SELFIES

2.2.0 verified Fri May 01 auth: no python

SELFIES (SELF-referencIng Embedded Strings) is a general-purpose, sequence-based, robust representation of semantically constrained graphs, primarily used for molecular structures in cheminformatics and ML. Current version 2.2.0, releases every few months.

pip install selfies

Common errors

error AttributeError: module 'selfies' has no attribute 'selfies' ↓

cause Wrong import style: trying to access selfies.selfies.

fix

Use from selfies import encoder, decoder directly.

error RuntimeError: The input SMILES could not be parsed due to semantic constraints. ↓

cause SMILES violates bond constraints (e.g., hypervalent atoms) and strict=True is default.

fix

Either fix the SMILES or pass strict=False to encoder.

error TypeError: encoder() got an unexpected keyword argument 'strict' ↓

cause Using version <2.0.0 where strict parameter didn't exist.

fix

Upgrade selfies: pip install --upgrade selfies

Warnings

breaking In v2.0.0, the encoder now defaults to strict=True, raising errors for SMILES violating bond constraints. Previously, invalid SMILES were silently allowed. ↓

fix Use encoder(smiles, strict=False) to mimic old behavior, or ensure input SMILES are valid.

deprecated Support for Python 3.5-3.6 dropped in v2.1.0. ↓

fix Upgrade to Python >=3.7.

Imports

encoder
```
from selfies import encoder
```
No common wrong import; just import the function.
decoder
```
from selfies import decoder
```
No common wrong import; just import the function.

Quickstart

Basic encode/decode a SMILES string.

from selfies import encoder, decoder

smiles = "CCO"  # ethanol
selfies_str = encoder(smiles)
print(selfies_str)  # [C][C][O]

smiles_back = decoder(selfies_str)
print(smiles_back)  # CCO