I’m considering testing DeepSeek through its OpenAI-compatible API in an app that already talks to another provider. The compatibility claim sounds convenient, but I’d rather hear from someone who has tried a real integration than assume every feature behaves the same.

Where have you hit differences—streaming responses, tool calls, error handling, or request parameters? I’m trying to gauge whether DeepSeek API compatibility makes a provider switch straightforward or still requires a careful round of testing.