| Mode | Endpoint | Body |
|---|---|---|
| Base | POST /predict_base | {"texts":["..."]} |
| Fine-tuned | POST /predict_finetuned | {"texts":["..."]} |
| RAG (Chat) | POST /predict_rag | {"texts":["..."],"top_k":5,"min_sim":0,"system_prompt":"..."} |
| RAG (kNN) | POST /predict_rag_knn | {"texts":["..."],"top_k":10,"min_sim":0,"strategy":"softmax","tau":0.10,"power":2.0,"abstain_min_conf":0.35} |