Navigator

Overview

Run the DMC Navigator production workflow on your own machine, fully offline — a resumable propose → dock → ingest loop over make-on-demand chemical space.

Suggest an edit

Run the DMC Navigator production workflow on your own machine, fully offline.

The on-prem package contains no source code and builds nothing. It pulls a pre-built, obfuscated, license-gated container image from Deep-MedChem and wraps it in a small navigator command. Your molecules, run state, and license never leave your host.

Production updates arrive through the stable image channel. Deep-MedChem moves that tag only after a reviewed on-prem release passes tests and image smoke checks; you choose when to install it by running navigator update.

Principles

  • Offline — after the image is pulled the workflow runs with no network access. The one exception is navigator data (downloading an encrypted database from Deep-MedChem's public bucket); databases are also installable fully offline from a local mirror (see Databases).
  • Licensed — the image only runs with a valid license issued for this machine (see Activate your license).
  • Self-contained — run state lives in ./runs, inputs in ./inputs, installed databases in ./databases.

The workflow

Navigator drives a resumable optimization loop. The optimizer runs in the licensed container; you own and run the scoring step (e.g. Glide docking) in your own environment — Schrödinger is never inside the image.

init → propose a batch → you dock/score it → ingest scores → propose … → complete

propose is a deterministic function of (observation archive, seed, iteration), so a run is reproducible, resumable, and idempotent on re-ingestion.

Get started

On this page