Changelog: 0.98 to 0.102

Archived CAMBER releases 0.98.0 to 0.102.0, in the same format as the current changelog, which holds the recent releases. The changelog index lists every archive.

[0.102.0] — 2026-10-10

0.102: follow-ups from 0.101, sample charts across the docs, a guide to the lab, and an access token on every lab route (#110–#121, #127). The suggester's unitless range check no longer overturns a strong name on a dirty series (#110); the air-economizer workbook exercise teaches the economizer low-limit lockout and the capstone adopts it (#111); and the RCx report ranks uncosted issues of equal severity by confidence before the issue key, so the capstone's stuck-damper issue ranks first again (#112). The docs site gains charts and screenshots rendered by CAMBER's own chart code (#113) and a step-by-step guide to camber lab (#121). Evidence charts shade exactly what each rule flags and name roles with their units (#114–#117, #119). A stuck supply-air sensor on a scheduled fan is now flagged (#118). Behaviour change (#120): free_cooling_missed no longer judges fan-off hours, so missed shares rise on units whose fan runs only when occupied; the air-economizer and capstone answers move with it. Security (#127): camber lab now requires a per-run access token on every route; open the URL it prints.

Added

• Sample charts and screenshots across the docs (#113). The docs site and README had no images. scripts/docs_figures.py renders 36 figures and three README thumbnails into docs/img/ with CAMBER's own chart code: the camber.charts primitives, the rules' evidence hooks and the drift detectors' frozen-baseline bands. The data is deterministic synthetic data (camber.synth, the ahusim, driftsim, vavsim and pumpsim simulators, the DCV simulator and small seeded generators); no downloaded dataset is read. A fixed style, the Agg backend, a fixed dpi and size and PNGs with no metadata keep reruns byte-stable on one machine.
• One figure per chart pattern in Visualization (#113), next to its code example: readiness ribbon, fault-annotated trend, load carpet (with a new short section), quality dashboard, OAT scatter, diagnostic scatter, the fitted drift band, rule evidence, cohort small multiples, M&V savings, load profile and load-duration curve, box by hour, and a new CUSUM and energy-signature section.
• A hero figure on the analytics pages (#113): free cooling, chiller, AHU, VAV and pump drift, M&V (the change-point fit and the savings chart), sensor health and schedule inference.
• The evidence chart each workbook exercise discusses (#113), with a caption that marks it as a synthetic illustration rather than the exercise's dataset: the stuck outdoor-air damper's OA fraction, the leaking cooling valve, the reheat penalty, chiller kW/ton, tower approach, the change-point baseline and the rest. The exercises' text and answers are unchanged.
• Screenshots (#113) of the RCx report, the site report, the camber lab catalog page and the camber serve trend viewer, on a synthetic two-AHU site. --screenshots takes them with headless Google Chrome, serving on 127.0.0.1 and stopping the servers afterwards; it is optional, as CI has no browser.
• A small gallery in the README (#113): three thumbnails linking to the docs pages.
• scripts/docs_figures.py --check (#113), run by scripts/gates.sh and by tests/test_docs_figures.py. It fails when an image referenced from docs/**/*.md or README.md is missing or over 150 KB, when a docs image has no alt text or caption, when a generated figure is missing, unreferenced or stale, or when docs/img/ exceeds 5 MB.
• Using the lab, a step-by-step guide to camber lab (#121), early in the docs navigation and linked from the README, DATASETS.md, CLI.md and the workbook. A first session on the open ornl-frp-ops dataset (install, start, pick a dataset, Fetch & ingest, trends, the report's top finding, the exercise, and the matching CLI commands), then how to read the trend viewer and the report, where the cache and the store live and how to clean them up, a classroom setup (pre-fetching for an offline room, manual downloads with --from-dir, a portfolio workspace, what loopback-only means on a shared computer) and troubleshooting. scripts/docs_figures.py --screenshots adds five walkthrough screenshots; unlike the other figures they show a real open dataset (credited in their captions), read from a local copy through the verified --from-dir path and never downloaded, and they are skipped when no local copy is present.

Changed

• Workbook air-economizer teaches the economizer low-limit lockout (#111). A new step 5 has learners read free_cooling_missed's warn on the fault-free unit (17.5 %). They then find from the trends and the unit's sequence that 899 of its 965 missed hours (93 %) fall below the documented 33.8 °F lockout, where the damper is held at its minimum by design. They set low_limit_f 33.8 and re-run: the fault-free unit reads ok (1.96 %), and the dampers stuck at 10 % and 25 % stay a fault (49.6 %). The exercise config leaves the low limit out on purpose, and its _comment says so. The declaration pins both readings, and the page (a new step and question, with later ones renumbered) and the instructor key are updated. The stand-in now models the lockout (four cold days below it), so both readings are also checked offline.

• Workbook capstone adopts the same low limit (#111), which intentionally changes its answers. free_cooling_missed in the capstone config now sets low_limit_f 33.8, with a basis, to match the dataset template and the economizer exercise. On the real data the control unit reads ok (2.0 %, was a 17 % warn) and loses its spurious "enable the economizer" issue and walk-down item. The onset unit reads a warn at 21 % (was a 30 % fault): six healthy months dilute a 1 July onset in a year-long share. Its damper issue keeps its cause heading and confidence H, and it now has stronger cause evidence (commanded_open_pct 89.9 % of 640 h, was 40.3 % of 673 h). With #112 it ranks first again: among uncosted issues of one severity, confidence now decides the order before the issue key. Drift is unchanged (-83, fault). The declaration, the page (questions 1 and 2, a caveat) and the key are updated.

• RCx: uncosted issues of one severity rank by confidence before the issue key (#112). link_findings used to order uncosted issues of one tier by issue key, a fixed but arbitrary tie-break. They now rank by input-trust confidence, H before M before L, and then by key. Severity, conditional-last and costed-first by dollars are unchanged, and costed issues keep their dollar order. This intentionally changes the RCx ranking where uncosted issues of one tier differ in confidence. On the 15 catalog run templates (default subsets), 4 of 96 issues change rank, in 2 reports, and each pair swaps. In lbnl-ddahu, DMPRStuck_OA_100's outdoor_air_fraction fault (H) moves from 2 to 1 above DMPRStuck_OA_0's g36_afdd fault (M). In ornl-frp-ops, base_heating's night_weekend_setback fault (H) moves from 2 to 1 above sb_heating's compressor_short_cycle fault (M). Both reports keep their chosen week; only its evidence score, which weights issues by 1/rank, changes. The RCx golden is unchanged: its synthetic site has no uncosted tie with mixed confidence. docs/RCX-REPORT.md describes the new order.

• Workbook capstone: the damper issue ranks first again (#112). On the real data, the onset unit's stuck-damper issue (an uncosted warn, confidence H) now ranks 1, above the onset unit's supply-air reset and the control's static-pressure reset (uncosted warns, confidence M). It ranked 3 in #111. The declaration asserts rank 1 in both modes. The page's question 1 asks what puts the issue first when nothing is costed. The instructor key explains the confidence order, lists the damper first in the walk-down, and rewords the common mistake about reading an uncosted rank as a measure of size.
• The unitless range check no longer overturns a strong name on a dirty series (#110). On the time-series path (the default when a series is passed, from 0.101), a point with no declared unit is range-checked in every plausible unit. Two changes act on that check only:
• an exact zero that the role cannot read in any plausible unit (a 0 °C room, 0 ppm CO2) is a dropout and is left out; an all-zero series gives no range verdict;
• when the role is the name's best lexical match and scores at least 0.6, the check keeps at least STRONG_NAME_RANGE_FLOOR (0.75) of its score. A series with no reading inside the bounds in any unit (a -999 dead channel) is still demoted in full.

This intentionally changes suggestions when a series is passed without a unit. A declared unit, the use_timeseries=False opt-out and the name-only default are unchanged. At default settings with the series passed, top-1 0.101 → 0.102: - held-out catalog names, real names, simulated LBNL sets, BTS anonymised: unchanged (73.0, 95.7, 74.4, 46.2 %); - BTS Brick-class names (upper bound): 89.4 → 93.4 % (name alone 93.2 %); - the five synthetic styles: 70.4-89.9 → 73.0-93.9 %; ahu_03_supply_temp 89.9 → 93.9 % (name alone 93.9 %).

No pool loses top-1 or top-3. Before and after tables, and the 4 synthetic points that now miss, are in docs/MAPPING-ASSIST.md. - camber.mapping_assist.STRONG_NAME_RANGE_FLOOR (new, provisional): the floor above, a module constant like WATER_AIR_PENALTY.

None of these figures is a gated benchmark.

Fixed

• camber lab needs its access token on every route (#127, security). The lab binds 127.0.0.1, but every GET was unauthenticated and the page carried the CSRF token, so any other account on a shared computer could load /lab, read the token, queue fetches and ingests, and read the ingested data, reports and trends. Each run now makes a random access token, separate from the CSRF token, and prints the launch URL http://127.0.0.1:<port>/lab?token=.... A valid ?token= sets an HttpOnly, SameSite=Strict session cookie and redirects to the URL without it; every request, GET or POST, the delegated /ui, /facilities, /points and /history included, then needs the cookie (or Authorization: Bearer <token>), else 401. POSTs still need the CSRF header and the Origin / Host checks. The URL is also saved to a 0600 launch file, lab-<port>.url, in a 0700 folder under $XDG_RUNTIME_DIR/camber or ~/.config/camber; a folder or file that is group- or world-writable or owned by another account is refused. The lab opens no browser, so the token never appears in ps. New in the provisional API: LabApp(access_token=...), LabApp.launch_url() and camber.lab.announce. SECURITY.md §11 describes the model and its residual risks; LAB.md, CLI.md, DATASETS.md, the README and the workbook pages now say to open the URL camber lab prints.
• A stuck supply-air sensor on a scheduled fan is now flagged (#118). With the fan gate on, a fan-dependent role's flat run was broken at every fan-off span, so a sensor frozen for four days on a 13 h schedule never reached the 24 h stuck limit. A run now joins across a fan-off span when the reading held the same value through it, and its length is the fan-on hours it covers, as the plant gate already counts running hours. A reading that moves with the fan off still starts a new run, so a supply air held at setpoint each day is not flagged.
• The simultaneous_heat_cool evidence chart uses the rule's threshold on the percent scale (#114). no_simultaneous_template assumed 0–1 valves (active=0.05), so on CAMBER's percent valves it shaded points from 0.05 % open. It now defaults to active=5.0, y_max=100.0 with "(%)" axes, and the rule's evidence passes its own flagged samples (occupied, both valves above 5 %, less the dehumidification-with-reheat classes it does not count) through a new violating= argument of diagnostic_scatter. A test checks that the shaded samples equal the rule's. economizer_template still reads its damper as 0–1.
• Evidence axes and carpet colour bars name the role and its unit (#115). The carpet colour bar always read "Load (kW)", and scatter axes showed raw role names such as COND_APPROACH_TEMP. Evidence carpets, OAT scatters, diagnostic-axis fallbacks, drift-band axes and multi-trend legends now read e.g. "cond approach temp (°F)", with the unit from camber.api.ui.role_units, the trend viewer's source. A multi-trend whose series share a unit shows it on the y axis. The site report's carpet colour bar names its column.
• savings_chart keeps the unit's case (#116). ylabel="kWh" rendered "Cumulative kwh". The default label is now "energy".
• Cohort small multiples get concise date ticks (#117). cohort_small_multiples panels use matplotlib's concise date formatter (at most four ticks), so multi-week dates no longer overlap. The docs-figure workaround is removed.
• Seven more rules shade their flagged samples in their evidence (#119). Each has a violation_mask(frame) built from what the rule counts:
• leaking_valve: the fan-on, both-valves-shut samples counted in hw_leak_pct / chw_leak_pct;
• static_pressure_reset: when not resetting, the judged setpoint samples on days it held;
• dcv_verification: high-demand samples the OA did not answer (static or uncorrelated), plus the setpoint-breach, below-floor and unventilated samples;
• co2_ventilation: the occupied samples counted as under-ventilated, or as over-ventilated when that is the finding;
• cooling_tower_approach: an approach-against-wet-bulb scatter of the judged samples, the ones above design + 3 °F in red;
• chw_plant_reset: low-ΔT running hours where ΔT is judged, and CHWST held at or below 46 °F when there is no reset;
• boiler_short_cycle: each start on a day with at least max_starts_per_day starts.

