ollama-serve

Keep your local Ollama server warm and your favorite models ready. This package offers small, composable helpers to check server health, start it when needed, and pull models on demand.

Install

uv add ollama-serve

Quick start

from ollama_serve import ensure_model_and_server_ready

if ensure_model_and_server_ready("llama3:latest"):
    print("Ready to chat.")

Common flows

from ollama_serve import (
    is_ollama_running,
    run_ollama_server,
    is_model_installed,
    install_model,
)

is_ollama_running()
run_ollama_server()
is_model_installed("llama3:latest")
install_model("llama3:latest")

Tips

• If Ollama is not on your PATH, install it from https://ollama.com/download.
• Use smaller timeout values for snappier "is it up?" checks in dev loops.

Compatibility

• macOS, Linux, and Windows (where Ollama runs locally).
• Requires the ollama CLI on your PATH.

Logging and config

• Uses standard Python logging via logging.getLogger("ollama_serve").
• Default settings can be overridden with env vars:
  • OLLAMA_SERVE_TIMEOUT (seconds, default 0.2)
  • OLLAMA_SERVE_RETRIES (default 1)
  • OLLAMA_SERVE_RETRY_DELAY (seconds, default 0.2)