ChemWorldModel
An open knowledge graph for chemistry. Query reactions, explore molecules, and plan synthesis routes using natural language — backed by 1.8 million reactions from the Open Reaction Database and enriched with data from PubChem, ChEBI, ChEMBL, and GHS hazard classifications.
What this project does
Ask questions in plain language — “What catalysts are used in Suzuki coupling reactions with yields above 80%?” — and get answers grounded in real experimental data, with full provenance back to the source.
Search molecules by structure, substructure, or fingerprint similarity. Every molecule is identified by InChIKey and enriched with properties, biological roles, bioactivity data, and safety classifications from multiple public databases.
Plan synthesis routes by traversing the reaction graph. Given a target molecule, find multi-step paths scored by yield, step count, and commercial availability of starting materials.
Data Sources
Every record traces back to a public, open-access data source with full provenance tracking.
Contributing
ChemWorldModel is open source. The codebase, schema migrations, and loader implementations are all available on GitHub. Contributions welcome — whether that’s adding a new data source, improving the query pipeline, or fixing a bug in the frontend.