README.md

---
title: SobroJuriBert
emoji: ⚖️
colorFrom: blue
colorTo: indigo
sdk: docker
pinned: true
license: apache-2.0
---

# SobroJuriBert - French Legal AI Assistant v2.0

Production-ready API for French legal document analysis powered by JuriBERT.

## 🆕 Version 2.0 Features

### Complete JuriBERT Integration
- **Mask Filling**: Complete masked tokens in French legal text
- **Embeddings**: Generate 768-dimensional semantic embeddings
- **Enhanced NER**: Extract persons, money, dates, organizations, courts, and legal references
- **Question Answering**: Answer questions about legal documents
- **Advanced Classification**: Classify documents with keyword analysis

### API Endpoints

#### Core Analysis
- `POST /mask-fill` - Fill [MASK] tokens in legal text
- `POST /embeddings` - Generate text embeddings
- `POST /ner` - Extract named entities (enhanced)
- `POST /qa` - Question answering
- `POST /classify` - Document classification

### Entity Types Supported
- **PER**: Persons (M. Dupont, Me Martin)
- **MONEY**: Monetary amounts (1000€, EUR 50.000)
- **LEGAL_REF**: Legal references (article L123-4, loi n°2024-001)
- **DATE**: Dates in various formats
- **ORG**: Organizations and companies
- **COURT**: Courts and tribunals

### Usage Examples

#### Mask Filling
```python
POST /mask-fill
{
  "text": "Le contrat est [MASK] entre les parties.",
  "top_k": 3
}
```

#### Named Entity Recognition
```python
POST /ner
{
  "text": "M. Dupont a payé 15.000€ selon l'article L123-4 du Code civil le 15 janvier 2024"
}
```

#### Question Answering
```python
POST /qa
{
  "context": "Le bail est signé pour une durée de 3 ans renouvelable.",
  "question": "Quelle est la durée du bail?"
}
```

## About
Created by Sobro Inc. for French legal professionals.
Powered by JuriBERT and state-of-the-art French NLP models.

Version: 2.0.0 - Full JuriBERT Integration
<!-- Deployed on Vercel -->