Document real-world peak detector validation

PROJECT_KNOWLEDGE.md

@@ -26,7 +26,8 @@ after a value falls below `drop_ratio * current_peak`, it emits that maximum
 only after actual subsequent below-threshold observations support a continuous
 `hold_seconds` interval. `min_peak` is a generic detector setting, along with
 `drop_ratio` (0 < ratio < 1), `hold_seconds` (>= 0), and an explicit positive
-`max_sample_gap_seconds`. Elapsed time, never a sample count, determines
+`max_sample_gap_seconds`. They are selected per signal, rather than being
+globally fixed process constants. Elapsed time, never a sample count, determines
 confirmation. An interval longer than the configured maximum sample gap
 restarts a reset candidate, so a timestamp gap is not
 continuous-below evidence. Reset values can be negative rather than zero,
@@ -35,10 +36,25 @@ The maximum gap is required rather than inferred or defaulted, making each
 source's continuity assumption explicit.
 
 K7 roll length, variable `bf81c547-dccf-4709-aee2-0f78366d1dfc` (m), is the
-first real-world validation case. The tested capture is continuous on a
+clean reference validation case. With `min_peak=10`, `drop_ratio=0.75`,
-10-second grid through its latest available sample; its shorter-than-requested
+`hold_seconds=60`, and `max_sample_gap_seconds=20`, its continuous 10-second
-result was caused by an end time queried in the future, not observed time-series
+capture produced exactly four plausible peaks of approximately 49.98 m. Its
-gaps. The raw capture remains ignored and deterministic tests use synthetic
+shorter-than-requested result was caused by an end time queried in the future,
+not observed time-series gaps.
+
+Bento 2 roll weight is the second, more realistic robustness validation case:
+machine `6201f3de-b940-45d7-930c-1481c5ae9e1d`, variable
+`059c0015-9aff-4270-a4ce-f21b83f99378` (kg). Its observed cadence is 10
+seconds, but it has long plateaus, reset values roughly 300–470 kg, and
+materially varying peaks. With `min_peak=800`, `drop_ratio=0.75`,
+`hold_seconds=60`, and `max_sample_gap_seconds=20`, the extended capture
+yielded 13 plausible completed peaks: 936, 919, 935, 930, 930, 920, 909,
+1195, 921, 930, 939, 912, and 914 kg (from 00:46:50Z through 03:06:30Z on
+2026-09-04). The approximately 1195 kg peak is retained as a real
+process/mechanical condition, not special-cased as a defect; it is emitted
+only after its following reset has enough below-threshold observations. A
+reset is not expected to approach zero, and peak values are not assumed to be
+constant. Raw captures remain ignored and deterministic tests use synthetic
 fixtures only.
 
 ## Central domain context

README.md

@@ -69,20 +69,34 @@ roll weight, and comparable sawtooth/batch signals. It retains the current
 maximum and emits it only after the value is below
 `drop_ratio * current_peak` continuously for `hold_seconds`. Configuration is
 `min_peak`, `drop_ratio` (strictly between 0 and 1), non-negative
-`hold_seconds`, and positive `max_sample_gap_seconds`. The hold uses elapsed
+`hold_seconds`, and positive `max_sample_gap_seconds`. These are per-signal
+configuration parameters, not globally fixed process constants. The hold uses elapsed
 timestamps, never a sample count; an interval longer than the configured
 maximum sample gap restarts a reset candidate,
 so a timestamp gap alone is not evidence that a signal remained below threshold.
 `max_sample_gap_seconds` is required rather than defaulted, so the source's
 continuity assumption is explicit for each detector configuration.
-Reset values can be negative and need not be zero. Equal timestamps are
+Reset values need not approach zero (and can be negative), and peak magnitude
-accepted in arrival order but add no elapsed hold time; backwards timestamps
+can vary materially between cycles. Equal timestamps are accepted in arrival
-are rejected.
+order but add no elapsed hold time; backwards timestamps are rejected.
 
-The first real-world validation case is K7 roll length (m), variable
+Real-world validations use ignored local captures and are not part of the test
-`bf81c547-dccf-4709-aee2-0f78366d1dfc`. When the ignored local capture is
+suite. K7 roll length (m), variable `bf81c547-dccf-4709-aee2-0f78366d1dfc`, is
-present, run `python scripts/validate_k7_roll_length.py` to inspect detected
+the clean reference case: its 10-second capture produced exactly four plausible
-peaks without adding the raw capture to tests or version control.
+peaks of about 49.98 m with `min_peak=10`, `drop_ratio=0.75`,
+`hold_seconds=60`, and `max_sample_gap_seconds=20`. Run
+`python scripts/validate_k7_roll_length.py` when that capture is available.
+
+Bento 2 roll weight (kg), machine `6201f3de-b940-45d7-930c-1481c5ae9e1d`,
+variable `059c0015-9aff-4270-a4ce-f21b83f99378`, is the more realistic
+robustness case. Its observed 10-second signal has plateaus, reset values of
+roughly 300–470 kg, and materially varying peak heights. With
+`min_peak=800`, `drop_ratio=0.75`, `hold_seconds=60`, and
+`max_sample_gap_seconds=20`, the extended capture produces 13 plausible
+completed peaks, including a legitimate approximately 1195 kg cycle. Run
+`python scripts/validate_bento2_roll_weight.py` when the ignored capture is
+available. Neither capture should be added to version control or automated
+tests.
 
