The Russian company Doomet has launched Doofram, an operating environment designed for managing large language models and building AI applications. The platform integrates essential components such as LLM interaction, tool management, context storage, and result evaluation into a single infrastructure.
Doofram serves as the technological foundation for the company's ecosystem, which includes the Doomet, Doutrace, and Dooaide products. These tools utilize the agentic framework for tasks ranging from document analysis and scientific text processing to software code management.
Over the past month, the company recorded approximately 47,000 user requests across its AI assistants, with over 45,000 resulting in successful execution. During this period, the infrastructure processed more than 30 billion words of input context and generated 150 million words in responses.
The platform is positioned as a secure, domestic solution that prevents data transmission to third parties. It supports various deployment models, including cloud-based configurations with proprietary models, integration with third-party LLM providers, and on-premises installation for corporate clients.
