-rake.

#freetalks | #communities | nexus.stc@proton.me

huggingface.co

BAAI/bge-m3 · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

*⚡ *Made of Iroh

We are testing another way for the replication of papers that allows us to mitigate existing issues. The IPFS requires us to release papers in large bulks and use DHT, which may behave unreliably under load.

Iroh is a novel p2p-protocol and implementation that provides users with a way to create a shared directory of unlimited size and replicate this directory across multiple peers in a faster way. Moreover, the Iroh client allows you to configure which exact parts of the directory will be replicated to your computer, meaning that small seeders may also participate without significant effort.

We would like to thank here the Iroh team and researchers of Earthstar and Willow protocol who provided a solid ground for the used approach.

How to use?

We have created a shared directory with DOI-named files and their extensions and are suggesting that you join its replication if you have OSX or Linux computer at hand.

We are making all papers uploaded through Nexus Aaron become available through Iroh within seconds, and we are also planning to publish all post-2021 papers within next months. You are free to take these papers, use them, repack them to torrents, upload them to LibGen, or do whatever you want.

A detailed guide and credentials are available in the /seed command for users who have more than 20 n-points. That limit is needed to manage initial load, the restriction will be removed in a week.

P.S. Both projects, Iroh and our replication that uses Iroh, are in an experimental stage right now, meaning that we may change anything or shut it down at any time.

Science Talks | Support | Twitter | nexus.stc@proton.me

iroh.computer

Documentation - Iroh

Learn about Iroh and how to integrate into your own project.

🎁 Rewards for uploads

Historically, all requests in bots are split into two categories: requests to our internal storage and online requests that attempt to retrieve papers from the web and then forwards requests to our group of human uploaders.

We are currently testing a reward system that will allow us to regulate downloading from the web and activity in the group of uploaders. Each Nexus user will be granted 10 points that can be spent on requesting papers from Internet sources, such as fresh journal articles, books not present in any free shadow library, and standards. Points can be earned by uploading books in papers in the same group.

This does not relate to the millions of books and papers already stored in STC and relate only to new content that is not present in STC. Most users won't notice any changes here.

How the entire process is organised? You request something with *📦 Request* button. The requested item will be downloaded from the Internet using our methods or forwarded to the @nexus_aaron group. People with institutional access (it may be you, just join the group, read pinned message and start uploading) fulfill others' requests in the group and earn 2 points for usual papers and 3 points for books and papers from the most famous scumbag publisher. Additionally, earned points can be spent to make queries to Cybrex LLM in the near future.

Once downloaded, the paper is placed into STC vaults and will be free for further requests.

Telegram

Nexus Aaron

Mutual aid society for exchanging scientific papers

"We have a good Xmas tradition, every year we shoot ourselves in the foot by trying to fight windmills, being afraid of the knowledge liberated for all people in the world." (c) Ass Chand Publishing Chaos

Merry Xmas, my friends! Do not forget to switch to other bots. The following year will bring you a lot of good surpirses.

FBI celebrates anniversary of Z-Library domain seizure.A lot more Z-Library domains have been captured and set to redirect to the FBI site two days ago. Another circus fighting against domain names is Austria’s Cybercrime Competence Center in Vienna. I don’t know who the fuck they are, but they have placed an entire logo on the seized sites pages. Possibly it is a design agency or something like that.

https://torrentfreak.com/fbi-austrias-c4-hit-z-library-with-a-massive-new-wave-of-domain-seizures-231108/

Science Talks | Support | Twitter | nexus.stc@proton.me