The LearnLukso project qualifies for the "Social & Community" category because it aims to gather and make knowledge about artificial intelligence accessible in various languages. Additionally, by creating a Custom GPT chatbot, it supports the AI community, and the provision of a landing page to raise awareness about LUKSO and AI enhances its social and community-oriented aspect.
The LearnLukso project qualifies for the "Infrastructure & Tools" category as it has developed a REST API for querying LUKSO Profiles and a custom GPT chatbot that complies with OpenAI's guidelines, both of which serve as valuable tools for developers and AI enthusiasts working in the field of artificial intelligence. Additionally, the project's focus on improving technology and its future plans for co-creating profiles and enhancing text generation further solidify its position as an infrastructure and tool for AI-related endeavors.
LearnLukso aims to curate and prepare learning resources and source code for Artificial Intelligence, making even the most current information on AI accessible and useful.
Try the chatbot using this link:
> https://chat.openai.com/g/g-CZS73iqGp-learn-lukso
## Why A.I.?
A key argument for using AI-based documentation is its multilingual capabilities. AI can speak many languages, helping to make information more shareable.
## Objective
In mid-November, OpenAI enabled the uploading of custom documents and querying these documents through a special interface. To gain approval for this function, certain conditions must be met, such as domain verification, HTTPS, API-Key, and the provision of the API in a defined schema.
Therefore, the goal was to develop a complete Custom GPT Chatbot that complies with OpenAI's guidelines.
The following steps were necessary to achieve this:
### Document
To make the context understandable to AI, information must be provided. For this, 15 repositories were selected, downloaded completely, and compiled into independent documents. Additionally, the transcripts of eight tutorials were downloaded and added. More information can be found here:
> https://github.com/a6b8/upGPT#documents
Two modules were programmed for generating the documents. They help to download many files, convert them, and compress them into a document:
- https://github.com/a6b8/sitemap2doc
- https://github.com/a6b8/repos2doc
The results that were finally used can be found here:
- https://github.com/a6b8/upGPT/tree/main/data
## REST API for A.I.
To query information about LUKSO Profiles, a REST API had to be set up according to OpenAI's guidelines. Thus, an API was connected to the subdomain api.learnlukso.xyz. In addition, a ./privacy
route and a /health
route were set up. The code for the project is here:
> https://github.com/a6b8/upAPI
The getUniversalProfile
route can be used to query metadata, which can be useful for creating images.
```
curl --location 'api.learnlukso.xyz/getUniversalProfile?address=0xB031363560403179Aac100d51864e27fFF4D7807&network=luksoTestnet&presetKey=LSP3Profile' --header 'x-api-key: MustHaveTheLengthOf42withAtLeastCharAnd123'
```
### Landing Page
OpenAI also plans a marketplace that will likely offer free Custom GPTs. This is an excellent opportunity to raise awareness about LUKSO. In preparation for a possible admission to the OpenAI "Marketplace," a landing page was created. This page links to the chatbot and the LUKSO documentation.
## Limitations
- Currently only available for OpenAI Plus customers.
- Custom GPT queries are significantly slower than regular GPT-4 queries.
- Some documents are not correctly recognized, and some could not be uploaded.
Nevertheless, all promised features are available, and some show impressive results.
## Future
The technology is still quite slow, but this is expected to change in the coming months.
- Co-creation of profiles through Dall-E or Midjourney.
- Reading and improving reputation to create enhanced texts.
- Faster development through simplified search and quick connection to the blockchain.