AutoWriter + ProteinMakerApp

What it is

Automated protein purification is normally driven by a proprietary GUI where a scientist configures every step of a multi-hour run by hand — buffer flows, wash volumes, gradients, elutions, plate handling — and saves the result as a machine-specific XML protocol. That workflow doesn't scale to the kind of parameter sweeps and per-sample customization real research needs.

AutoWriter replaces the GUI drag-and-drop with a plain CSV. Rows describe the fluid steps; columns describe the parameters. Save the file, and a headless engine translates it into a hardware-executable XML protocol — ready to load into the robot. ProteinMakerApp is the Electron desktop wrapper around it: a polished editor + live protocol preview + one-click XML generation, all built for chemists who don't want to think about command lines.

The core engine (AutoWriter)

Pure Python (stdlib + win32file for Windows file locking), PyInstaller-packaged as a single standalone .exe. The main ProtocolBuilder is ~1,860 lines and handles the interesting orchestration work:

• Deck geometry. A DeckPlanner allocates well positions across a 4×5 deck grid (24-, 48-, 96-well plates and 50 mL tubes), scaling with sample count and auto-splitting volumes across wells when a plate can't hold the requested elution volume.
• Lag-smart pipe simulation. The instrument's fluid path has real volumetric latency — when buffer B displaces buffer A, there's a lag-volume "slug" of A left in the line. AutoWriter models this as a true FIFO: gantry moves are delayed by lag volume, syringe operations stay unchanged, and the last buffer of the run gets one extra lag-volume flush automatically.
• Per-sample parameters. Each of the 24 syringe lanes can run on independent flow rates and volumes in the same protocol — useful for gradient variation across a sample set.
• Aspirate-on-move. The XML step model doesn't natively support "aspirate while the gantry moves." AutoWriter splits each syringe step into an AspirateOnly phase that runs parallel with the gantry, then a DispenseOnly phase — a real throughput win on long protocols.
• Custom fractionation. A repeat-shorthand for elution collection: 3(5),2(4) expands to 5,5,5,4,4, one number per fractionation well.
• Named speed parameters. Every named speed ("app speed", "Wash A1 speed", etc.) is registered in the XML's <globalparameters> so the operator can tweak them from the instrument's UI without regenerating.

All of that gets serialized to a very specific XML dialect the robot understands, and the whole thing is verified against a canonical-diff test harness — generated XML is compared against golden files, with an --analyze mode that groups diffs by normalized pattern to catch systematic changes.

The desktop app (ProteinMakerApp)

An Electron + TypeScript app that calls AutoWriter's cli_bridge.py as a subprocess and gets JSON back {xml, timing, errors}. It gives chemists a proper editor experience — syntax-aware CSV, live protocol preview, estimated run duration — without dropping them into a terminal. The two projects share knowledge of the CSV schema and the Protein Maker XML format, but neither depends on the other at runtime: AutoWriter runs headless in any pipeline; the app is a nice-to-have front door.

Fast facts