ModKie lets you deploy HuggingFace weights on RunPod (vLLM or custom workers) and expose them through the same API key as marketplace models. After deployment completes, use register-gateway to obtain a model ID usable in createTask.
High-level flow: pick a HF repo → launch RunPod template → wait for health check → register endpoint → test in Playground → call createTask from your app. Custom models appear alongside Kie-routed catalog entries in GET /models.
Training-focused teams often pair this with vision workflows documented in our YOLO training post.

