Pitch: ScienceStr

Research and knowledge should be free

The idea behind this project is to enable broader and more unrestricted access to scientific knowledge while at the same time combine that with advancements of large language models to provide short answers based on derived information.

ScienceStr stands on the shoulders of great people and projects who have started liberating academic research years ago, from Aaron Swartz, Alexandra Elbakyan with sci-hub or the team behind standard template construct to name just a few.

Vision

Make running sciencestr incredibly simple so people can easily deploy it on their own hardware and run their own instances of it.

Tech

The tech and bot itself are very simple, specially in the current iteration.

It consist of couple parts:

1. nostr bot

2. api that acts as intermediary between backend and the bot

3. ipfs gateway

4. quant

5. cybrex

Code is accessible in the repository, due to zaps not being implemented and bot running on OpenAI api i'm not sharing the pubkey.

How it looks in practice

Next steps

• Current PoC uses OpenAI api due to simplicity of deployment but cybrex supports self hosted LLMs which need to be tested

• Simplify deployment (current state of docker deployment is not great)

• Improve functionality of the bot to make it aware of the conversation and able to expand on topics, run more extensive queries etc