🛠️ Roadmap

# MedAI / VetAI / SantAI Roadmap

## Phase A – Prototype minimal - On Premise ONLY

- Curated RAG structure
- Core modules integration (SantAI = common technical core to VetAI and MedAI)
  - Transcription: Whisper (local)
  - LLM: Llama 3.1 70B via Ollama local
  - RAG: LlamaIndex ✅ Vous maîtrisez déjà
  - Framework: Python + FastAPI ✅ Parfait
  - Base vectorielle: Qdrant ou ChromaDB (local)
  - Interface: Streamlit (rapide) ou Gradio

```
santAI/
├── transcription/
│ ├── whisper_local.py # Whisper base model
│ └── audio_processor.py # Preprocessing audio
│
├── rag/
│ ├── vectorstore/
│ │ ├── chroma_db/ # Base ChromaDB locale
│ │ └── index_manager.py # Gestion index LlamaIndex
│ │
│ ├── medical_corpus/
│ │ ├── guidelines/ # Vos JSONs guidelines
│ │ ├── pharmAI/ # Données Pharmacopéé
│ │ ├── protocols/ # Protocoles thérapeutiques
│ │ └── tropical_diseases/ # Spécifique Réunion
│ │
│ ├── retriever.py # ⭐ LLAMAINDEX
│ ├── embeddings.py # Modèles embedding (sentence-transformers)
│ └── corpus_loader.py # Chargement JSONs → LlamaIndex
│
├── llm/
│ ├── qwen3_interface.py # Interface principale Qwen3
│ ├── qwen3_thinking.py # ⭐ Mode thinking médical
│ └── config.py # Configuration modèle
│
├── prescription/
│ ├── dosage_engine.py # Calculs posologies
│ ├── interaction_checker.py # Vérification interactions
│ ├── molecule_visual.py # Active molecule visual
│ │ https://molecule-icon-generator.streamlit.app/
│ └── vidal_interface.py # Interface base Vidal
│
├── api/
│ ├── main.py # FastAPI principal
│ ├── routes/
│ │ ├── consultation.py # Endpoints consultation
│ │ ├── diagnostic.py # Endpoints diagnostic
│ │ └── prescription.py # Endpoints prescription
│ └── models.py # Pydantic models
│
├── interface/
│ ├── app.py # Streamlit principal
│ └── components/
│ ├── transcription_ui.py # UI transcription
│ ├── diagnostic_ui.py # UI diagnostic différentiel
│ └── prescription_ui.py # UI prescription
│
├── requirements.txt # Dépendances
├── config.yaml # Configuration globale
└── README.md # Documentation
```

- First local LLM test
- Proof Of Concept

## Phase B – Academic pilot

- Clinical and veterinary validation
- User feedback loop
- Improved testing and CI

## Phase C – LTS (Long-Term Support)

- Stable release
- Debian packaging & integration
- Documentation & multilingual support
- Governance & funding model operational

## Future

- International regional deployments under SaintAI core
- Integration with public health datasets and partners