Neural Datafari is an extension that allows you to benefit from generative AIs at every step of your knowledge management strategy, from data crawling up to their usage.
Similarly to how a brain works, a generative AI cannot generate an answer by using only a specific subsets of documents used at the learning phase. Therefore, companies that have document level permissions, cannot train a model with this corpus, unless they want to run the risk of disclosing confidential information to unauthorised employees. The Retrieval Augmented Generation approach brings in a solution.
RAG is a two steps approach: when a user queries Datafari (1), a vector search is run (2), to quickly identify the relevant authorised documents; within these documents, it identifies the most relevant chunks. The top chunks are then sent to the generative AI model as a context to the user query (3); the AI model is tasked to generate an answer using only the chunks provided with the query (4). This way, we can guarantee that the user only gets an answer generated with documents he is authorised to access (5).
More about Datafari
Crawling
This is the step where data is fetched from the knowledge sources. AI can extract information, classify, summarise documents, and much more
Search
This is the phase where queries are analysed, documents are ranked, and results are displayed. AI allows for the use of vector search to manage natural language queries and semantics.
AI Chatbot
Together with standard search, our AI chatbot generate answers based on internal documents, or targeting a specific document. A quick action mechanism gives employees a way to act fast, for instance to translate a document.
Security
Everything that relates to security, be it users authentication, authorisation, encryption and respect of the document level access rights. RAG guarantees you can use gen AI respecting documents confidentiality.
Used by diverse and renowned companies
Digital Sovereignty
Via the NGI Search European project, Neural Datafari is supported by the European Union to bring in the power of generative AI within Datafari Enterprise Search, thanks to vector search and RAG. An EU and open source approach guarantees digital sovereignty while respecting european freedoms and legislation.
Online demo