curieai

The official JavaScript/TypeScript SDK for Curie — inference infrastructure for scientific AI.

Run biology, chemistry, and physics AI models through one unified API.

Installation

npm install @curieai/curieai
# or
pnpm add @curieai/curieai
# or
yarn add @curieai/curieai

Quick start

import { Curie } from '@curieai/curieai';

const client = new Curie({ apiKey: 'sk-...' });

// Biology: Predict protein structure
const fold = await client.fold('MKTIIALSYIFCLVFA...');
console.log(`Confidence: ${fold.confidence}%`);

// Chemistry: Analyze molecular properties
const molecule = await client.analyzeMolecule('CCO');
console.log(`MW: ${molecule.properties.molecularWeight}`);
console.log(`Drug-like: ${molecule.isDrugLike}`);

// Physics: Predict atomic forces
const forces = await client.predictForces({
  elements: ['H', 'H', 'O'],
  positions: [[0, 0, 0], [0.96, 0, 0], [0.24, 0.93, 0]],
});
console.log(`Energy: ${forces.energy} eV`);

Environment variable

export CURIE_API_KEY="sk-..."

import { Curie } from '@curieai/curieai';
const client = new Curie(); // reads from CURIE_API_KEY automatically

Biology

Fold proteins

const result = await client.fold('MKTIIALSYIFCLVFA...');
console.log(`Confidence: ${result.confidence}%`);
console.log(result.pdb); // Full PDB structure

Generate protein embeddings

const result = await client.embed('MKTIIALSYIFCLVFA...', {
  returnPerResidue: true
});
console.log(`Embedding dim: ${result.dim}`);

Design protein sequences

const result = await client.design(pdbString, {
  numSequences: 5,
  temperature: 0.1
});
console.log(result.best.sequence);

Chemistry

Embed molecules

const result = await client.embedSmiles('CCO');
console.log(`Embedding dim: ${result.dim}`);

Analyze molecular properties

const result = await client.analyzeMolecule('CCO');
console.log(`MW: ${result.properties.molecularWeight}`);
console.log(`LogP: ${result.properties.logP}`);
console.log(`Lipinski compliant: ${result.isDrugLike}`);

Translate molecules

// Text to SMILES
const smiles = await client.translateMolecule('aspirin', 'text-to-smiles');
console.log(smiles.output);

// SMILES to text
const text = await client.translateMolecule('CCO', 'smiles-to-text');
console.log(text.output);

Physics

Predict atomic forces

const result = await client.predictForces({
  elements: ['H', 'H', 'O'],
  positions: [[0, 0, 0], [0.96, 0, 0], [0.24, 0.93, 0]],
});
console.log(`Energy: ${result.energy} eV`);
console.log(`Forces: ${result.forces}`);

Models

| Category | Model | Method | Slug | |----------|-------|--------|------| | Biology | ESMFold v1 | client.fold() | esm/esmfold-v1 | | Biology | ESM-2 650M | client.embed() | meta/esm2-650m | | Biology | ProteinMPNN | client.design() | bakerlab/proteinmpnn | | Chemistry | ChemBERTa-2 | client.embedSmiles() | seyonec/chemberta-2 | | Chemistry | RDKit | client.analyzeMolecule() | rdkit/cheminformatics | | Chemistry | MolT5 | client.translateMolecule() | google/molt5-large | | Physics | MACE-MP-0 | client.predictForces() | cambridge/mace-mp-0 | | Physics | NequIP | client.predictForces() | mir-group/nequip | | Physics | DeePMD | client.predictForces() | deepmodeling/deepmd |

Docs

Full API reference: curie.sh/dashboard/docs