BookyAI says Ollama is "not reachable" while a large model is loading, but a smaller model works. Why?

Before each request, BookyAI checks that Ollama answers. While Ollama loads a very large model, the disk is busy and that check could take longer than BookyAI waited, so it reported Ollama as not reachable even though it was running. Update to version 1.9.4 or later: the check is more patient, a slow answer no longer stops the request, and "not reachable" is only reported when the connection is genuinely refused. In that case the message includes the system error code (for example ECONNREFUSED) to help with troubleshooting.