At Lizard Soft, we continue to innovate our products with the newest Artificial Intelligence (AI) features and data processing tools. Now we significantly updated the content indexing mechanism with which the UniBot Personal Assistant & Corporate Chatbot works. This mechanism greatly enhances the Retrieval Augmented Generation (RAG) architecture. And it, in turn, allows a Large Language Model (LLM), similar to ChatGPT, to use the most powerful information search system to ensure maximum accuracy of answers.

The integration of such an information search system in UniBot v2 greatly increases control over the data used by LLM in formulating answers. For enterprises, this means that the capabilities of generative AI can be limited to enterprise document content derived from vectorized documents, images, and other data formats. As long as the appropriate embedding models exist for this content.

Before this update, UniBot v2 used Azure Account Storage Blobs to store intermediate indexing results, which was sufficient for many situations. However, due to the growth of data volumes and increasing requirements for processing this data, we have added Cosmos DB support. This way not only did we ensure the necessary flexibility and scalability, but also optimized the indexing process for huge databases.

The key advantage of this solution is the ability to manage the indexing process by your IT administrators: they can now change the number and order of steps, which makes the indexing system adaptive. And thanks to this, enterprises can quickly integrate the latest AI models and embedding models without re-indexing the entire wealth of information. And that significantly saves time and resources!

Let's consider a practical usage:

• a Bank using UniBot v2 can promptly update information in its call center with ongoing changes in its products and services.
• a Law Firm can significantly increase the productivity of searching through documents of the legal base (laws, by-laws, court decisions, etc.), and this is a key factor in the efficiency of lawyers' work.
• Pharmaceutical Companies can accelerate the analysis of scientific data for their developments, relying on accurate and fast indexing of scientific publications and available clinical data, both domestic and foreign.

In short, the updates in UniBot v2 in fact are the new capabilities for businesses to manage large volumes of data. At the same time, the information used by AI systems is being provided with the most accurate and up-to-date data. The scalability and adaptability of the new indexing architecture ensure that businesses can quickly respond to changes without the extra effort of re-indexing. So, to fulfill their business tasks, the potential of AI will be used in its entirety.