# The Agent Memory Benchmark's dataset is a byte-exact published artefact, and this file
# exists because nothing declared that.
#
# `tests/test_agent_memory_bench.py` regenerates the dataset and compares the bytes
# against the committed copy, and the CI benchmark job runs `diff -r` over the same pair.
# The generator writes LF explicitly. Git for Windows checks text files out as CRLF by
# default, so without this line the committed copy and the generator's own output differ
# by one byte per line — on Windows, and only on Windows. That is how it was found: every
# other job in the matrix passed.
#
# Scoped to the dataset rather than declared repository-wide. A blanket `* text=auto
# eol=lf` would change what every contributor's working tree looks like, which is a
# larger decision than this change gets to make. These three files are the ones whose
# bytes are asserted on.
benchmarks/agent_memory/datasets/** text eol=lf
