Open-source Go tool explains production logs locally using an on-device AI model

logwhisper is a small command-line tool written in Go that reads production log lines and describes what happened, the likely cause, and what to check next. It runs entirely on the user's machine through Ollama and an open-weight Llama 3.2 model, so no API key is needed and logs are never sent to the cloud. The tool was built to speed up first-pass log reading during late-night incidents, and a sample log in the repository shows a service running out of Postgres connection slots. The author notes that the smallest model mostly narrates log lines one by one, while the larger 3B model synthesizes cause and effect more effectively. The project is a dependency-free Go module with chunking for small context windows and a prompt template designed to discourage small models from inventing root causes.