Implemented the baseline/noise and SQLite growth improvements.

README.md

@@ -237,6 +237,29 @@ The continuous monitor also stores a local SQLite behavior baseline at
 five-minute windows. Historical rate and hitcount-rate deviations then contribute
 to its anomaly score. Set `FGAI_BASELINE_DB` to use another location.
 
+SignalScope prunes old baseline buckets during each monitor cycle. The defaults
+keep 14 days of buckets and dedupe history, prune stale one-off categorical
+values after 7 days, and cap high-cardinality values per stream/entity/field.
+Tune these in the dashboard or in `state/fgai-config.json`:
+
+```json
+{
+  "baseline_retention_days": 14,
+  "baseline_value_retention_days": 7,
+  "baseline_max_values_per_field": 2000
+}
+```
+
+If an existing baseline database has already grown large, stop the monitor and
+run a manual prune plus SQLite compaction:
+
+```bash
+signalscope baseline-maintenance --baseline-db state/fgai-baseline.sqlite3 --retention-days 14 --value-retention-days 7 --max-values-per-field 2000 --vacuum
+```
+
+`VACUUM` can take time on a large database and should not be run while the
+monitor is actively writing.
+
 Analyze local logs:
 
 ```bash