Show HN: Frugon – Find which LLM calls a cheaper model could handle (local, MIT) (github.com)

Frugon, an innovative, open-source cost analysis tool for local large language model (LLM) use, has been launched to help developers track and minimize their LLM expenses. By analyzing call logs from LLM interactions, Frugon assesses whether switching to cheaper models could generate significant savings, all while ensuring user data remains private and local. Users can easily start analyzing their costs with minimal setup, using commands like `frugon analyze ./logs.jsonl` to get a detailed savings report. This tool is crucial for the AI/ML community as it addresses the growing concern of spiraling LLM costs, particularly for developers leveraging high-expense AI models. Frugon's unique ability to provide routing recommendations—such as suggesting which calls can be redirected to less expensive models—can result in substantial monthly savings (potentially over 30-50% or more, depending on usage patterns). With features that measure the performance of alternative models against the user's own data, Frugon empowers developers to make informed decisions about their AI workloads without compromising on quality.