The analysis functions behind them take an optional masks_out dict; their results are unchanged. A one-sample shaded span in a multi-trend is now drawn to the next sample instead of as a hairline.

No verdict or finding changes. In the RCx report these issues, and simultaneous_heat_cool, now report violation hours instead of "not measured", and the evidence-mode representative week can move to the top issue's fault week. The synthetic RCx golden's week moves from 2026-03-09 to 2026-03-16, the planted reheat week of its top issue. The workbook cooling-tower and CHW-reset figures are now drawn from those rules' evidence. - free_cooling_missed judges fan-on hours only, and its evidence plots only the hours it judges (#120). The rule had no fan gate: overnight and weekend hours with the supply fan stopped counted as free-cooling weather, so they diluted the missed share, and a cooling-valve output parked open with the fan off read as missed free cooling. It now uses the fan-on gate the other air-side rules use (camber.schedules.fan_on_mask: fan status, else speed, else airflow), with a fan_gate parameter (default on) and the fan_gate and n_masked_fan_off metrics. A unit with no fan signal is judged ungated, as before, and names supply_fan_status in _missing_optional. The evidence chart is now a diagnostic scatter of the judged samples (fan on, OAT inside the free-cooling window) with the missed ones in red; integrated-economizer samples are drawn but never red, so the chart's out-of-band share is missed_pct (the rule's mask is what the chart shades, through #114's violating=), and the axes are labelled by role and unit (#115). It used to scatter every sample, fan-off and warm-weather hours included. No gated benchmark moves.

This intentionally raises the missed share on units whose fan stops, and the workbook answers move with it (real lbnl-sdahu data; the fan runs only in occupied hours there): - air-economizer: the fault-free unit is a fault at 34.1 % (was a 17.5 % warn); its 965 missed hours, 899 of them below the lockout, are unchanged, but the fan-off hours no longer dilute them. With low_limit_f 33.8 it reads ok at 3.63 % (was 1.96 %), with 1,014 hours set aside (was 2,141). The dampers stuck at 10 % and 25 % miss free cooling in 100 % of the fan-on free-cooling hours, with and without the lockout (were 49.4 % and 49.6 %). The page now frames step 5 as the healthy unit reading as a full fault until the lockout is set. The Irish unit trends no fan signal and is unchanged (8.0 %, ok). - capstone: the onset unit is a fault at 41.7 % (was a 21 % warn) and the control ok at 3.6 % (was 2.0 %); commanded_open_pct is 90.3 (was 89.9) on the same 640 h. The damper issue is now the report's only fault and still ranks first, with confidence H; the rest of the ranking is unchanged.

The declarations, the two exercise pages and the instructor key are updated.

[0.101.0] — 2026-10-04

0.101: a seasonal OA damper minimum, data-led point-role suggestions by default, and a walk-down that follows the heading cause (#105–#109). g36_afdd takes a seasonal OA damper minimum, oa_damper_min_by_month (#105). The lbnl-ddahu template uses it: 45 %, and 28 % in June to August. The unit's winter heating hours return to OS#1 there, and its winter hours idling at the minimum are no longer read as free cooling. #105 has a limit: FC6 still cannot catch the dual-duct damper stuck shut (DMPRStuck_OA_0) on this data. Outside the summer that run's damper command saturates open, so its hours are OS#5, where FC6 does not apply. The run is still caught, through FC8, FC10, FC11 and FC12. The time-series point-role suggester gains a setpoint guard and a range check for points with no unit (#106). Default change (#107): suggest_roles, review_unmapped and review_bacnet now read the data whenever a series is passed; use_timeseries=False opts out and gives the 0.100 output byte for byte. With a series passed, top-1 rises against 0.100 on every evaluation pool. Against the name alone, pools with strong names dip slightly on dirty series: BTS Brick-class names read 89.4 % with the series against 93.2 % from the name alone, because site C's negative airflows and dropouts fail the range check. docs/MAPPING-ASSIST.md advises use_timeseries=False, or cleaning the series, where the names are already reliable. The RCx walk-down's equipment item now follows the cause that heads the issue (#108), which changes the two re-headed catalog issues (lbnl-ddahu DMPRStuck_OA_0, lbnl-sdahu damper_stuck_075). The open-fdd cross-check leaves the per-verdict list out of its JSON by default (#109). It was re-run once on the merged code with all three engines: only CAMBER's lbnl-ddahu verdicts change, as #105 reported, and every engine's any-FC result is unchanged. The synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmark gates did not move.

Changed

• The walk-down's equipment item follows the cause that heads the RCx issue (#108). Since 0.100 (#101) a member finding's equipment-level cause can head an issue, but the "Verify on site" equipment item still followed the root finding. camber.walkdown.site_checks takes heading_causes={issue key: (rule, cause key)} (additive, provisional), and the RCx report passes the cause it chose for each re-headed issue. That issue's equipment item is then the member rule's template for that cause, with the member rule's references; an issue the root heads keeps the root's item. This intentionally changes the Verify on site section for re-headed issues only. On the 15 catalog run templates (default subsets), 2 of 98 issues change, the two #101 re-headed: lbnl-ddahu DMPRStuck_OA_0 and lbnl-sdahu damper_stuck_075 now ask for the outdoor-air damper's blades, linkage and actuator under a full open and close command, not the minimum-position setting or the damper on a hot hour. The RCx golden is unchanged (its synthetic site has no re-headed issue).
• open-fdd cross-check: the per-verdict list is left out of the JSON by default (#109). run_crosscheck.py now omits the per-verdict list and collapses the per-equipment "not evaluated" reasons to counts unless --keep-verdicts is passed; a month-window run otherwise writes about 200k more lines. --omit-verdicts is still accepted (it is the default). The Markdown output is the same either way. The committed results were re-run at integration with the new default, so none of them carries the per-verdict list.
• The lbnl-ddahu template sets the unit's seasonal damper minimum (#105). g36_afdd now runs with oa_damper_min 45 and oa_damper_min_by_month 28 in June to August (the documented sequence positions), where the learned minimum was the 28 % summer position all year. Hourly, every verdict holds: the fault-free and DMPRStuck_OA_100 runs stay ok and DMPRStuck_OA_0 fault. The winter heating hours at the 45 % minimum return to OS#1, so FC6 judges 1,111 h on the fault-free run (was 744 h; 0.09 %) and 1,273 h on DMPRStuck_OA_100 (was 762 h; 0.08 %), and FC5 and FC7 are evaluated there (367 h and 529 h, 0 %). Winter hours idling at the 45 % minimum are no longer free cooling: the fault-free run's 4 h of FC9 hits go, and on DMPRStuck_OA_0 FC8 reads 100 % of 48 h (was 39.7 % of 121 h) and FC9 0 % (was 20.7 %). FC6 still does not catch DMPRStuck_OA_0 (1.96 % of 766 h, unchanged), which #105 set out to do. Outside June to August that run's damper command saturates (median 100 %) while the hot deck heats, or both coils run, so its hours are OS#5, where G36 does not apply FC6, and the temperature balance there reads a median 17.7 % OA, within G36's 30-point tolerance of the 31.8 % minimum anyway. lbnl-sdahu is unchanged. g36_afdd is not a scored target, so camber datasets score and the gated benchmarks do not move. The open-fdd cross-check's CAMBER verdicts on lbnl-ddahu change (any-FC unchanged in both windows); docs/ECOSYSTEM.md has the before/after.
• Default change: the point-role suggester reads the data when a series is passed (#107). suggest_roles, review_unmapped and camber.interop.bacnet.review_bacnet now use the time-series path (FeatureSuggester(use_timeseries=True)) when a series is passed for a token and no suggester is given. This intentionally changes their output whenever a series is passed, including the BACnet review path when series_by_name is given. The new keyword use_timeseries=False opts out and gives the 0.100 output byte for byte. Without a series the output is unchanged, and so are roles_from_bacnet and vendor_aliases, which pass no series. The FeatureSuggester class default stays name-only. At default settings with the series passed (no unit, no outdoor-air series), top-1 before → after:
• held-out catalog names: 49.4 → 73.0 %;
• real names: 84.8 → 95.7 %;
• simulated LBNL sets: 73.1 → 74.4 %;
• BTS anonymised: 0.0 → 46.2 %;
• BTS Brick-class names (upper bound): 25.5 → 89.4 %;
• the five synthetic styles: 15-26 → 70-90 %.

On strong names the unitless range check costs a little against the name alone. With BTS Brick-class names it is 89.4 % against 93.2 %, because site C's negative airflows are out of range. The full table is in docs/MAPPING-ASSIST.md. - Time-series suggester, step 1: held-out measurement and the two known fixes (#106). Both fixes act only on the time-series path, which from #107 is the default when a series is passed. - Setpoint guard. When a point's name says setpoint (setpoint, SP, STPT, RMCLGSPT), the data ranks only the setpoint roles. robod's temp_setpoint stays cool_sp, where the data alone had made it space_temp. - Range check without a unit. A point with no declared unit is now range-checked on the time-series path. Before, it skipped the check. It is read in every plausible unit of the role (the units the role templates try), and the best reading counts. A °C room temperature passes, and a sentinel such as -999 is demoted, as on the default path. - Measured on the held-out catalog names with their data (catalog_names.py --data, new; 89 of the 223 names have a usable open-tier series). Top-1 / top-3 %: - name only: 76.4 / 79.8; - name plus data: 76.4 / 82.0 before, 75.3 / 80.9 after; - the default suggester with the series passed: 49.4 / 50.6. Its range check reads a unitless temperature as °F.

After the fixes the data helps 3 points and hurts 4. The 4 losses are refrigerated-case air
temperatures labelled as air-handler roles. Two of them, at about -3 °F, are below those
roles' physical bounds.

• Other evaluation pools (name plus data, top-1): real names 95.3 → 95.7 %, with 0 losses (was 2); real names excluding in-sample names 89.1 → 90.7 %. Simulated LBNL sets and BTS are unchanged. The synthetic names move by at most +0.1. Details and the helped and hurt lists are in docs/MAPPING-ASSIST.md.
• real_names.dataset_points takes keep_series and returns each point's 15-minute series with its profile. The evaluation scripts only.

None of these figures is a gated benchmark. The synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmark gates are not affected.

