🏗️ A1–A5 Upfront Embodied Carbon Calculator (Local LLM)

Overview

This repository contains a lightweight, deterministic calculator for estimating A1–A5 upfront embodied carbon in building materials and construction processes.
It was designed as part of a sustainability engineering project to make carbon assessments more transparent, modular, and interactive.

The tool includes:

  • Rule-based carbon computation following EN15978 and RICS conventions.
  • Local LLM integration (google/flan-t5-base) for generating carbon-reduction suggestions.
  • A Gradio GUI interface that allows users to input materials, quantities, and emission factors, and instantly view results and insights.

⚙️ Features

  • Input: A1–A5 modules, materials, quantities, units, emission factors, notes.
  • Output: Per-row and per-module emissions (kgCO₂e, tCO₂e) + total embodied carbon.
  • Integrated suggestion engine:
    • Rule-based heuristics for reduction strategies.
    • LLM-generated design insights using a small local model (no external API).
  • Built with strict unit validation to prevent inconsistent calculations.

🧠 Technical Notes

This calculator runs entirely offline using Hugging Face’s google/flan-t5-base model for text generation.
It does not require API keys and can run on CPU (no GPU necessary).

💡 The local LLM is not used to compute carbon, only to suggest design improvements.


🧩 How to Run

pip install -r requirements.txt
python app.py
Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support