In an exciting development for those working with AI technology, new features have been introduced to the GenAI Platform. These enhancements are designed to simplify the process of constructing, managing, and refining the knowledge bases that support AI agents. The upgrades include capabilities such as web crawling, custom crawling rules, and one-click reindexing. These tools allow users to keep their AI agents updated with pertinent, real-world information without the need for manual data collection or reliance on external storage. This suite of enhancements, along with the recent improvements to the Retrieval-Augmented Generation (RAG) system, enables AI agents to provide quicker, more precise, and contextually aware responses by drawing from well-organized and rich data sources.
The proficiency of AI agents is intrinsically linked to the data they can access. With the introduction of the web crawling feature, users can now gather data from publicly available websites and incorporate it directly into their knowledge base. This capability significantly reduces the burden of manual data collection. It is particularly beneficial for AI agents that depend on public web content to generate insights and drive actions.
Key Features of the Enhanced GenAI Platform
Custom Crawling Rules: This feature allows users to specify particular pages or entire domains from which their AI agents should extract data. As a result, the agents can collect information from precisely targeted sources, ensuring that the data is both relevant and useful.
One-Click Reindexing: Keeping your knowledge base current and relevant is crucial, and one-click reindexing facilitates this by refreshing the data effortlessly. This feature is particularly advantageous for applications such as financial analysis and competitive research, where up-to-date information is vital.
Integrated Storage Solution: By integrating crawled content directly into the knowledge base, the platform eliminates the need for external storage solutions. This not only simplifies the setup but also reduces the overhead associated with managing external data repositories.
Through web crawling, AI agents can continuously update themselves with real-time information, maintaining their effectiveness and relevance.
To further enhance the capabilities of AI agents, significant upgrades have been made to the Retrieval-Augmented Generation (RAG) system.
Enhancements to the Retrieval-Augmented Generation (RAG) System
Improved Accuracy for Text-Based Queries: The accuracy of responses to text-based questions has seen a substantial increase, now reaching up to 95 percent. This improvement means that AI agents are more reliable and can provide more accurate information.
Enhanced Handling of Tables and Graphs: The platform has made strides in processing tabular and graphical data. With improved layout models and GPU-accelerated Optical Character Recognition (OCR), accuracy for these types of data has increased nearly fourfold.
Metadata-Aware Responses: AI agents can now provide more contextually rich and clear responses by referencing metadata such as source URLs, file names, and PDF page numbers. This enhancement allows for more transparent and informed responses.
Metadata-Based Queries: Users can now filter results through natural language queries that incorporate metadata. For instance, a user can request a summary of a specific document, and the AI agent will be able to deliver precise results based on the provided metadata.
These enhancements to the RAG system make AI agents not only smarter but also more precise and transparent, enabling them to be useful across a wide range of real-world applications.
Additionally, the GenAI Platform has integrated Claude 3.7 Sonnet, offering access to Anthropic’s latest model, which is focused on advanced reasoning. Claude 3.7 Sonnet features an extended thinking mode, enabling deeper analysis and more accurate answers to complex queries. This model improves upon the strengths of the Claude 3 family, delivering enhanced performance, natural language fluency, and reliability. It is particularly suited for AI agents tasked with problem-solving, advanced comprehension, and providing trusted responses.
In conclusion, the latest features of the GenAI Platform, including real-time web crawling, improved RAG capabilities, and the powerful Claude 3.7 Sonnet, provide a robust foundation for advancing AI-driven projects. These enhancements are designed to elevate the capabilities of AI agents, making them more effective and versatile in a variety of applications.
For more information on these updates, you can explore the GenAI Platform and take advantage of these new features to enhance your AI projects.
*Note: The performance metrics mentioned are based on an internal benchmark study conducted in the first quarter of 2025. The study evaluated the accuracy and time-to-build metrics for AI agents with a knowledge base, using independent public datasets across various data types including text, tables, graphs, and multimodal data. Actual performance may vary, and a detailed methodology is available upon request.
For more Information, Refer to this article.