Added

• g36_afdd takes a seasonal OA damper minimum (#105). A new oa_damper_min_by_month parameter ({month: position %}, mirroring min_oa_pct_by_month) overrides oa_damper_min in the months it names, so the operating states judge each interval's OA damper against its own month's minimum position: heating (OS#1) at it, free cooling (OS#2) beyond it. It needs oa_damper_min, the position in the other months. run_g36_afdd gains the same keyword and G36Result a trailing oa_damper_min_by_month field; the finding reports the metric only when it is set, and a caveat names the seasonal minimum. The default is None, so default outputs are byte-identical. Param docs and docs/THRESHOLDS.md cover it.

[0.100.0] — 2026-10-04

0.100: the dual-duct unit's seasonal minimum, equipment causes heading RCx issues, and an opt-in leak drift detector (#97–#104). g36_afdd FC6 takes a seasonal minimum OA (#97). The lbnl-ddahu mapping adds its cold-deck SAT setpoint, so G36 FC9, FC11 and FC13 are evaluated there (#98); the dataset re-ingests once, and the open-fdd cross-check was re-run on the merged code. reheat_penalty takes a box_type that caps its fan-heat estimate by box (#99). A new opt-in coil_leak_drift judges coil-valve leaks against the unit's own baseline (#100). Its results are measured records, not gated. An RCx issue is now headed by a member finding's cause when that cause names the equipment at fault (#101). The lbnl-ddahu and lbnl-sdahu templates now run free_cooling_missed, which re-heads two catalog issues. The point-role suggester reads more public naming conventions and keeps weather-station points outdoors (#102). This intentionally changes the default suggester output. Setpoint-level grouping for the sensor-health outlier read was measured and rejected (#103). Config runs resolve a facility's time zone the way the read API does (#104). The synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmark gates did not move.

Changed

• Dataset templates: lbnl-ddahu and lbnl-sdahu run free_cooling_missed (#101). Both templates now run it at the documented 60 °F high limit, reading each unit's own OA damper command and mixed-air temperature. Hourly, lbnl-ddahu DMPRStuck_OA_0 reads warn (14.5 % of free-cooling hours, cause: damper not delivering), and its fault-free and DMPRStuck_OA_100 runs read ok. lbnl-sdahu also sets the unit's documented 33.8 °F low-limit lockout (low_limit_f, below), so 3,374 of its 5,515 sub-60 °F hours count as free-cooling weather. There damper_stuck_010 and damper_stuck_025 read fault (49.6 %), and onset_damper_stuck_025 and damper_stuck_075 read warn (21.1 % and 20.8 %), all with cause damper not delivering. The fault-free and coi_leakage_010 runs read ok (1.96 % and 1.93 %). Without the low limit they read warn (17.5 % and 18.3 %), because 93 % of their missed hours fall below the lockout, where the sequence holds the damper at its minimum by design. The template comment records both. The lbnl-sdahu RCx report has 11 issues, up from 8. free_cooling_missed is not a scored target, so camber datasets score and the gated benchmarks do not move. The other four air-handler templates do not run it, and each comment says why. nuig-ahu101 is 100 % outdoor air with no mixing box. On lbnl-b59, ornl-frp-ops and ornl-frp-vav the rooftop units trend no cooling-valve command, so the rule would decline.
• An RCx issue is headed by a member's cause when it names the equipment at fault (#101). An issue's heading came from its chain's root finding, which can hide a more specific cause on a member. A cause is equipment-level when it names a component that does not do what it is told: a damper that does not deliver the outside air it is commanded to (free_cooling_missed, damper_not_delivering), a reheat valve whose position does not follow its demand, a leaking valve, a stuck actuator, or a drifting OA damper. The precedence: a root with an equipment-level cause keeps the heading; otherwise the most upstream member with one heads the issue; otherwise the root's cause, as before. The action, its title and its links stay the root's, and the page says in one line which member named the cause. The executive summary's Issue column and the cause key of RcxReport.to_dict() issues follow the heading; each issue gains cause_rule, the rule of the finding that names it. See docs/RCX-REPORT.md, "Which cause heads an issue". On the catalog run templates (15 datasets, default subsets, 98 issues) two headings change, both through free_cooling_missed, which the lbnl-ddahu and lbnl-sdahu templates now run (see Dataset templates below): lbnl-ddahu DMPRStuck_OA_0 reads "Outdoor-air damper not modulating (stuck low)" (was "Outside air below the ventilation minimum"), and lbnl-sdahu damper_stuck_075 reads "Outdoor-air damper not modulating (stuck part open)" (was "Economizer open above the high limit"). No other heading changes. The RCx golden file is unchanged: its synthetic site has no member with an equipment-level cause.
• lbnl-ddahu maps its cold-deck supply-air setpoint (#98). CSA_TEMPSPT (the publisher's Brick model types it as a supply-air temperature setpoint of the cold deck; 55.0 °F in every row) now maps to supply_air_temp_sp, next to the cold-deck CSA_TEMP. Both roles are the cold deck. The hot deck's own 90 °F setpoint is not mapped, and the mapping, template and cross-check caveats say so. The dataset content hash changes, so stores re-ingest lbnl-ddahu once.
• g36_afdd now evaluates FC9, FC11 and FC13 on the unit. DMPRStuck_OA_0 adds FC9 (20.7 %) and FC11 (12.2 %) to the FC8, FC10 and FC12 it already flagged. FC13 reads 0 % on every run. The fault-free run's FC9 hits fall on its 4 free-cooling hours, under the 24-hour floor, so it stays ok. No run's severity changes.
• The open-fdd cross-check was re-run (examples/openfdd_crosscheck/results/). Only lbnl-ddahu verdicts change. Every engine's any-FC result per run is unchanged. The open-fdd engines now evaluate FC7, FC9, FC11 and FC13 there: FC11 fires on DMPRStuck_OA_0 in all four open-fdd runs, and FC13 in the SQL engine and the pandas engine at its own defaults. In month windows, the pandas engine at G36 tolerances goes from TPR 0.58 with FPR 5/12 to TPR 0.62 with FPR 6/12, because FC9 fires on one fault-free month. docs/ECOSYSTEM.md has the before/after.
• Workbook air-sat-reset. supply_air_reset now reads the flat cold-deck setpoint on DDAHU__fault_free: "not reset (setpoint flat at ~55F)", where it read "no reset (SAT pinned low at ~55 F regardless of OAT)" from the supply air before. The severity (warn), the direction (flat) and the 66 % below the G36 target are unchanged. The instructor key, the page, the exercise config comment and the stand-in follow, and a new check pins sp_range_f 0.0.
• Unchanged. No gated benchmark key moves: the synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmarks all hold. The LBNL benchmark's DDAHU family scores outdoor_air_fraction only, which does not read the setpoint.
• Point-role suggester: vocabulary from public naming conventions (#102). This changes the default name-only output. The 0.96 real-name evaluation showed most misses were vocabulary gaps. FeatureSuggester now reads:
• new abbreviations: AF, WH, STA, Enable, SS, PM / PMP, CHWP / HWP / CWP, CT, SAF / SF, BOI, SW / RW / SWT / RWT / LWT / EWT, HWL / CHWL / CDWL, HValve / CValve, HC / CC / LAT, presence, occupant, PIR, weather / WX, and dry-bulb, wet-bulb and dew point;
• run-together words (OADMPR, RMCLGSPT) split into known abbreviations, and vowel-dropped abbreviations (Sply).

Context rules use the same lexical score: - a water point is not given an air-side role (a chilled-water flow is not airflow); - a chiller, boiler or tower places a point on its loop; - an unlocated temperature (air_temperature) reads as a space temperature, not outdoor air; - temp_setpoint reads as a zone setpoint, not supply_air_temp_sp; - a wet-bulb is not a dry-bulb oat; - pump and tower-fan speeds no longer read as supply_fan_speed; - a UUID or PM2.5 names nothing.

On a 6,546-case name × unit corpus, the top-1 role changed in 854 cases. The 0.95 golden cases are unchanged. Module constants WATER_AIR_PENALTY (0.4) and UNLOCATED_TEMP_PENALTY (0.6) hold the two new multipliers. Results on the same inputs (examples/suggester_eval): - real names, pooled, name only: 82.5 → 94.3 % top-1 (in-sample: the vocabulary was chosen from these misses); - held-out catalog names, name only (new catalog_names.py, out of sample): 72.2 → 83.0 %; - BTS anonymised: unchanged. - Weather-station guard on the time-series path (#102). With use_timeseries=True, the data ranks only the outdoor roles when a point's name places it outdoors. For a weather station (weather, WX, meteo, the Synoptic _set_N suffix), it ranks only oat, outdoor_rh and outdoor_co2. A wet-bulb needs the name to say so. The data no longer turns an outdoor temperature into a wet-bulb or an outdoor humidity into a supply-air humidity. On real names, the data's top-1 losses fell from 5 to 2. - Not changed: the time-series path stays opt-in. docs/MAPPING-ASSIST.md recommends turning it on when series are passed. That needs maintainer sign-off.

The synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmark gates are not affected.

