Metadata-Version: 2.4
Name: serigy-jev
Version: 0.1.2
Summary: Inference-only PT-BR fake-news classifier (load Hub weights + predict)
Author-email: Wendell Barreto <wendellbarreto@gmail.com>
License-Expression: MIT
Project-URL: Homepage, https://github.com/wendellbigato/serigy-jev
Project-URL: Hugging Face, https://huggingface.co/wendellperbar/serigy-jev
Requires-Python: >=3.14
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.26.0
Requires-Dist: pydantic>=2.5.0
Requires-Dist: torch
Requires-Dist: transformers>=4.40.0
Requires-Dist: sentence-transformers>=3.0.0
Requires-Dist: safetensors>=0.4.0
Requires-Dist: huggingface_hub>=0.24.0
Requires-Dist: python-dotenv>=1.0.0
Dynamic: license-file

# Serigy-Jev

Classificador local de fake news em PT-BR. Carrega pesos do Hub e faz `predict`.

```bash
pip install serigy-jev
```

```python
from serigy_jev import SerigyJev

model = SerigyJev("wendellperbar/serigy-jev")
out = model.predict(
    "Senado aprova lei que PROÍBE jogo de videogame para menores de 18 anos."
)
print(out["predicao"], out["score_final"], out["confianca_calibrada"])
```

```bash
serigy-jev --texto "texto da notícia..."
```

Se o sensor for gated, configure `HF_TOKEN`.

### Saída

| Campo | Significado |
|-------|-------------|
| `predicao` | `Claramente Falso` / `Incerto/Misto` / `Claramente Verdadeiro` |
| `classe_ordinal` | `0`, `1` ou `2` |
| `score_final` | veracidade `0.0`…`1.0` |
| `confianca_calibrada` | perto de `1` = mais concentrado |
| `probabilidades` | massa em cada classe |

Pesos: [huggingface.co/wendellperbar/serigy-jev](https://huggingface.co/wendellperbar/serigy-jev)

MIT — Copyright © 2026 Wendell Barreto.
