Ecosynth

AI-assisted retrosynthesis planner that combines symbolic chemistry rules, RAG over literature/reaction data, and a green-chemistry scorer to propose, validate, and rank synthesis routes — with built-in hallucination detection and a React frontend for exploring reasoning traces.

What it does

• Retrosynthesis engine (ecosynth/) — wraps AiZynthFinder for route search, builds a constraint graph over reaction steps, repairs broken branches, and scores routes for green-chemistry metrics (solvents, reagents).
• Validity firewall — checks proposed reactions against a taxonomy of known hallucination/failure modes before surfacing them.
• RAG engine — retrieves supporting literature/reaction precedent from a Chroma vector store (data/chromadb/) to ground route suggestions.
• API (api/) — FastAPI service exposing synthesis, steering, and health endpoints.
• Frontend (frontend/) — React + TypeScript UI: synthesis tree view, reasoning/constraint timelines, preference sliders, failure console.
• Models (models/) — trained green-chemistry random forest scorer.
• Setup scripts (setup/) — download reaction data (USPTO-50k), build the Chroma DB, train the green-RF and ChemBERTa models.

Stack

Python (FastAPI, AiZynthFinder, ChromaDB, scikit-learn, transformers/ChemBERTa) + React/TypeScript/Vite frontend.

Getting started

conda env create -f conda_env.yml
conda activate ecosynth
pip install -r requirements.txt

# one-time data/model setup
python setup/01_download_data.py
python setup/02_build_chromadb.py
python setup/03_train_green_rf.py
python setup/04_train_chemberta.py

./run.sh

Frontend:

cd frontend
npm install
npm run dev

Tests

pytest tests/

Repo layout

api/        FastAPI app + routes (health, steer, synthesize)
ecosynth/   core pipeline: retrosynthesis, constraint graph, validity firewall, green scorer, RAG
frontend/   React/Vite UI
setup/      data download + model training scripts
models/     trained model artifacts
data/       reaction datasets, solvent/reagent tables, Chroma vector store
tests/      pytest suite
report/     project writeup (LaTeX)

---

Author

Jeyadev R · AI/ML Engineer & Full-Stack Developer · Chennai, India

• Portfolio: jeyadev.dev
• GitHub: @jeyadev-jd
• LinkedIn: r-jeyadev

If this project is useful to you, a ⭐ on the repo helps others find it.