I have a small app built around an OpenAI-compatible client and am considering pointing it at DeepSeek. I know changing the base URL and model name may be straightforward, but I don’t want to assume every behavior will carry over unchanged.

For people who have made this switch, which parts needed the most attention: request parameters, streaming, error handling, or something else? I’m looking for practical migration surprises rather than a general model-quality comparison. Did your existing OpenAI-compatible client work with only configuration changes?