Skip to content
Home » ExaWizards launches “RAGOps,” which enables improvements while operating generative AI and business data col laboration

ExaWizards launches “RAGOps,” which enables improvements while operating generative AI and business data col laboration

ExaWizards Co., Ltd.
ExaWizards launches “RAGOps,” which enables improvements while operating generative AI and business data collaboration
~If you are not satisfied with the answer quality, human intervention will be required to make additions and corrections, and services will be provided using exaBase Studio~
ExaWizards Co., Ltd. (Minato-ku, Tokyo, President and CEO: Makoto Haruta, hereinafter referred to as ExaWizards) has begun offering “RAGOps”, an application that can continuously improve the quality of responses related to business data through in-house operations. . By providing it on the AI ​​software development environment “exaBase Studio”, you can quickly optimize processing according to your company’s needs, using available AI models and a wide variety of in-house data. We have adopted a Human-in-the-Loop system that allows humans to intervene if the quality of answers is not satisfactory, creating a system that can be put to practical use in managing specialized documents and sales knowledge, responding to internal and external inquiries, etc.
[Image 1:×630.jpg] Data collaboration with generation AI is called RAG*, and demand is rapidly increasing because customers’ unique data can be processed and analyzed with generation AI. However, there are issues such as “not getting the expected results,” “I don’t want people to access data they don’t have permission to view,” and “I’m worried about the spread of incorrect answers with Hallucination,” making it difficult for companies to start using it in earnest. There are also quite a few. Therefore, ExaWizards uses exaBase Studio, which allows users to develop AI software in-house, to develop a RAGOps application that allows users to improve and operate RAG in-house, and provides it as one of exaBase Studio’s templates. In the future, you can continue to expand the functionality by combining parts that will be added to exaBase Studio templates. RAGOps is an application that aims for a “quick win” that allows you to start using it for business quickly, and achieves a return on investment sooner. In the future, we plan to support data that includes images, tables, and handwritten characters in addition to text.
*Also called lag, an abbreviation for Retrieval Augmented Generation.
[Image 2:×726.png ]
Figure RAGOps is provided as a template on the AI ​​software development environment exaBase Studio. By managing the collaboration between various databases, LLMs, and users and operators, we are able to improve the quality of answers in practice.☑︎️RAGOps Functions RAGOps mainly provides the following functions with an easy-to-use user interface. This enables user companies to operate and improve their own generated AI and data collaboration.
(1) Function to expand data during operation RAGOps accumulates information on whether the user is satisfied with the answer as feedback in the cache DB. For queries similar to questions with highly rated answers, cache is returned directly without using LLM, improving answer quality and cost efficiency. In addition, if the evaluation of the answer is low, an operator will intervene and prompt the user to input the answer data. This Human-in-the-Loop design improves operations while enriching missing information based on user needs. (2) Access management function according to user privileges: It has a function to manage the related data that RAG refers to when a user asks a question, according to user privileges. This function generates answers using only the appropriate data depending on the user’s department, position, question subject, etc.
(3) Customizing the addition and replacement of parts to suit data and business requirements.Each part of the application can be freely added or replaced in accordance with changes in business or data. This is a feature of applications on exaBase Studio. For example, data pre-processing/post-processing, user interface, etc. can be easily changed during operation to match the LLM model and data structure being adopted.
[Image 3:×424.png ]
Figure RAGOps application screen. RAG can be used in actual work, such as the function to contact the operator and request improvements if you are not satisfied with the response results (left), and the function to grant access privileges to files targeted for RAG data linkage at the time of upload (right). provides functions with an easy-to-use user interface.
☑︎️Examples of intended uses for RAGOps ・Providing a wide range of use cases that require expert knowledge such as searching legal documents ・Publishing documents such as internal regulations and product manuals, and responding to inquiries from people Establishment of a system that requires continuous expansion of data, such as efficiency in response to use cases that require both, knowledge management of sales knowledge, etc. ☑︎Provision conditions From May 2024,
applications on exaBase Studio templates Publish. Available now to licensed customers using exaBase Studio and through ExaWizards Professional Services.
☑︎Features of exaBase Studio exaBase Studio is a development environment that allows you to build AI software by combining internal and external AI models, services, and data. We utilize an intuitive design and development UI (user interface) called Canvas to visualize processes, allowing for open design and updating that involves not only engineers but also managers and internal and external personnel who are familiar with the business. It will be easy. It is equipped with a mechanism to semi-automate the complicated steps of
development, which will greatly change the development process. We support in-house development using an agile style of “improving moving things by touching them.”
Please see below for
details.・RAGOps・exaBase Studio
[ExaWizards Company Profile] Company name: ExaWizards Co., Ltd. (stock code 4259) Location: 5th floor, East Building, Sumitomo Fudosan Mita Twin Building, 4-2-8 Shibaura, Minato-ku, Tokyo Established: February 2016 Representative: President and CEO Makoto Haruta Business details: Industrial innovation and solving social issues through service development utilizing AI URL:
More details about this release: