Paramus — Polymer Formulation Triage

Overview

Take a full formulation and a target property profile, predict the profile, check additive compatibility, and return a keep/adjust/reject assessment with the binding trade-offs. This is the highest-level polymer skill — it composes the other polymer screens into one decision.

Guidance vs execution: every property and compatibility signal comes from tested Paramus tools. Do not blend properties or invent mixing rules in ad-hoc code.

When to Use

• "Here's my formulation — does it hit the targets, and what breaks?"
• Comparing formulation variants against one property profile Do not use for: process/manufacturing simulation (hand off to the engineering workflow).

Workflow

Base polymer + additives/fillers/plasticizers + target property profile
  ↓ 1. Parse & validate all components   → Paramus polymer + cheminformatics tools
  ↓ 2. Predict target properties          → compose Tg / mechanical / solubility screens
  ↓ 3. Check additive compatibility       → solubility/miscibility signals
  ↓ 4. Fuse vs profile                     → per-target pass/fail + trade-off surface
  ↓ 5. Assess & report                     → keep / adjust / reject + binding constraints

Procedure

1. Discover the component tools with search/get_schema; reuse the Tg, mechanical, and solubility screens as sub-steps rather than re-deriving them.
2. Parse and validate every component; quarantine anything that fails.
3. Predict each property in the target profile, carrying uncertainty and model versions throughout.
4. Screen plasticizer/filler compatibility via solubility signals; flag likely phase separation.
5. Fuse results against the profile; identify which target is the binding constraint.
6. Report an overall verdict (keep/adjust/reject), the binding trade-off, and concrete adjustment levers (e.g. "raise plasticizer → gains flexibility, loses Tg margin").

Domain Judgment

• The binding constraint matters more than the average score — name it explicitly.
• Compatibility failure vetoes a good property profile — a phase-separated blend won't deliver.
• Propagate uncertainty from every sub-screen; the verdict inherits the weakest confidence.

Fallbacks

• A sub-screen model unavailable → deliver a partial verdict, clearly scoped to what was assessed.
• Endpoint unreachable → stop and report; no local blending rules.

Promotion Note

This composite is a strong promotion candidate: if "triage a formulation against a profile" proves frequent and stable, the fused verdict step earns implementation as a single tested Paramus tool.