github.com/Cooksey99/llm-workspace ↗
Nucleus is a modular Rust library for building advanced local AI tools and applications, built-in RAG for private documents, and an extensible plugin system for custom workflows.
Open this visualization on its own page →
Contributors
2
Lines of Code
140
From
2025-11-08
To
2025-11-11
About Cooksey99/llm-workspace
Nucleus is a modular Rust library for building AI-powered developer tools that keep data private by running entirely on local machines. Rather than a finished application, it provides infrastructure components for integrating local or self-hosted language models with real capabilities like reading files, searching code, and executing commands. The system uses a tool-augmented approach where an LLM can recognize when it needs to take actions and invoke plugins to accomplish tasks, enabling AI assistants that understand codebases and filesystems without sending information to external services.
The library is organized as a Rust workspace with distinct modules: nucleus-core handles LLM orchestration and chat management, nucleus-plugin provides the plugin trait system with permission controls, nucleus-std offers standard plugins for file operations and command execution, and nucleus-dev includes developer-specific plugins with git and LSP integration support. It also includes built-in vector database support for retrieval-augmented generation, enabling persistent storage of indexed documents with automatic deduplication and efficient vector search across large codebases.
The project targets developers building custom AI assistants and internal tools where privacy and customization matter more than off-the-shelf solutions. Use cases range from AI-enhanced terminal wrappers and code review systems to interactive documentation and project-specific helpers. The library is under active development with an MIT license, and the maintainer explicitly welcomes contributions in areas like new plugins, additional LLM backend support, documentation, and testing.