CandyLLM is a Python library for working with Hugging Face and OpenAI text-generation models. It emphasizes production-oriented features such as monitoring, event systems, model selection helpers, and an extensible plugin/tool architecture.
The project targets practical LLM applications: multi-provider model support (including PyTorch/Hugging Face stacks), Docker deployment options, caching and performance helpers, and configuration patterns for agentic workflows.