aakothari/Conformer

1---2title: Conformer3emoji: 🧬4colorFrom: gray5colorTo: blue6sdk: static7app_file: index.html8pinned: false9license: mit10short_description: AI molecular workbench for strings11---12 13# Conformer14 15A browser workbench for the BRAID molecular line notation and the BRAIDBERTa /16DeepBERTa encoders. Load molecules, convert between representations, draw17structures, compute descriptors, benchmark tokenisations, embed and compare18encoders, export datasets.19 20Nothing to install. Structure parsing, drawing, descriptors and fingerprints run21entirely in your browser via RDKit's WebAssembly build, so most of the workbench22works with no server at all.23 24BRAID conversion and the encoders need Python — `braids.codec` depends on RDKit25APIs that the WebAssembly build does not expose — so those routes are served by26a small FastAPI service:27 28**https://braid-9wc2.onrender.com**29 30The encoders run there from ONNX graphs exported from the checkpoints below,31on CPU. That service is on a free instance and sleeps when idle: the first32request after a quiet spell can take ~50 seconds to wake it. Everything33browser-side stays instant meanwhile.34 35- Code: https://github.com/AayushK-othari/braid36- Models: https://huggingface.co/aakothari/BRAIDBERTa37  and https://huggingface.co/aakothari/DeepBERTa_zinc_base_100k_v438 39`GET /version` on the backend returns the git SHA, the pinned model revisions40and the installed package versions behind whatever you are looking at.41 42## Pointing at your own backend43 44The workbench's settings panel takes any endpoint. To run the converters and45encoders locally instead — nothing leaves your machine:46 47```bash48pip install -r requirements-backend.txt49python backend.py50```51 52then set the endpoint to `http://127.0.0.1:8000`.