Added

• free_cooling_missed takes an economizer low-limit lockout, low_limit_f (provisional). Many sequences lock the economizer out below a low outdoor-air temperature and hold the OA damper at its minimum, so mechanical cooling in that weather is not missed free cooling. With low_limit_f set, hours below it are neither available nor missed. They are reported as low_limit_excluded_hours (and low_limit_cooling_hours, those with mechanical cooling running), with a caveat. camber.freecooling.free_cooling_opportunity takes the same parameter through the same weather test and reports hours_low_limit_excluded. The RCx report's economizer page passes the rule's value to it. The default None keeps both outputs byte-identical, with none of the new metrics or fields. lbnl-sdahu sets 33.8 °F. lbnl-ddahu does not: no lockout is documented for that unit, and on its fault-free run the OA damper modulates above its minimum down to about 15 °F. Its findings are unchanged. The two #101 RCx headings hold. The air-economizer and capstone workbook exercises keep their own configs and do not set it, so their answers do not move.
• reheat_penalty(box_type=None): the "auto" fan-heat cap by box type (#99, provisional). The "auto" estimate of a fan-powered box's closed-valve lift was clipped to 0-8 °F whatever the box. On the LBNL series boxes (lbnl-fpu SFPU runs) the real lift is 13.7 °F, so the clip always bound. box_type declares the box: "single_duct" (no fan) caps the estimate at 3 °F, and "parallel" caps it at 8 °F, as before. "series" caps the no-rise allowance (valve open) at 10 °F and keeps the big-rise allowance (valve shut) at 8 °F. The big-rise check judges the closed-valve samples that the estimate is learned from, so with a higher cap a passing valve's heat is learned as fan heat and the check cancels itself out. With box_type set, the finding also reports box_type, fan_lift_f (the uncapped estimate) and fan_heat_closed_f. The default (None) and a numeric fan_heat_f are unchanged, and default outputs are byte-identical. Measured on all 31 PFPU and 31 SFPU runs (hourly, read through the measured position and through the demand alone): "parallel" matches the default on every PFPU run. "series" leaves every SFPU verdict unchanged and widens the declines' margin. The valve stuck at 20 % (12.9 °F rise) was declined 0.06 °F under the old 13 °F bound and is now 2.1 °F under the 15 °F bound, while the lowest working full-valve rise (17.6 °F) stays 2.6 °F above it. A symmetric series cap of 9-11 °F would have dropped the passing-valve caveat on the 50 % and 80 % leaks and on the valves stuck at 80 % and 100 % (read through the demand). The caps were set on these labelled runs, so they are an in-sample fit. The lbnl-fpu exercise configs are unchanged, because their boxes are parallel. docs/THRESHOLDS.md documents box_type.
• g36_afdd FC6 takes a seasonal minimum OA (#97). A new min_oa_pct_by_month parameter ({month: pct}, the override outdoor_air_fraction already takes) makes FC6 judge each sample against its own month's minimum. It needs min_oa_pct for the other months, and the FC6 caveat names the seasonal minimum. The default is None, so default outputs are unchanged. Param docs and docs/THRESHOLDS.md cover it.
• The lbnl-ddahu run template runs g36_afdd (#97). FC6 is judged against the unit's documented seasonal minimum: 31.8 %, and 11.9 % in June to August (the values outdoor_air_fraction uses), with a basis map. Hourly, FC6 reads 0.13 % on the fault-free run (ok), 1.96 % on DMPRStuck_OA_0 and 0.13 % on DMPRStuck_OA_100. No verdict depends on FC6. FC6 is not a stuck-closed detector on this unit. Since #95 its applicable hours are almost all in June to August, the learned OA damper minimum being the 28 % summer position, and against an 11.9 % minimum a damper stuck shut is off by less than G36's 30-point tolerance. A fixed 31.8 % would flag DMPRStuck_OA_0 (32.1 %), but only by judging summer hours against the winter minimum. The #94 figures for a single minimum (fault-free FC6 20.5 % at 31.8, 47.1 % at 11.9) predate #95. On the current code both single minima read the fault-free run ok.
• A coil-valve leak drift detector against the unit's own baseline (#100; opt-in, provisional). This is the follow-up planned in #84. leaking_valve catches the published 10 % leak on lbnl-sdahu only with a fan heat calibrated on the fault-free run, and that run is also scored. The new coil_leak_drift (camber.rules.coil_leak_rule.CoilLeakDrift) calibrates nothing. It fits the coil's valve-shut air rise (the coil's leaving air, or the supply air, minus the mixed air, on fan-on hours with every mapped coil valve shut) against the mixed air on a known-good window. It then flags a shift in the current window at the same mixed-air temperature: down for a cooling leak, up for a heating leak.
• How it runs. It uses the existing drift machinery, so the baseline can be a frozen one or a declared reference from 0.98 (another unit, or a known-good period of the same unit).
• What it judges. Only current hours inside the baseline's mixed-air range are judged. The rest are counted and caveated, never extrapolated.
• What it reports. A finding reports the shift in °F and in baseline sigmas, the basis sensor, the gates applied and a sustained-shift alarm. A warn or fault also says that a shifted supply-air or mixed-air sensor, or a valve stuck partly open, reads the same.
• How to switch it on. An ahu drift family entry opts in with "coil_leak": ["cooling"] (or ["cooling", "heating"]) and may tune it with "coil_leak_params": {...}. build_drift_suite and refit_baselines take the matching coil_leak= and coil_leak_params= keywords.
• Roll-up. diagnose_ahu_drift puts a leak on the coil side as coil_leak_drift:<coil>. The scorecard counts it under maintenance.
• Defaults are unchanged. Without the key no suite, config, template or benchmark changes.
• Documentation. Its parameters are documented in the new camber.rules.param_docs.DRIFT_PARAM_DOCS and rendered in docs/THRESHOLDS.md under "Opt-in drift detectors". Also updated: CLI.md ("Leak drift (opt-in)"), TUNING.md, AHU-DRIFT.md and API-STABILITY.md.
• Validation on the LBNL leak runs (measured, not gated). examples/lbnl_fdd/leak_drift.py scores the detector against each unit's fault-free run three ways: a 60/40 split, a declared twin, and a mid-year onset splice. It reports TPR/FPR with 95 % Wilson intervals.
• lbnl-sdahu. The 10 % leak reads −1.05 to −1.11 °F (−3.5 to −3.7σ). TPR is 1/1 (0.21-1.00) in every mode and FPR 0/2-0/3, with no calibration.
• lbnl-fcu. Cooling and heating leaks at 20 / 50 / 80 %: TPR 6/6 (5/5 onset), each on the right coil. FPR is 1/33 split (a −4 °C room-sensor bias, +2.5σ) and 0/32-0/33 for twin and onset. The point-in-time leaking_valve at its defaults reads 24/33 there.
• Not scored. Stuck-open valves (the same symptom) fire on 4/4 SDAHU and 9/10 FCU runs. The SDAHU supply-air sensor biases −2 / −4 °C fire as cooling leaks; this sensor confound is stated in every finding.
• No leak runs elsewhere. The dual-duct archive has none.
• Not gated (maintainer decision). The detector's results stay measured, opt-in records. No drift.coil_leak_drift key is added to any benchmark baseline, and the LBNL benchmark does not run the detector.

Documentation

• Sensor health: the setpoint level as a grouping key is measured and rejected (#103). The alternative to the residual read deferred from #87 judges a controlled point's outliers within each level of its setpoint, when the setpoint takes at most 4 distinct values. It was measured for supply air and duct static on every catalog RCx report (511 trust-table rows). As proposed, it moves no row: no catalog setpoint has two to four levels. The LBNL single-duct runs hold one constant level, and lbnl-b59 and nuig-ahu101 reset theirs. Grouping by the four most common levels does reach those two units, and it lowers every supply air it touches by 0.01-0.12. lbnl-b59 RTU03 and RTU04 go from trusted to suspect. Not adopted; there is no code change. docs/SENSOR-HEALTH.md records the measurement, and a new test pins that mode="auto" does not group by setpoint.

Fixed

• Config runs resolve a store facility's time zone the same way as the read API (#104). Since 0.99.1 /facilities reads a facility's zone from its registry entry's timezone first. Config runs on a store source read only the catalog entry's zone or the open-fdd provenance, so an explicit timezone on the registry entry was ignored by runs. Both now use one lookup: the entry's timezone, the dataset block on the entry, the catalog entry, then the open-fdd provenance. Behaviour change: a per-site zone in a facility's dataset block now also sets a run's zone, as /facilities already reported it. Each BDG2 site records one. On the ten BDG2 sites' template runs (65 daily meter baselines) the fall-back and spring-forward days now have their 25 and 23 hours; no verdict changes, CV(RMSE) moves by at most 0.0003 and R² by at most 0.0012.

[0.99.1] — 2026-10-03

0.99.1 patch: open-fdd-ingested facilities report their time zone (#96). The synthetic, fleet, LBNL and BDG2 benchmark gates did not move.

Fixed

• The read API did not report the time zone of an open-fdd-ingested facility (#96). camber interop openfdd ingest recorded the site zone only in the facility's openfdd provenance. Config runs found it there, but /facilities looked a zone up only on the registry entry or through a catalog dataset_id, so these facilities had no timezone and the trend viewer labelled their site wall clock as UTC. The importer now also writes the zone to the facility's meta["timezone"]. The read API looks up a facility's zone in this order: the explicit timezone, the catalog dataset zone, then the openfdd provenance zone, so facilities ingested with 0.99.0 report their zone without a re-ingest. An unchanged re-ingest, which is skipped, also fills in the missing key. Facilities with no recorded zone are unchanged.

[0.99.0] — 2026-10-03

0.99: open-fdd interop and G36 heating at minimum OA (#22 items 1–4, #95). CAMBER and open-fdd now meet at a file and process boundary, with every result labelled by the engine and version that produced it. A reusable cross-check (#22 item 1) scores CAMBER's g36_afdd and open-fdd's pandas and SQL engines (commit 32a6d44, PyPI 4.4.9) on the same labelled LBNL air-handler data, and explains each difference from public code and open data. An importer (#22 item 2) reads an open-fdd building package or historian layout into a CAMBER store, so drift, M&V and sensor-trust checks can run on data collected at the open-fdd edge. Columns map through a versioned role crosswalk (#22 item 4), and the results go back as a draft findings-exchange JSON (#22 item 3). docs/ECOSYSTEM.md opens with an entry point for open-fdd users. In g36_afdd, heating (operating state 1) now needs the OA damper at its minimum, the companion of #94's free-cooling test (#95).

Added

• G36 cross-check against current open-fdd (#22 item 1). examples/openfdd_crosscheck runs CAMBER's g36_afdd and open-fdd's two engines on the same labelled frames, pinned to open-fdd commit 32a6d44 (PyPI 4.4.9). It works through files and processes only: the pandas engine runs in its own venv as a subprocess, and the SQL engine runs as fdd_cli in a container with no network and read-only mounts.

The harness ships: - a versioned role mapping (role_map.json); - tolerance profiles (profiles.json): open-fdd's defaults, and the G36 Table 5.16.14.7 tolerances CAMBER uses; - one normaliser per engine, recording each verdict's denominator; - per-engine, per-FC scoring with Wilson intervals, where "not evaluated" is kept apart from "not detected"; - synthetic probes that isolate single engine behaviours.

Results for all three engines on lbnl-sdahu (full subset) and lbnl-ddahu replace the "not yet re-compared" section of docs/ECOSYSTEM.md, which now also says where the 0.1.5 comparison no longer applies. CAMBER runs with each dataset's run-template parameters, so FC6 is evaluated on lbnl-sdahu (#94). Synthetic probes run through fdd_cli confirm which G36 tolerances the SQL tuning file cannot set (FC7, FC9, FC11, FC13–FC15). Offline tests run without Docker or open-fdd: tests/test_openfdd_crosscheck.py. - CAMBER reads open-fdd data (#22 item 2, provisional). The new camber.interop.openfdd module reads an open-fdd building package (openfdd_package_v1, a folder or a .zip) or open-fdd's historian Parquet layout. CAMBER's drift, M&V and sensor-trust checks can then run on data collected at the open-fdd edge. - Boundary. It works through files and processes only. No open-fdd code is imported or copied, and nothing is written back. - Required inputs. read_package / read_historian take the site time zone and the unit system as required arguments, because a package carries neither and CAMBER will not guess. - Mapping. Columns map through the package's own map, which is authoritative. Every other column is counted and reported with its reason: no name, unknown name, deliberately unmapped, wrong equipment type, unusable unit or duplicate role. - Units and time. It handles UTC and local stamps, SI and IP units, declared units, and 0–1 vs 0–100 percent. It reads equipment classes from the equipType stamps and the inventory, never from ids. - ingest_package and camber interop openfdd ingest (provisional). They write one facility per building into a ParquetStore or a portfolio workspace. - Lifecycle. In a workspace the facility is registered through the lifecycle (provisioning, or active with --activate), the write takes the workspace lock and is audited (interop.openfdd.ingest), and a facility in any other state is refused. - Provenance. Recorded under openfdd on the registry entry: the source and schema version, the sha256 of every file read, the crosswalk version and its pinned docs commit, the zone, units and coverage. - Re-ingest. It is idempotent, and data is staged and swapped in. --config-out writes a starting run config: rules the mapped roles can run, the package weather as the shared outdoor temperature, and a daily M&V baseline per metered class. camber interop openfdd inspect does the same read without writing. - A versioned open-fdd → CAMBER role crosswalk (#22 item 4, provisional). It ships as camber/interop/openfdd/crosswalk.json, version 1, written from open-fdd's documented Haystack-name and SQL-role vocabulary at a pinned commit. 50 of its 71 rows map; the other 21 are deliberate non-mappings, each with its reason. camber interop openfdd crosswalk [--json] prints it. - A findings-exchange JSON, draft 0.1 (#22 item 3, provisional). findings_document, run_findings and camber interop openfdd findings CONFIG --out FILE emit an engine-labelled document that a separate process (an open-fdd agent or pipeline) reads back. - Every record carries {engine: {name, version}}. - declined and not_evaluated are distinct statuses. - magnitude names its denominator. - Fields CAMBER does not yet produce are null. - See docs/INTEROP-OPENFDD.md.

