This system consists of three components:
1. Database registration function for vector search
2. Trial function for vector search
3. RAG utilization function
With this system, enterprises can easily create RAGs. Let's take a closer look at the features.
1. Database Registration for Vector Search
Input side: By entering the URL of a website, the data from the site will be saved. During this process, the data is divided into chunks, with the chunk size being adjustable. Additionally, considering the continuity of the data, there is an overlap feature that can be specified.
Output side: This displays the number of indexes in the registered database, which is the number of chunks the retrieved data was divided into. The name of the saved database is also displayed.
2. Trial Function for Vector Search
This feature allows you to test similarity searches using the database you just created. In RAG, an initial search is conducted. This feature provides the top 5 similar sentences based on the input prompt, ranked by similarity.
Input side: The database name and the prompt, referred to as a query. The output is the top 5 candidates ranked by similarity.
3. RAG Utilization Function
Finally, we have the RAG function. This allows further refinement and generation of text using the created database and prompt.
Input side: You enter the database name, create a prompt template that includes search results, and finally enter the query to conduct the search. The output is the text generated using RAG.
Our company provides a GPU environment, but since this is a managed service, you can experiment extensively without any hassle. It's extremely convenient, so please contact us for more information. Due to the use of GPUs, this service is offered on a paid basis to enterprises. Thank you for your understanding.