feat: increase MetricSamples to 100 points (~15 min resolution)

Modern LLMs have 100k+ token contexts. 100 samples over 24h gives
~15 minute resolution while adding minimal token overhead.

This lets the LLM see fine-grained patterns, short spikes, and
accurately distinguish anomalies from normal behavior.
This commit is contained in:
rcourtman 2025-12-21 22:25:54 +00:00
parent 818fbffd66
commit 49fc86c70b

View file

@ -486,9 +486,9 @@ func (b *Builder) computeGuestMetricSamples(guestID string) map[string][]MetricP
if len(points) < 3 {
continue
}
// Downsample to ~24 points (roughly hourly over 24h)
// This gives the LLM good resolution to spot patterns and spikes
sampled := DownsampleMetrics(points, 24)
// Downsample to ~100 points (~15 min resolution over 24h)
// Modern LLMs have 100k+ token contexts - we can afford detailed history
sampled := DownsampleMetrics(points, 100)
if len(sampled) >= 3 {
samples[metric] = sampled
}