Changed

• A store facility ingested from open-fdd knows its zone. A store-source config's source.timezone now defaults to the zone the facility was ingested in, as it already did for catalog datasets. This is additive. A facility without an openfdd provenance block is unchanged.
• G36 heating needs the OA damper at its minimum (#95). g36_afdd now reads a fan-on hour with the heating coil alone active as operating state 1 only when the OA damper is at its minimum position, per the G36 §5.16.14 operating-state definitions. It uses the same learned or configured oa_damper_min and oa_damper_tol as the free-cooling test from #94. A heating hour with the damper open beyond the minimum is state 5, where only FC1–FC4 apply. A missing damper reading there leaves the hour unclassified. Without an OA damper point the valves-only reading stays, and the caveat now names both states.
• New outputs (provisional). The G36Result.n_heating_above_min_oa field and the finding metric heating_above_min_oa_hours.
• Before/after. No change on lbnl-sdahu (no heating coil), nuig-ahu101 (no OA damper point) or the synthetic G36 scenarios, and no gated benchmark moved. On the dual-duct lbnl-ddahu the hot deck heats while the OA damper command is well above the learned 28 % minimum (median 55 % on the fault-free run). Between 885 and 1,520 heating hours per run move from state 1 to state 5, so FC5 is no longer applicable on the two stuck-damper runs. No run's flagged FCs or severity changed.

[0.98.0] — 2026-10-03

0.98: hardening from the workbook (#84-#93). Working through the 0.97 workbook on real data showed where CAMBER's answers were wrong, silent or hard to tune. On the air side (#84) the damper census and supply_air_control gate on trended occupancy, and leaking_valve can credit a unit's measured fan heat. For terminal units and ventilation (#85) a new rule, actuator_stuck, finds a box's damper or valve stuck against the zone's demand, reheat_penalty reads a valve's measured position, a heating shortfall is graded on how often it happens as well as how deep it goes, and cohort deviation gains opt-in grouping and normalisation. The central plant (#86) checks the sign of the chilled-water reset, learns a pump's VFD floor, runs the boiler rules without a run status, and accepts a declared drift reference. Sensor health (#87) reads a fan-less unit's outliers per operating mode and names clipped readings, stray rows and scheduled status points. The RCx report (#88) names each issue's cause, lists the checks that did not run, says what data an M&V refusal needs, and ends the issues with a generated "Verify on site" checklist. More rules link to the PNNL re-tuning material and ornl-frp-vav declares a scored detector (#89). Every tunable threshold is documented with its basis and a way to calibrate it, and run configs read from YAML as well as JSON (#90). Three follow-ups close small inconsistencies: one 60 °F free-cooling high limit (#91), a site elevation for derived wet-bulbs (site_elevation_ft, #92), and named fan-off hours with an opt-in duration fault in dcv_verification (#93).

Four maintainer decisions shape the benchmarks. S1: the LBNL benchmark reads the leaking_valve parameters from the lbnl-sdahu template, which moves the SDAHU and pooled leaking-valve keys (the fan heat was calibrated on a run that is also a scored negative; docs/VALIDATION.md states the circularity). S2: faultlab's healthy chilled-water reset now runs the right way, with no synthetic key moving. S3: actuator_stuck is a scored synthetic scenario (46 scored rules, new actuator_stuck TPR/FPR keys). S4: drift accepts a declared, never-stored reference (another unit or a known-good period). For sensor health, decision (a) applies: only units with no fan signal are read per inferred operating mode, and fan-gated units are unchanged. Every other benchmark is unchanged.

Added

• roles_any_of (#86 item 4a, provisional). A rule may declare roles_any_of, a tuple of role groups of which each needs at least one role present (a run status or a gas input). The runners load every role of a group, skip the equipment when a group has none, and include the present group roles in the sensor-health gate. camber.model.entities.missing_inputs is the one test shared by the runners and runnable_rules, and Runnable.missing_any_of names the unmet groups. The mechanism alone leaves default outputs unchanged; the first rules to declare groups are the boiler firing rules (see Changed, 098-plant-boiler).
• Rules that did not run are listed (#88 item 3, #86 item 4b). Registry.run, run_periods and run_fleet take an optional skipped= list and append a RuleSkip (rule, equipment, class, missing inputs, reason missing_inputs / no_data / no_verdict) for each equipment a rule applies to but produced nothing on. A skip is recorded only when the rule applies: its declared classes match the equipment's, or at least one of its inputs is present (a chiller rule on an air handler is not applicable, not skipped). The default (None) records nothing, so the benchmarks and faultlab are untouched. run_config collects them in RunResult.rules_skipped, adding a rule-level record when a configured rule found none of its inputs on any equipment; they stay out of findings (and findings.json, so datasets score decline counts do not move).
• RCx report. Appendix A gains a "Checks not evaluated (missing inputs)" table (rule, missing inputs, equipment: one row per rule and missing set, six units named, then "and N more"), and the cover a "Checks not evaluated" row, both only when there is something to list. no_verdict records are kept for later use but not listed.
• An M&V refusal says what data is needed (#88 item 4). New camber.mandv.sufficiency: baseline_need(interval, n_have, min_n=...) returns the required, available and missing amount, its unit, the calendar days it takes to collect, and a one-line text ("1,440 baseline hours needed, 168 available: 1,272 more hours (about 53 days of data)"). caltrack_savings and caltrack_savings_hourly raise InsufficientBaseline, a ValueError whose message is byte-identical to before, with the gap on .need: too few days or hours, missing hour-of-week bins, or thinly observed bins. The config's daily M&V decline for too few days adds metrics["data_needed"] to its mv_baseline and mv_savings findings (declined_reason and summary unchanged), and the RCx "M&V and drift" section prints a "Data needed" paragraph per meter. Hourly M&V has no config path, so its need is reported through the exception only. Workbook capstone pins the one-week refusal's need at 1,272 hours.
• Tunable thresholds, documented (#90, provisional). camber/rules/param_docs.py documents every numeric, flag and enumerated constructor parameter of every built-in rule, including the extra instances. Each entry gives the unit, a sensible range, the basis of the default (standard: ... only where the code cites a section, public source: ..., CAMBER judgment or calibrated on ...) and how to calibrate it from your own data. Defaults are read from the constructors, never copied. FIXED lists thresholds that are still hard-coded, and a test fails on any undocumented or stale parameter.
• camber rules params [RULE] [--json|--yaml]. Prints each parameter with its default, basis and calibration hint, plus a ready-to-paste config snippet. The YAML snippet carries each note as a comment beside its value.
• Every built-in rule is tunable from a config. make_rule and a config's params now also cover the extra instances (cohort_airflow, cohort_space_temp, sat_/static_reset_effectiveness, sat_/static_rogue_zone_census, sat_/static_cohort_starvation); their identity arguments (the cohort role, the reset kind) stay fixed. rule_factories() lists them.
• YAML run configs (the [yaml] extra, pyyaml>=6). camber run, report, explain, ask, fleet, drift, mv, python -m camber.config, the portfolio migration and the edge config read .yaml / .yml. Without PyYAML the command stops with an error that names the extra. The loader types values as JSON does (dates, 07:00, no / on and 012 stay strings), so equivalent JSON and YAML configs give identical results, as a test asserts. JSON stays the dependency-free default.
• camber datasets config --format yaml (and --out *.yaml) writes a template as YAML, with its _comment notes turned into comments; config_template(..., format=). Writing YAML needs no extra.
• Calibration provenance in findings. A rule entry may carry a basis map ({"fan_heat_f": "calibrated on ..."}), which each finding of the rule records as metrics["param_basis"] = {param: {"value", "basis"}}. A misspelt parameter name is an error. Existing configs are unaffected.
• Docs. docs/THRESHOLDS.md is generated from the registry by scripts/thresholds_doc.py (--check is a test). The new guide docs/TUNING.md, "Tuning thresholds with your own data", covers calibrating on a known-good period, avoiding circular calibration, recording provenance in config comments and re-checking against labelled data. Its worked examples use lbnl-sdahu and nist-heatpump-fdd. Also updated: CLI.md, API-STABILITY (provisional), the mkdocs nav, the README, and a "going further" pointer in the workbook index.
• A declared, never-stored drift reference (#86 items 1 and 5, decision S4; provisional). A drift.families[] entry may carry "reference": {"equip": ID} (score every other unit of the class against a named healthy one, optionally with its own "period") or "reference": {"period": [start, end]} (a known-good window of the same unit). The reference is fitted in a scratch in-memory store on every run and never saved, so a run still cannot mint its own baseline. Findings carry metrics["baseline_source"] (reference:<equip> or period:<start>..<end>) and a caveat, and their summaries read "vs the reference ..." instead of "vs frozen baseline". The reference unit declines as is_reference; a reference that cannot serve a detector declines every target (reference_missing_inputs, reference_untrusted, empty_reference). A section whose families all declare a reference needs no store and no windows. DriftFamilyResult.reference is new (also in as_dict); the drift report and camber drift run name the reference. Nothing changes for configs without one.
• The LBNL plant templates score the plant drift detectors. lbnl-chiller.json and lbnl-boiler.json (and the plant-cooling-tower / plant-boiler exercise configs) declare PLANT__fault_free as the reference for the tower and boiler families, and the catalog declares cooling_tower_fan_effort_drift: [tower_fouling] and boiler_efficiency_drift: [boiler_fouling] as scored targets. Through camber run the verdicts match examples/lbnl_fdd/plant_detectors.py run by run: tower fan effort 2/3 (fouling 065 and 080 at +17.8 / +10.9 fan %-points; 095 at +3.3 is under the floor), boiler efficiency 3/3 (+53.7 / +24.9 / +5.1 %; 095 clears the 5 % warn floor by 0.1 point), no false alarm on the 20 / 13 other runs. The scorer counts the reference run as a correct negative (0/21, 0/14).
• heat_valve_position: a reheat valve's measured position beside its demand (#85 item 3, provisional). A new role for a unit that trends both the controller's demand and the valve's feedback: map the demand to heat_valve and the feedback to heat_valve_position. It is a percent role (0-1 fractions are rescaled), physically bounded like the other valves, carries the Haystack tags heating valve sensor and the 223P quantity of heat_valve, and is offered by the mapping assistant for percent units. It is not exported as a Brick point (the importer reads a heating coil's position sensor as heat_valve). lbnl-fpu maps RH_VLV_S to it; its catalog known issue and mapping comment say why.
• reheat_penalty(fan_heat_f=None) (#85 item 3). A fan-powered box's own fan (and, in a parallel box, the plenum air it mixes in) lifts the discharge above the entering air with the valve shut. A number of °F raises both valve-vs-discharge bounds (5 °F no rise at full valve, 10 °F big rise with it shut) by that much; "auto" estimates it per box as the median closed-valve, airflow-bearing lift over the entering air (fan-on samples only when the fan status is mapped, >= 12 samples), clipped to 0-8 °F. The value used is reported as fan_heat_f. The default None changes nothing; the two lbnl-fpu exercise configs set "auto", with its basis.
• Why free cooling was missed (#88 item 2). free_cooling_missed records the cause as additive metrics; its severity and missed_pct do not change. missed_cause is damper_not_delivering when, on missed hours with a usable temperature balance (|OAT - RAT| >= 5 °F), the OA damper was commanded at least 90 % open while the measured OA fraction stayed below 80 %, on at least 20 % of those hours and 24 h. It is economizer_not_commanded when a damper command is trended but stayed below open, and undetermined otherwise. The rule also reports commanded_open_pct, commanded_open_hours, commanded_open_oaf_median_pct and missed_damper_cmd_median_pct. The new parameters (cmd_open_pct, oaf_open_pct, stuck_min_share_pct, stuck_min_hours, stuck_low_oaf_pct) are documented in docs/THRESHOLDS.md. Measured on real data: the four lbnl-sdahu stuck-damper runs read damper_not_delivering (40-87 % of those hours) and the fault-free and valve-leak runs 0 %; on lbnl-ddahu the damper stuck closed reads 95 % and the fault-free run 0 %.

