hearth-llm acts as a proxy for llama.cpp's llama-server, maintaining persistent key-value caches to keep local LLM contexts warm across sessions. This avoids expensive recomputation when resuming conversations or tasks with large language models. It is a lightweight developer tool aimed at improving the performance and responsiveness of local LLM inference setups.
In the AI & ML space, hearth-llm takes a focused approach. It focuses on losing LLM context and having to rebuild KV caches every time a local model session restarts. It is built as an open-source project for developers. hearth-llm is open source under the MIT license. hearth-llm is available on the command line, and it can be self-hosted.
hearth-llm first shipped in 2026. Key capabilities include KV Cache Persistence, LLM Proxy, and llama.cpp Integration.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Same category — not a similarity match