UNPKG

tkyodrift

Version:

Lightweight CLI tool and library for detecting AI model drift using embeddings and scalar metrics. Tracks semantic, conceptual, and lexical change over time.

github.com/oslabs-beta/tkyo-drift

oslabs-beta/tkyo-drift

16 files

	Filename	Content Type	Size
	../
	DriftModel.js	text/javascript	19 kB
	captureSharedScalarMetrics.js	text/javascript	1.93 kB
	compareScalarDistributions.js	text/javascript	2.51 kB
	downloadTrainingData.py	application/octet-stream	495 B
	loadScalarMetrics.js	text/javascript	3.22 kB
	makeErrorLogEntry.js	text/javascript	994 B
	makeLogEntry.js	text/javascript	2.88 kB
	oneOffEmb.js	text/javascript	6.62 kB
	printLogCLI.js	text/javascript	5.16 kB
	printScalarCLI.js	text/javascript	6.81 kB
	pythonHNSW.py	application/octet-stream	5.69 kB
	pythonKMeans.py	application/octet-stream	1.71 kB
	pythonTrainingEmb.py	application/octet-stream	13.8 kB
	tkyoDriftSetTraining.py	application/octet-stream	2.27 kB
	tkyoDriftSetTrainingHook.js	text/javascript	2.23 kB
	writeSharedScalars.py	application/octet-stream	3.98 kB