• Recommendation.cause (#88 item 2, provisional). Every recommender names the finding's cause in a short phrase built from the metrics it reads, e.g. "Outdoor-air damper not modulating (stuck low)" or "Hot-water pump pinned at its minimum speed". title stays the action. free_cooling_missed with damper_not_delivering gets a new recommendation, "Repair the outdoor-air damper or actuator", linked to the PNNL economizer guide and Re-tuning ch. 6. "Stuck low" or "stuck part open" follows commanded_open_oaf_median_pct against stuck_low_oaf_pct (default mirrored in DEFAULT_PARAMS["econ_stuck_low_oaf_pct"]).

• Clipped readings are named (#87 item 2, provisional). sensorhealth.clipped_at_limit(series, role) finds a pile-up at a round-number range limit (at least 12 samples and 0.5 % within 0.1 % of the span of the extreme, ten times the density of the adjacent 5 % band, and the limit a round number to within 0.01 %, in degF or degC for temperatures). frame_checks flags the point clipped and fills SensorTrust.clipped (side, limit, limit_label, n, frac, frac_fan_off); no trust penalty. Checked on CO2, OAT, wet bulb, space and return air (both ends) and airflows, water flows and duct static (high end). On the catalog: nuig-ahu101's CO2 at 2,000 ppm (all fan-off), lbnl-b59's OA flow at 20,000 cfm, and the simulated 140 °F bound of one LBNL dual-duct fault run.

• Outliers read per operating mode (#87 item 1a, provisional). sensor_trust(mode=, mode_source=) and frame_sensor_health(mode=None | "auto" | Series) judge a fan-dependent duct point's outliers within each operating mode; "auto" infers an off-mode (OA damper <= 2 %, every coil valve <= 1 %) on a unit with no fan signal only. New SensorTrust.mode_source and mode_outlier_frac.
• Stray lead / tail rows (#87 item 1b). A point's rows beyond a gap of 30 days or more (and 20 % of its span) that hold at most 1 % of its samples are left out of the judgement: flag stray_lead / stray_tail, with SensorTrust.main_start, main_end and n_stray.
• Scheduled status points (#87 item 3). gapfill_signature(series, role=None): for a stepwise point (a status role, or >= 95 % of samples on two levels) days sharing a pattern with two or more others are reported as "N days follow a fixed schedule" (scheduled_days, n_schedule_patterns) rather than warned on; repeated_days keeps only unexplained repeats.

• "Verify on site": a generated walk-down checklist in the RCx report (#88 item 1). New provisional module camber.walkdown: site_checks(issues, *, recommend, rule_of, overrides, trust, skipped, declined) returns SiteCheck items (issue_key, equip, kind, look_at, point, confirms, refutes, references, plus rule and rank), ordered sensors, equipment, design values, data. Sensors: each sensor a conditional issue leans on (with its gated trust), a sensor_drift issue's sensor, and each input a check declined as untrusted. Equipment: one item per issue from SITE_CHECKS[rule][cause], a template for every rule with a recommender, its cause read from the finding's metrics (CAUSE_KEYS: missed_cause, the CHW reset direction and flow mode, a pump's inferred VFD floor, the reheat valve divergence, the DCV causes, ...); other rules get a generic item built from their required inputs. Design values: DESIGN_PARAMS (site facts such as a minimum outdoor-air fraction, a high limit or an occupancy schedule, never detection thresholds) still at the rule's default. Data: the checks not evaluated for missing inputs. The texts describe what a technician checks on site, in CAMBER's own words; every item links PNNL Re-tuning chapter 9 through the references registry (camber.references.WALKDOWN_REFERENCES, not a rule mapping), and nothing from the chapter is reproduced.

• actuator_stuck: a terminal or fan-coil damper or valve stuck against the zone's demand (#85 item 1, provisional). A new rule for terminal boxes and fan coils (an air handler's outdoor-air damper is out of scope). It finds the runs where an actuator holds one position over occupied (trended occupancy, else the schedule), fan-on hours, through a thin wrapper around the sensor-health run finder (camber/rules/_flat_runs.py), and judges each run of at least min_flat_hours by what the zone asked for. Contradicted runs can reach fault: a damper shut through occupied hours with the airflow at or below 5 % of AIRFLOW_SP or min_airflow (with neither, judged against the occupied mode alone, with a caveat); a damper or cooling valve below its open limit while the zone runs warm_margin_f over its cooling setpoint for a quarter of the run (a heating valve: under its heating setpoint); fully open while the zone sits satisfied_margin_f inside its setpoint for half the run; a heating-valve position flat while its demand moves 20 points. Unexplained runs (one value for whole_day_share of a day's active samples while the demand, airflow setpoint, a setpoint or the zone temperature moves) warn at most, and a run at a limit the demand agrees with is saturated, not stuck. Roles: damper, heat_valve_position (else heat_valve), cool_valve. Metrics per role: flat_runs, stuck_share, value, tier, reason, driver and the flagged runs. Registered in RULE_CLASSES, the applicability table (terminal, fan_coil), the scorecard (maintenance), the references (PNNL chapter 7), the parameter docs, and the advisory recommender (cause "Damper stuck at 20 %: the zone runs warm while it holds still"; stroke the actuator before retuning anything). On ornl-frp-vav it finds 6 of 6 stuck days on the default subset and 17 of 18 on the full one, with no false alarm on the 13 fault-free and airflow-bias days (see Changed).

• actuator_stuck is a scored synthetic scenario (S3, approved). faultlab gains a VAV box whose damper sticks at 30 % through warm afternoons; coverage.n_scored and n_single go 45 -> 46 and the synthetic benchmark gains actuator_stuck.tpr 1.0 and .fpr 0.0. Every other synthetic key is unchanged.

• Cohort-deviation options (#85 item 2, opt-in). CohortDeviation (cohort_airflow, cohort_space_temp) takes group_by_topology (compare only the units behind one air handler, grouped like the rogue-zone census), normalise ("design_max": by design_max={equip: cfm}, else the peak AIRFLOW_SP; "reference": by reference={equip: reference_equip}), summary="variability" (the standard deviation) and tail ("low" / "high"). Every default reproduces the earlier result. camber.charts.cohort gains the variability summary, a tail argument and cohort_deviation_from_values. Size normalisation alone cannot isolate a stuck box: on the ORNL set the share of design airflow flagged six healthy box-days and one stuck day of six; each box against its own fault-free day flagged all six.

• Shared M&V test vectors (examples/mv_vectors/). These are inputs, CAMBER's expected outputs and a standalone checker, so that another change-point / Guideline 14 implementation (open-fdd's helpers first) can be cross-checked through files alone.

• Synthetic cases. 8 seeded cases with exact truth: 2P; 3PH at 58 °F; 3PC at 65 °F; 4P; 5P at 55 / 68 °F; a noise pair either side of the baseline gate; and a weak-weather load.
• BDG2 meters. 6 meters (electricity, chilled water, steam and gas) as daily and calendar-month aggregates, plus one irregular-bill variant. Each has a baseline year and a reporting year, both raw and with a 10 % injected saving.
• BDG2 data stays local. CAMBER redistributes no datasets, so the BDG2 inputs and predicted series are not committed; expected.json keeps only their statistics. fetch_bdg2.py (numpy, pandas and the standard library) rebuilds them:
  • it downloads the publisher's files at the catalog's URLs and checks their sha256 pins;
  • it rebuilds the aggregates deterministically and checks each derived CSV against its own pinned sha256;
  • it writes them to a git-ignored local/ folder, and prints the citation and the CC BY-SA 4.0 licence.
• Bill cases. 3 cases (1 synthetic, 2 BDG2) with mid-month 28–35-day reads, one estimated read and one missing bill. They run through CAMBER's billing config path: base_f: "auto" bases with their ranges, degree days built from each day vs from the bill's mean, the degree-day model against the change-point models by BIC, Portfolio Manager calendarization and avoided cost at each bill's own (synthetic) rate. The synthetic bill case is committed; the 2 BDG2 bill cases are rebuilt locally.

• Expected outputs. expected.json uses the versioned schema mv_vectors/1 and records, per fit:

  • the selected kind, every candidate's BIC and the BIC gap;
  • the coefficients, in CAMBER's form and in a convention-free slopes_dEdT form;
  • n, p, R², adjusted R², CV(RMSE) and NMBE;
  • the baseline-gate, calibrated-simulation-gate and SEP verdicts side by side;
  • Option C savings with FSU.

  predictions/ holds CAMBER's predicted series row by row. - Consumer tools. These never import CAMBER, so a pandas library and a SQL twin are checked alike: - check_vectors.py (numpy and pandas only) rebuilds the expected numbers (--self-test), compares another implementation's results JSON with the documented tolerances and prints CAMBER's own results (--template); - export_parquet.py writes typed Parquet copies of every CSV; - example_results.json is a worked results file. - Documentation and licence. SCHEMA.md documents every field, the two tiers, the method (grid, BIC, p counting, day weighting, the bill choices) and the tolerances. Everything committed is Apache-2.0. The locally rebuilt BDG2 files are CC BY-SA 4.0, with attribution. - Regression test. tests/test_mv_vectors.py regenerates the synthetic tier offline and requires an exact match, so any move in CAMBER's M&V numbers shows up there. - With examples/_data/bdg2 present, it also rebuilds the BDG2 tier, checks every sha256 pin and requires the BDG2 statistics to regenerate exactly. - -m network does the same from a fresh download.

  No default output changes.

