Automated Data Reduction on VULCAN

Abstract

The automated data reduction on VULCAN is overhauled by using multithreading-capable python codes. Other than just creating reduced histogram data in GSAS format and reduce meta data as the prior mantid based routine, more automated services are included in the new reduction pipe line, including, generating GSAS-II format data, neutron and sample environment data slicing/chopping, plotting chopped meta data, matching calibration files based on sample environments or instrument configurations, reduction of VNb files, invoking AutoSAC, notification to Slack channel on completion of data reduction, accessible via Slack APP commands, marking on data readiness, notifying PI via email when mail-in proposal is complete, and parsing user’s request from run note. The whole program is coded through vibe-coding fashion, where an AI agent is used with multiple rounds of prompt and testing. Thanks to this, further improvements and modification can be done by employing an AI agent after it digests the routines or the markdown files. The current version has been delivered to VULCAN beamline for running/testing with a full cycle of data, where frequentl demands in either data size or runs are encountered, and strategic approaches were introduced for a more robust execution. While not included, the workflow allows further automated data analysis given a complete data analysis routines to be ready and tested in command line interface (CLI). Current version also can be invoked by using a specially constructed agent based on an open large language model (LLM), thanks to the CLI design.