Python script queries local Ollama model using only the standard library

A developer describes calling a locally running Ollama model from Python using only the standard library, with no SDK or external packages. The approach posts a request to Ollama's HTTP endpoint on port 11434 and reads the generated text from the response. The author used this method in a small tool, about 306 lines long, that explains Swiss administrative letters in plain Italian. In testing, a 1B-parameter Gemma model took roughly 40 seconds on a CPU to explain a payment order, a trade-off accepted for privacy since no data leaves the machine. The author notes that for needs like retries, pooling or async calls, a dedicated client library may still be worthwhile.