Changed

• Equipment class on role frames; the DCV return-air caveat only where it applies (#85 item 4). The rule runners (and a run's lazy frame_for) set frame.attrs["camber_equip_class"] to the equipment's class. dcv_verification keeps its "typically return-air CO₂" caveat on air handlers and on equipment of absent or unrecognised class (direct API calls are unchanged), and drops it on terminals and fan coils, whose CO₂ is the room's own: the finnish-dcv, b4b-windesheim and sdu-ou44 rooms (class VAV) lose it. Only that caveat changes; severities and metrics do not. Workbook zone-dcv pins the caveat absent and its instructor discussion point is rewritten.
• damper_census reads trended occupancy (#84). Each box's occupied hours now come from its own trended occupancy point when it has any non-null value, and from the assumed weekday 07:00-18:00 schedule only when it has none; trended warm-up / cool-down flags drop prep-mode samples (the rule declared them but never read them). A weekend test day now gets a census instead of "no damper data". New rule param occupancy_gate ("trended" default, "schedule" for the old behaviour, "off" for every sample), new occupancy_gate metric and DamperCensusResult.occupancy_gate field (trended occupancy / the assumed schedule / mixed / off), and damper_census(..., use_trended_occupancy=True). Intended default change: on ornl-frp-vav the fault-free day's census median moves 39.3 -> 38.7 % (still a fault, all 10 boxes low), and the weekend days d3_stuck_000, d3_stuck_060 and d3_stuck_100, which returned "no damper data", now read 36.8 % / 40.3 % / 41.8 %.
• supply_air_control gates on trended occupancy (#84). New param occupancy_gate: "trended" (default) judges only fan-on samples that the unit's trended occupancy marks occupied, and keeps the fan-only gate when no occupancy is trended (no assumed-schedule fallback); "schedule" falls back to the weekday 07-18 schedule; "off" is the pre-0.98 fan-only gate. The gate sits in the running mask, so the finding, its evidence chart and the triage violation mask judge the same samples. New occupancy_gate metric; the summary says "occupied running hours" when gated. Intended default change: the fault-free lbnl-sdahu unit goes from a warn at 12.1 % too warm (78 % of those hours were unoccupied fan cycling) to ok at 2.97 %; AHU__damper_stuck_075 stays a fault, too cold 32.6 -> 31.2 %; the capstone's RCx report loses the fault-free control's supply_air_control issue (6 issues, was 7).
• irish-ahu template comment corrected (#84). The excess outdoor air is not "mostly the COVID-19 period": 53 % of the cooling-weather hours before it (2017-06 to 2020-02) against 45 % during it (2020-08 to 2021-11).
• Workbook. air-sat-reset (fault-free unit ok at 3 %, the stuck damper 31 %, question 5 recast around the gate, with a pinned check that occupancy_gate: "off" restores the 12 % warn), air-static-pressure (census median 38.7 %, the weekend caveat and common mistake rewritten) and the air-economizer instructor key (the stale sentence about the catalog note removed) updated; capstone re-verified.
• overcooling_severity grades a heating shortfall on share as well as depth (#85). The shortfall grade was the deepest tier sustained for an hour, so a few cold hours a year with the reheat saturated read fault. It is now the lesser of the depth tier and a share tier set by the share of occupied samples in a sustained shortfall at least warn deep (new shortfall_share_pct, default {"warn": 5, "fault": 20}; below 5 % a shortfall is info). Intended default change: on lbnl-fpu the fully open damper (PFPU__VAVDMPRStuck_100pct) goes from shortfall fault to info (fault-deep, but 0.81 % of samples) and its finding from info to ok; the stuck-shut reheat valve stays a shortfall fault (29.9 %). New metric shortfall_depth_severity keeps the depth-only grade, and a caveat says when the share lowered it. shortfall_share_pct=None restores the old grading. The same gate is available for the overcooling tiers as share_pct, off by default (no overcooling verdict changes). Workbook zone-reheat-saturated: answers 3 and 5 updated, and the new grade pinned.
• More rules link to the PNNL Re-tuning material (#89). RULE_REFERENCES maps 48 rules, up from 35, each checked against the guide's own headings. sat_rogue_zone_census links the discharge-air-temperature guide, whose reset section bases a zone-driven reset on the zones served, setting aside the warmest and coolest. static_rogue_zone_census, static_cohort_starvation and damper_census link the static-pressure guide, whose "too high or too low" section reads the box damper positions. sat_cohort_starvation maps to chapters 5 and 7; reheat_capacity_shortfall, cohort_airflow and cohort_space_temp to chapter 7. condenser_bypass_leak, chiller_approach_fouling, chiller_staging_fleet, cooling_tower_fan_effort_drift and boiler_efficiency_drift map to chapter 8. The G36 reset-effectiveness rules, filter_fouling, g36_afdd, the DX and heat-pump rules and the other drift detectors stay unmapped, with the reason in a comment and in docs/REFERENCES.md. Reports gain "Learn more" links for these findings; no finding changes.
• lbnl-chiller data issue condenser-bypass-runs-implausible (#89, annotate). The five condenser-bypass runs hold a fixed bypass all year (0.917 / 0.967 / 0.988 of the condenser flow for 25 / 50 / 75), drive the condenser loop to 118-162 °F at p90 / maximum, and the two 75 % runs are the same data with the valve command at 0.0 in every row. The runs stay as published and scored. The ingest is unchanged, so every content_hash is too. The plant-cooling-tower workbook caveats and instructor notes now say "both 75 % runs" and link the issue.
• rbc-g36-ahu: the stale leaking_valve note is corrected. The rule has gated on a mapped fan status since 0.93 (#42). The catalog note and the template comment now say so, and say that the template still leaves the rule out until it is re-checked on these baselines.
• chw_plant_reset checks the reset's sign and recognises constant-flow plants (#86 item 2). New params design_deltaT_min_f (8 °F, was fixed), expected_reset_sign ("negative" default: an outdoor-air reset lowers CHWST as OAT rises; "positive", or "any" for the old either-way test), flow_mode ("auto" default, "constant", "variable") and constant_flow_cv (0.05). A clear slope the wrong way is no longer a reset: chwst_reset_present is false, the new chwst_reset_direction metric reads reverse (expected / flat otherwise), a caveat says so, and the finding warns. Role.CHW_FLOW is now an optional input: with flow_mode="auto", a flow whose coefficient of variation over the running hours is at most constant_flow_cv (on at least 24 hours) marks a constant-flow plant, whose low loop delta-T is reported but left out of severity, with a caveat. New metrics flow_mode (constant / variable / unknown), flow_cv and design_deltaT_min_f; CHWPlantResult gains trailing flow_cv / n_flow. The recommender gives no "fix low ΔT" advice on a constant-flow plant and has a new branch for a reversed reset, "Find why the chilled-water supply warms in hot weather". Intended default change on lbnl-chiller (chiller 1's flow varies by at most 0.11 % on every run): the fault-free plant goes fault -> ok (delta-T median 6.4 °F, 80.1 % of hours below 8 °F, now not judged), and so do the other 18 non-bypass runs (the chiller-bias runs' fault/warn were never this rule's to raise); the five tower-bypass runs go fault -> warn, their CHWST rising with OAT at +0.57 to +0.95 °F/°F.
• The pump rules learn the VFD floor (#86 item 3). chw_pump_dp_reset and hw_pump_dp_reset take near_min_pct (a number, or "auto") and floor_tol_pct (1.0). "auto", the chilled-water default, learns the pump's minimum speed from a plateau in its running speeds (camber.chwpump.learn_vfd_floor: the 2nd percentile, accepted when >= 10 % of running samples sit within floor_tol_pct of it and the 90th percentile is >= 20 points above it) and counts speeds at or below max(25, floor + floor_tol_pct) as near the minimum; with no plateau it falls back to 25. The hot-water rule keeps the fixed 25 % band by default. New metrics vfd_floor_pct, near_min_band_pct and near_min_source (learned / default / fixed) on both rules; the summary names the band. Intended default change on lbnl-chiller: the secondary pump's floor is learned at 34.5 %, so the fault-free run counts 31.3 % of hours near minimum (was 0), and the two high-reading chiller-sensor-bias runs, where the pump idles at its floor 68 % of the time, go ok -> warn. No other severity moves.
• Parameter registry: keyword-or-number ranges. A parameter that takes a keyword or a number lists the keywords first and ends with the numeric range (("auto", 10.0, 60.0)), printed as '"auto", or 10.0 to 60.0' by camber rules params and docs/THRESHOLDS.md. chw_plant_reset, chw_pump_dp_reset and hw_pump_dp_reset move from fixed-in-code notes to documented tunables.
• Workbook. plant-chw-reset-pumping: the healthy plant's chw_plant_reset is pinned ok with the constant-flow check, the stuck bypass warn with a reversed reset (+0.76 °F/°F), and the learned floor (34.5 %, 31.3 % of fault-free hours; the chiller-bias run a warn at 68.6 %); the page's setup, steps, questions and caveats and the instructor key (answers 1, 2, 4 and 5, discussion, mistakes) are rewritten. plant-sensor-vs-equipment: answer 2 notes the high-reading runs' pump now warns at its floor.
• The boiler firing rules run without a boiler run status (#86 item 4a). boiler_summer_lockout, boiler_short_cycle and hw_plant_deltat now declare roles_any_of = ((boiler_status, gas_input_rate),) instead of requiring boiler_status. With no run status mapped they read firing from the gas input above 5 % of its own 95th percentile (camber.schedules.plant_run_mask); a sample with no gas reading stays missing rather than counting as a stop. Such findings carry the metric run_source: "gas" and a caveat that a firing shorter than the resample interval is invisible. A frame that has a run status gives byte-identical findings. On lbnl-boiler (hourly) the three rules now give 17 findings each, all ok: the fault-free boiler fires 37.8 % of hours with 0.92 starts a day, 0 % of firing hours above the summer lockout, and a 36 °F loop delta-T median (12.2 % of hours below the 20 °F floor on the worst fouling run). The three rows leave the RCx report's "Checks not evaluated" table. The online monitor, the default evidence chart, the RCx confidence inputs and the triage sensor-precedence roles now count a rule's roles_any_of inputs too. The workbook exercise plant-boiler is updated.
• camber drift freeze refuses a config whose families declare a reference (exit 1, naming them), and drift accept / drift_refit leave those families out: there is nothing stored to freeze or move. The config docstring's "a run never mints its own baseline" now states the S4 exception. Docs: CLI.md ("A declared reference"), TUNING.md ("Drift references"), PLANT-DETECTORS.md and VALIDATION.md.
• Workbook. plant-cooling-tower pins the fan-effort findings and cooling_tower_fan_effort_drift TPR 67% (step 6, question 6, the instructor's answer 6 and a discussion point on the bypass runs, where the matched-load model extrapolates). plant-boiler: the point-in-time rules still miss the fouling, but boiler_efficiency_drift catches all three (TPR 100%, replacing the overall TPR 0 % answer); page steps 5 and 7, questions 4 and 5 and instructor answers 4 and 5 rewritten.
• reheat_penalty and overcooling_min_flow read the valve position when it is mapped (#85 item 3). Both judge heat actually delivered, so with heat_valve_position mapped they read it in place of the demand (new metric valve_signal: position or demand). When the demand is at or above 90 % while the position is at or below 5 % on at least 25 % of the occupied full-demand samples (>= 12 of them), the finding carries a "stuck or failed valve, not a reheat penalty" caveat (new metric valve_divergence_share). Sites with one valve point are unchanged. Intended default change (from the new lbnl-fpu mapping): on PFPU__ReheatVLVStuck_0pct reheat_penalty goes from fault (51 % "open", read from the demand) to ok (0 %) with the caveat, and overcooling_min_flow from fault to ok with the caveat; the healthy runs' figures hold (11 / 67 / 99 % open). reheat_capacity_shortfall, overcooling_severity and the reheat-valve drift detector keep reading the demand. Workbook: zone-reheat-overcooling question 5 and zone-reheat-saturated question 4 rewritten, with their instructor keys.
• leaking_valve can credit a unit's measured fan heat (#84 item 1, provisional). New opt-in params, all off by default so default outputs are byte-identical: measured_fan_heat_f (the unit's own fan rise; a cooling leak on the supply-air path is then a rise below measured_fan_heat_f - cool_delta_thr_f), cool_delta_thr_f (the cooling margin, default delta_thr_f), occupied_only (judge occupied samples, from the trended occupancy when mapped, else the weekday 07-18 schedule; Role.OCCUPANCY becomes an optional input only on such an instance) and judge_heating_on_supply_air (False judges a heating leak only on the heating coil's own leaving air). New metrics cool_shift_f and occupancy_gate appear only when configured, with caveats; LeakValveResult gains cool_shift_f, occupancy_gate and hw_judged, and analyze_leak_valves the same keywords. Documented in param_docs.py and docs/THRESHOLDS.md.
• The lbnl-sdahu template catches the published valve leak. Its leaking_valve entry sets measured_fan_heat_f: 1.0, cool_delta_thr_f: 1.0 and occupied_only: true, with a basis map. Intended default change: AHU__coi_leakage_010 goes ok -> fault (60.5 % of occupied, fan-on, valve-shut hours below the mixed air, was 0.6 % beyond the 3 °F margin); the fault-free run stays ok at 2.4 % and the damper runs under 2 %. The 1.0 °F was calibrated on the fault-free run, which is also a scored negative: its verdict is in-sample, and the template comment, docs/VALIDATION.md and docs/TUNING.md say so; a half-year split (calibrate on one half, judge the other) keeps the fault-free run ok (2.4 % / 2.7 %) and the leak a fault (54.3 % / 77.2 %). The lbnl-ddahu template comment records why the rule is not run there (its mapped supply air is the cold deck; run anyway it false-faults DDAHU__DMPRStuck_OA_0).
• LBNL benchmark (approved gated move, S1). examples/lbnl_fdd/benchmark.py builds LeakingValve from the same template params. SDAHU TPR 0.4 -> 0.6, accuracy 0.5 -> 0.6667, correct diagnosis 0.4 -> 0.6; pooled TPR 0.7 -> 0.8, accuracy 0.7692 -> 0.8462, correct diagnosis 0.7 -> 0.8; FPR unchanged at 0. No other benchmark key moves. The committed baseline is refreshed by the integrator.
• Workbook. air-heat-cool: the leak run is now a fault (TPR 100 %, FPR 0 %), the fault-free median rise reads +1.1 °F over occupied hours (was +1.0 °F over all fan-on hours); the page, questions, going-further steps and instructor key are rewritten around the calibrated fan heat and its circularity.
• RCx issue headings name the cause, not the remedy (#88 item 2). An issue page is headed "Issue N: {cause}", and the executive summary's Issue column shows the cause. The action paragraph reads "Recommended action — {title}: {action}", so the action titles stay in the report. RcxReport.to_dict() issues gain title and cause. The RCx golden file changes (intended). Workbook capstone: the top issue is now headed "Outdoor-air damper not modulating (stuck low)", with a damper repair as its action. The answer key and the page questions are updated.