 `docker compose up -d timescaledb` is an optional local database design for a
 future persistence milestone. It is not required for the bootstrap tests.

docs/architecture.md

@@ -52,6 +52,9 @@ This is calculation-domain logic, with no ENLYZE transport dependency.
 The detector uses timestamps rather than a sample count. An interval longer
 than the configured positive `max_sample_gap_seconds` restarts a reset candidate,
 so a timestamp alone does not establish that a signal was continuously below
-threshold. Equal
+threshold. These parameters (`min_peak`, `drop_ratio`, `hold_seconds`, and
-timestamps are processed in arrival order but contribute no elapsed time, and
+`max_sample_gap_seconds`) are selected per signal rather than treated as global
-backwards timestamps are rejected. Reset values are not assumed to be zero.
+process constants. Equal timestamps are processed in arrival order but
+contribute no elapsed time, and backwards timestamps are rejected. Reset values
+are not assumed to approach zero, and peak values may vary materially between
+cycles.

docs/roadmap.md

@@ -31,7 +31,10 @@ tests against sanitized fixtures.
 The generic configurable sawtooth detector is implemented: it retains the
 current maximum, closes a cycle after a configurable below-peak fraction has
 actual sample support for a configurable elapsed duration, and emits one
-completed peak. Its pure state can later be persisted for incremental use.
+completed peak. Its pure state can later be persisted for incremental use. It
+has been validated with a clean K7 roll-length reference signal and a more
+variable Bento 2 roll-weight robustness signal; the latter includes plateaus,
+non-zero resets, and varying peak heights without detector special cases.
 
 ## Later
 

scripts/validate_bento2_roll_weight.py

@@ -0,0 +1,82 @@
+#!/usr/bin/env python3
+"""Validate the peak detector against an ignored local Bento 2 capture.
+
+This developer utility reads no network data and does not modify the capture.
+Run it from the repository root when
+data/raw/enlyze/bento2-roll-weight-extended.raw.json is available.
+"""
+
+from __future__ import annotations
+
+import json
+from datetime import datetime
+from pathlib import Path
+
+from production_analytics.calculations import (
+    PeakCycleDetector,
+    PeakDetectorConfig,
+    TimeSeriesSample,
+)
+
+
+CAPTURE = Path("data/raw/enlyze/bento2-roll-weight-extended.raw.json")
+VARIABLE_UUID = "059c0015-9aff-4270-a4ce-f21b83f99378"
+EXPECTED_PEAKS = (
+    ("2026-09-04T00:46:50+00:00", 936.0),
+    ("2026-09-04T00:57:10+00:00", 919.0),
+    ("2026-09-04T01:07:20+00:00", 935.0),
+    ("2026-09-04T01:17:20+00:00", 930.0),
+    ("2026-09-04T01:27:20+00:00", 930.0),
+    ("2026-09-04T01:37:30+00:00", 920.0),
+    ("2026-09-04T01:48:00+00:00", 909.0),
+    ("2026-09-04T02:14:20+00:00", 1195.0),
+    ("2026-09-04T02:24:10+00:00", 921.0),
+    ("2026-09-04T02:35:40+00:00", 930.0),
+    ("2026-09-04T02:46:10+00:00", 939.0),
+    ("2026-09-04T02:56:20+00:00", 912.0),
+    ("2026-09-04T03:06:30+00:00", 914.0),
+)
+
+
+def main() -> int:
+    if not CAPTURE.is_file():
+        print(f"Bento 2 capture is not available: {CAPTURE}")
+        return 1
+
+    payload = json.loads(CAPTURE.read_text(encoding="utf-8"))
+    data = payload["response"]["body"]["data"]
+    value_index = data["columns"].index(VARIABLE_UUID)
+    samples = (
+        TimeSeriesSample(
+            datetime.fromisoformat(record[0].replace("Z", "+00:00")),
+            float(record[value_index]),
+        )
+        for record in data["records"]
+        if record[value_index] is not None
+    )
+    detector = PeakCycleDetector(
+        PeakDetectorConfig(
+            min_peak=800.0,
+            drop_ratio=0.75,
+            hold_seconds=60.0,
+            max_sample_gap_seconds=20.0,
+        )
+    )
+    peaks = list(detector.process_many(samples))
+    observed = tuple((peak.peak_timestamp.isoformat(), peak.value) for peak in peaks)
+
+    print(f"Detected peaks: {len(peaks)}")
+    for peak in peaks:
+        print(
+            f"peak={peak.value:.0f} kg at {peak.peak_timestamp.isoformat()} "
+            f"confirmed={peak.confirmed_at.isoformat()}"
+        )
+    if observed != EXPECTED_PEAKS:
+        print("Validation failed: detected peaks differ from the expected extended-window result.")
+        return 2
+    print("Validation passed: detected peaks match the 13 expected completed peaks.")
+    return 0
+
+
+if __name__ == "__main__":
+    raise SystemExit(main())