|
--- |
|
license: apache-2.0 |
|
--- |
|
|
|
# GreenChat |
|
An environmental and sustainability focused QA system. |
|
|
|
|
|
## 🌍 Background |
|
GreenChat is a domain-specific RAG model designed to support environmental decision-making across multiple domains relevant to UN SDGs. The system integrates knowledge from: |
|
- Climate Science |
|
- Renewable Energy |
|
- Environmental Protection |
|
- Economics |
|
- Natural Technologies |
|
|
|
## 📚 Dataset (GREEN Database) |
|
Currently available datasets: |
|
``` |
|
G - Global Climate and Weather (7.03k samples) |
|
R - Renewable Energy (2.36k samples) |
|
E1 - Environmental Protection (10.3k samples) |
|
E2 - Economics and Policy (31.4k samples) |
|
N - Natural Technologies for Sustainability (57 samples) |
|
``` |
|
|
|
## 🛠️ Technical Stack |
|
- **Base Model**: Mistral Instruct 7B |
|
- **Framework**: LangChain |
|
- **Embeddings**: Sentence-Transformers (all-MiniLM-L6-v2) |
|
- **Vector Store**: FAISS with IVF-PQ indexing |
|
- **Evaluation**: RAGAS |
|
- **Interface**: Streamlit |
|
|
