Scenario
Reason over the whole recipe object — components and weight fractions — rather than a single molecule, demonstrating that the platform understands formulations, not just structures.
Worked example. A waterborne or solvent-borne coating recipe, given as components plus weight percentages.
Agent workflow
- Read components and weight fractions from the recipe.
- Deterministic roll-up — VOC content per the 2004/42/EC definition, plus a flash-point / physical-property estimate from component data.
- Screening CLP mixture classification — apply component classifications with generic concentration limits and M-factors to produce a preliminary mixture hazard view.
Demo output
Recipe in → VOC value plus preliminary hazard flags out, with every number traceable to a specific component and rule.
Honesty line
This is a screening / pre-check, not a legally binding CLP determination. The engine does the arithmetic and shows its work; the human signs off. Over-claiming legal classification is exactly what a weak vendor would do — so the platform does not.
Deterministic vs LLM
- Deterministic — VOC and concentration-limit arithmetic.
- LLM — explanation of why a given flag fired.
Data sources
- Native (offline): CoolProp (thermophysical properties of known fluids); RDKit.
- Public reference layer: CLP Annex VI Table 3 (harmonised C&L); ECHA C&L Inventory; generic/specific concentration limits and M-factors (CLP Annex I/VI); the 2004/42/EC VOC limit table.
Limits
Mixture classification is a screening, not a binding determination. It is a starting point for a qualified human assessor.
Skills orchestrated by this use case
This scenario routes an AI agent through the following curated skills.
-
Solubility Formulation Screen
FormulationScreen solvent and co-solvent systems to improve solubility of a poorly soluble compound, ranking solvent blends and flagging formulation strategies (co-solvency, pH adjustment, surfactants) — orchestrating Paramus solubility-prediction and property tools.When to use: a user needs to raise the solubility of a hard-to-dissolve active.Multi-tool WORKFLOW with solubilization-strategy judgment. -
Formulation Property Optimization
FormulationOptimize a multi-component formulation to hit property targets by defining the component/ratio design space under mixture constraints, screening candidates, and optimizing the blend against multiple objectives — orchestrating Paramus mixture-DOE, property-prediction, and optimization tools.When to use: a user needs to find component ratios that meet several property targets at once.Multi-tool WORKFLOW with mixture-design judgment.
Tools it reaches for
- CoolProp thermophysical properties
- RDKit molecular descriptors
Browse the full deterministic layer in the tool browser.