Call a local model from plain Python: no SDK, no dependencies

A minimal Python 3.10 script without external libraries can invoke a local Ollama model by sending a POST request to http://localhost:11434/api/generate and rec