Local data
User documents and context can remain in infrastructure chosen by the user.
YANDI AI
Personal and local artificial intelligence — not simply another chat with an LLM.
YANDI AI is designed as an environment that can combine local and external models, long-term memory, reasoning, tools, expert modules, local data, network resources and other YANDI nodes.
Intelligence beyond one model's weights
A model does not have to store all knowledge in its own parameters. External memory, knowledge structures and specialized skills can be full parts of the system, connected when needed.
Tools
Specialized components handle specific tasks while AI helps select and connect them.
Memory
Long-term memory should be controllable, portable and separate from any particular model.
Experts and Skills
Language and domain modules can specialize in programming, mathematics, search, document analysis and other tasks. They do not need to turn the base language module into an encyclopedia of every tool.
This direction is in development.
Memory belongs to the user
Users should be able to decide what is saved and deleted, what is shared and with whom, which models can access it, and which nodes are trusted.
Some of these capabilities are under research and are not promises of a finished production service.