I m preparing Navor AI, a unified developer gateway for DeepSeek, Qwen, Doubao, Seedream, Seedance, voice, and 3D models. The goal is to make authentication, credits, documentation, and usage logs consistent across providers.

For developers working with Chinese AI models, which capability or provider is still the hardest to integrate today? I d especially value feedback on API ergonomics, model coverage, and the next feature that would save you the most time.