chematic

Cheminformatics for Python, Rust, and the browser. chematic is built in pure Rust, with bounded input handling, typed errors, and optional native InChI.

日本語 · 中文 · Documentation · Live demo

Install

pip install chematic
cargo add chematic --features "smiles,perception,chem,3d,fp"
npm install @kent-tokyo/chematic

Python needs no C/C++ compiler. Rust and WebAssembly builds use the same core.

v1.0.13 release boundary

This release records expanded reaction, V3000, parser-safety, CIP, and cross-binding validation. RDKit-compatible fingerprints, Markush/polymer semantics, and browser artifact parity remain bounded; see the compatibility scope and CHANGELOG.

Python

import chematic

mol = chematic.from_smiles("CC(=O)Oc1ccccc1C(=O)O")
print(mol.mw, mol.logp, mol.tpsa)
print(mol.has_substructure("[OH]"))

report = chematic.report([mol], names=["aspirin"])
report.save("report.html")

For a small RDKit-compatible subset:

from chematic import rdkit_compat as Chem
from chematic.rdkit_compat import Descriptors

mol = Chem.MolFromSmiles("CCO")
print(Descriptors.MolWt(mol))

This is not a full RDKit clone. Unsupported options fail explicitly; see the migration guide.

Rust

use chematic::smiles::parse;

let mol = parse("c1ccccc1")?;
println!("{}", mol.atom_count());
# Ok::<(), Box<dyn std::error::Error>>(())

The workspace also contains focused crates for SMILES/SMARTS, descriptors, fingerprints, reactions, SDF/MOL/CDXML, 2D/3D, crystal formats, MCP, and bindings. See the format matrix.

JavaScript / WebAssembly

import init, { parse_smiles, get_descriptors_json } from "@kent-tokyo/chematic";

await init();
const mol = parse_smiles("CCO");
console.log(mol.atom_count());
console.log(JSON.parse(get_descriptors_json(mol)));

The WASM package supports molecule parsing, descriptors, fingerprints, reactions, selected 2D/3D operations, and format conversion. The current export surface is documented in the WASM README.

What is stable

• SMILES and SMARTS parsing/writing
• Descriptors, fingerprints, Tanimoto similarity, and substructure search
• SDF/MOL V2000/V3000 and selected chemical formats
• Python, Node/WASM, and Rust bindings
• Bounded parsing, typed failures, and deterministic batch APIs

Experimental or intentionally bounded areas include 3D generation, pKa/ADMET screening, IUPAC names, Markush/polymer expansion, CDXML editing, and the RDKit-compatible subset. Canonical SMILES is not always a safe deduplication key; use the fail-closed canonical_smiles_stable_key() API where required.

See compatibility scope, validation, and error and resource limits for exact guarantees.

For research software users

Use the researcher guide for a short path from installation to a reproducible result. It explains which APIs are stable, which comparisons are version- and corpus-pinned, how unsupported cases are reported, and how to cite a specific release. The benchmark index keeps performance claims separate from correctness and compatibility evidence.

MCP server

chematic-mcp provides local chemistry tools over stdio for MCP-compatible agents. It performs no network access except for the optional name lookup tool. See its README.

Development

cargo build --workspace
cargo test --workspace --all-targets --locked
cargo clippy --workspace --all-targets --locked -- -D warnings

Benchmark methodology and dated results are kept in docs/benchmark.md and benchmarks/. Release history is in CHANGELOG.md.

License

Licensed under either Apache License 2.0 or MIT, at your option. See NOTICE for attribution details.