/classify/zero-shot

Zero-shot classification (encoder-only NLI). Runs the NLI model against the hypothesis template for each candidate label and returns softmax scores. A single input returns a single object; a list returns an object with `results` per input.

curl -X POST "https://example.com/classify/zero-shot" \  -H "Content-Type: application/json" \  -d '{    "model": "mnli",    "input": "The player scored a hat-trick",    "candidate_labels": [      "sports",      "politics",      "cooking"    ]  }'

{  "object": "classification",  "model": "string",  "labels": [    "string"  ],  "scores": [    0.1  ],  "usage": {    "prompt_tokens": 0,    "total_tokens": 0  }}