Monster 17: Gigi AI search

The main articles are:

• Speech Recognition (Technology, Market)
• Speech technology: On the path from recognition to understanding
• Data mining Data mining

2025

Move to SSE streaming to improve real-time AI search stability

Monstr17 has introduced SSE (Server-Sent Events) technology into the architecture of its AI search, providing more stable data exchange and accurate synchronization of user sessions. The update is aimed at improving the reliability of the interface, in cases where users work simultaneously with several chats or their Internet connection is unstable. The company announced this on December 12, 2025.

Previously, the service used a Websocket connection to transmit information about search steps to the user, as well as to broadcast a response from AI in real time. However, with short-term communication losses, users may not have received the current status about the search, or the response from the AI was interrupted.

The transition to SSE, as well as the implementation of an additional caching mechanism, completely fixed this problem. Thanks to the built-in automatic reconnection mechanism and the transmission of the last processed event, the server continues to stream data strictly from the last successfully delivered event.

SSE technology is lighter than Websocket, so along with this update, the load on the servers has also decreased, which allows you to handle more traffic at the same capacity.

When a user runs multiple queries, the system must maintain the transparency and accuracy of each query. SSE provides streaming data and gives us the opportunity to scale the service without losing stability, "said Grigory Antonov, CTO and co-founder of Zhizhi.

The developer noted that the modernization of the communication layer was a key step before expanding research and increasing the number of active users: the infrastructure will be ready to support more parallel operations and long computing processes.

Zhizhi AI Search View

Russia The " AI Gigi" search is presented in, using generative models and adapted for Russian Internet the -segment. This product has become a domestic alternative to foreign services Perplexity.ai ChatGPT and Search and allows you to receive answers quoting sources. The company Monstr17 announced this on November 24, 2025.

Unlike traditional search engines, this service generates responses in a natural language and accompanies them with clickable links to sources, allowing the user to check the accuracy of information and delve into related topics. The search works in real time, analyzing data from open sources.

Among the functionality of "Zizhi" is the "Study" mode, in which the system performs a comprehensive analysis of data from more than 40 sources in 5-7 minutes. The Related Topics feature forms a pool of recommendations on related issues to continue to immerse yourself in the topic. In addition, the service provides an intelligent chat that allows the user to discuss the found data in a dialogue format and refine requests without having to formulate them again. The ability to analyze documents and images is also available: the system recognizes text, highlights key data and connections, and interprets visual information, helping the user to work with reports, research or photos without using third-party tools. This combination of modes makes the service not just a tool for finding information, but a universal research platform.

The service is developed on the basis of open models GPT OSS and DeepSeek and is optimized for work in the Russian cultural and legislative field. This enables fast data processing at moderate infrastructure costs, reducing system cost of ownership and simplifying implementation in enterprise environments Zhizhi is focused on both the mass user and corporate IT teams. The product is available on the desktop and in mobile versions with support for the App Store, Google Play and RuStore. The service supports integration with external systems through the API, which allows you to use as a component of corporate ecosystems.

In the next five years, an additional $20 million is planned to be invested in the development of Zhizhi. According to the team, these funds will be used to increase computing power and develop their own AI model for search tasks in the Russian-speaking segment, optimized for the context, structure and sources of the Runet. This approach will allow you to quickly update the technology stack and quickly respond to market changes.

The company's goal is to reach 10% of the Russian search market within three years. To do this, Zhizhi plans to consistently improve the quality of user experience, develop the service as a product alternative to existing search engines, and expand integrations with Russian corporate ecosystems through the API. According to the founders, by 2028 the audience of the new service may exceed 20 million active users.

The main quality of AI search is that it goes beyond the traditional search, limited by the issuance of a list of sources on demand. AI search analyzes many sources, understands the context and forms a meaningful response - a conclusion similar to an expert opinion. Thus, search turns from a data extraction tool into intelligent user interaction, "added Grigory Antonov, technical director and co-founder of Zhizhi.

The launch of AI search with citation is an important step in the development of the domestic AI ecosystem. We see how Russian AI-based products are catching up and in many aspects are starting to outstrip global trends. Our goal is to make the search not just convenient, but transparent when the user understands where the information comes from, - commented Alexey Nechaev, CEO and co-founder of Zhizhi.