Pure familiarity scoring and decay.
Familiarity rises with how often you open and edit a file (with diminishing returns) and decays toward zero the longer you go without touching it. A 0.0..1.0 score maps to a 0..4 heat bucket the file tree renders as a tint (0 = coldest/unfamiliar, 4 = warmest/familiar).
ponytail: open + edit counts and recency approximate "how well you know
this file". Real dwell time (focus duration) is a stronger signal; add it
if the buckets feel too coarse.
Summary
Functions
Maps a 0.0..1.0 familiarity score to a 0..4 heat bucket.
Familiarity in 0.0..1.0 for stats evaluated at now (unix seconds).
Types
@type stats() :: %{ first_seen: integer(), last_seen: integer(), open_count: non_neg_integer(), edit_count: non_neg_integer() }
Functions
@spec level(float()) :: 0..4
Maps a 0.0..1.0 familiarity score to a 0..4 heat bucket.
Familiarity in 0.0..1.0 for stats evaluated at now (unix seconds).
Activity saturates (so the 100th edit adds little), then decays by how long
it has been since last_seen.