mianabubaker/sparrowgate

SparrowGate is a small Python tool that sits between your application and a large language model provider like OpenAI. Its job is to cut down how much you spend on API calls. It does this by remembering prompts it has seen before and by sending simple queries to cheaper models instead of always using the most expensive one. The author, a 16 year old developer, built it to solve a problem he kept hitting in his own projects: tiny throwaway questions were burning through his API budget for no good reason. The tool works as a proxy. When your app wants to ask a language model something, the request goes to SparrowGate first. SparrowGate checks a local database to see if it has already answered that exact prompt. If it has, it returns the stored answer right away, so no new API call is made and no money is spent. If the prompt is new, SparrowGate looks at how long or complex it is. Short or simple prompts get sent to a cheaper model. Bigger or harder prompts go to the original model you wanted to use. After the answer comes back, SparrowGate saves it for next time. There are two ways to run it. You can start a dashboard mode with a visual interface, or you can run it as a pure API endpoint. Either way, you install the Python dependencies and run a single script. When running in API mode, SparrowGate adds a savings percentage directly into the response data your app already receives, so you can see how much money the caching and routing saved on that call. The README is short and honest about the state of the project. It is still early in development and the author is actively working on it. There is no feature bloat here. No dashboards for enterprises, no complex setup. It is meant to be something you drop in and forget about. The full code lives in a Python file called app.py, and the license is MIT, which means anyone can use or modify it freely. The full README is longer than what was shown.