• RCx report: a "Verify on site" section, on by default (#88 item 1). Section id verify (slot section:verify, so --notes-template writes it), after the issue pages and before Further reading; omitted when it would be empty, or left out with report.rcx.sections. A lead paragraph links chapter 9, then one table per kind of item: # (linked to the issue, or A for Appendix A), Equipment, Look at, Point, Confirms, Refutes. Further reading adds chapter 9 when the section is present. The RCx golden file changes (intended). Workbook capstone: step 2 now compares the student's checklist with the generated one, answer 4 of the instructor key points to it, and a new check pins the section's static-setpoint sensor item (trust 0.40) and the onset unit's damper item. The capstone's minimum outdoor air and high limit are site parameters in its config, so the section lists no design values for it; the key now says so.

• RCx trust table (#87). A unit with no fan signal is scored per its inferred operating mode in the gated column, and Gate used says "outliers read per mode: inferred off-mode (OA damper and coil valves closed)". On the catalog only irish-ahu changes: its supply air reads trusted 0.89 (was untrusted 0.19; plain sensor_trust now 0.27), its stray 2015 rows are left out (return air, OA damper and both valves suspect -> trusted), and its economizer high-limit issue's confidence moves M -> H. Every point flagged clipped shows the flag (one LBNL dual-duct return air). Fan-gated units are unchanged: reading them per fan mode too was measured and rejected (see docs/SENSOR-HEALTH.md). The workbook exercise data-trend-quality answers 4, 6 and 7 are rewritten.

• One free-cooling high limit, 60 °F (#91). camber.freecooling.free_cooling_opportunity defaulted high_limit_f to 65 °F while the free_cooling_missed rule used 60 °F. Both now read one constant, DEFAULT_FREE_COOLING_HIGH_LIMIT_F (60 °F), CAMBER's deliberately conservative screening default. Only a direct library call with no high_limit_f changes (it counts fewer free-cooling hours). Rule findings and reports are unchanged (the RCx economizer page passes economizer_high_limit's value and prints it), and so are the synthetic and fleet benchmarks. docs/TUNING.md gains guidance on a climate-appropriate dry-bulb high limit (the unit's sequence, the energy code's high limit for the climate zone, or the trends).
• condenser_water_reset takes the site elevation (#92). CondenserWaterReset and analyze_cw_reset gain elevation_ft and pressure_psia (default None), passed to stull_wetbulb_f when the wet-bulb is derived from OAT + RH. Without either, a derived wet-bulb is now caveated as sea-level, as the cooling-tower rules already did; that caveat is the only change to default output. A new top-level config key, site_elevation_ft (feet; validated by camber.config.site_elevation_ft), sets the elevation once for the site. It reaches cooling_tower_approach and condenser_water_reset and, through the drift section (run_drift, refit_baselines and build_drift_suite gain elevation_ft), cooling_tower_approach_drift and cooling_tower_fan_effort_drift. A rule's own elevation_ft or pressure_psia wins. On a tower that resets 1:1 with the true wet-bulb at 1,600 m in hot, dry air, the sea-level slope reads 0.93 and the corrected one 1.00, so reset_present can flip near reset_slope_flat. The catalog's plant data trend a measured wet-bulb and do not change. Documented in docs/CLI.md and param_docs.

• dcv_verification names fan-off hours and gains an opt-in duration fault (#93). Occupied hours below oa_floor_cfm with the supply fan off (from SUPPLY_FAN_STATUS, else SUPPLY_FAN_SPEED at or below the new fan_off_speed_pct, 5 %) are named in the summary and a caveat ("supply fan off while scheduled occupied"), counted as fan_off_occupied_pct / fan_off_occupied_hours, and left out of below_floor_pct, which is now the shortfall with the fan running. The share fault still reads every below-floor sample (below_floor_total_pct, the old below_floor_pct), so severity does not move; with no fan signal nothing changes. The new below_floor_fault_hours (default None, opt-in) faults a contiguous below-floor episode of that many occupied hours whatever its share; the longest episode (below_floor_longest_h, _start, _fan_off_h) is reported either way. assess_dcv gains fan_off_mask and the matching DcvResult fields. The ASO recommender reads a fan_off_occupied cause ("Run the supply fan whenever the space is occupied"). On lbnl-b59 the October and December 2020 days are now named fan-off (41, 82, 40 and 40 h); below_floor_pct goes 2.2 / 4.2 / 2.1 / 2.0 % -> 0.1 / 0.0 / 0.1 / 0.0 %, severity stays info, and the 2020 smoke-mode window stays unflagged. b4b-windesheim is unchanged and finnish-dcv (no fan signal) gains only the new metrics. docs/VENTILATION.md, docs/VALIDATION.md and the workbook zone-min-oa (caveat, answer 5 and two new checks) are updated.

• ornl-frp-vav declares actuator_stuck as its detector (#89 item 3, #85 item 2). labels.targets is now {"actuator_stuck": "terminal_damper"} and the template runs it; unmet_setpoint_hours stays as context. Full subset: TPR 17/18, 94 % [74-99 %] (Wilson 95 %), FPR 0/13, 0 % [0-23 %], against unmet_setpoint_hours' 10/18 and 1/13; default subset 6/6 and 0/1 (was 2/6 and 1/1). The miss is room 106 stuck at 100 % on a warm day, where an open damper is what the zone asked for. known_issues gains three lines: the airflow-bias days have no detector, a detection pointing upstream on the duct-static-collapse days is legitimate, and fleet and neighbour findings are not scored. The workbook exercise zone-bad-box scores actuator_stuck and reads its verdict on each day.

• A reheat valve that diverges from its demand gets a repair recommendation. When a reheat_penalty finding's valve_divergence_share is at or above 0.25 (the rule's caveat threshold; DEFAULT_PARAMS["reheat_valve_divergence_share"]), the advice is "Repair the reheat valve or actuator", with the cause "reheat valve stuck or failed shut (the controller calls for heat the valve does not deliver)", instead of "Minimize reheat".

Documentation

• docs/DATASETS.md: "Running the workbook answer checks on local files" (#89). Covers CAMBER_WORKBOOK_FROM_DIR, the <dir>/<dataset-id>/ layout, symlinking a checkout's differently named files, the manual lbnl-b59 files (Building_59.zip, README_Dryad_Bldg59.txt) and the per-dataset file list, which a test keeps in step with the harness and the catalog.
• ECOSYSTEM: open-fdd's ECM tooling. A reciprocal note covers open-fdd's ECM workbooks, their reference calculators and EnergyPlus-twin comparison, and its change-point and G14 helpers. It explains how its pre-retrofit estimates and CAMBER's post-retrofit measurement fit together, with measured inputs for the calculators (#22), and links the shared M&V vectors. The "not yet re-compared" note now applies to the fault conditions only.

Fixed

• docs/VENTILATION.md overstated the lbnl-b59 wildfire result (#85). It said the 2020 smoke mode was a below-floor fault on two rooftop units. Re-measured on the default subset (which holds the whole OA-flow record, April–December 2020): all four units are below the 750 cfm floor in 2.0–4.2 % of occupied hours, info under the 10 % fault share, and those hours are fan-off days in October and December. In the smoke-mode weeks (2020-08-24 to 09-06) the dampers sat at their 10 % minimum and the units still took in 1,130–3,625 cfm, 0–0.9 % below the floor. The page, docs/VALIDATION.md and the zone-min-oa caveat now say so.
• tests/test_references.py sees instance-named rules. _rule_names() now includes builtin.rule_names(), so a mapping for cohort_airflow or the census and starvation rules no longer fails the "every mapped rule exists" check.
• Stale rule docstrings and comments corrected (text only, no behaviour change). The supply_air_reset_compliance module described the SAT reset as a positive slope against OAT (it is negative since 0.92); dcv_verification said a trended occupancy is AND-ed with the schedule (it replaces it); the rogue-zone census said the fleet runner carries no topology (it auto-builds a naming one, which takes precedence over groups); chw_supply_tracking left out the chiller-power run gate added in 0.92; and the compressor_short_cycle / heatpump_defrost default comments got their arithmetic wrong (a 5 min timer allows about 12 starts an hour; hourly defrost is about 48 reversing-valve transitions a day), so 12 and 24 a day are screening ceilings, not generous ones. The camber rules params basis text for those two says so.
• Parameter registry: the occupancy_gate and share-gate entries. supply_air_control and damper_census occupancy_gate, and overcooling_severity shortfall_share_pct and share_pct, are documented in camber/rules/param_docs.py and docs/THRESHOLDS.md. A dict-valued tier map names its keys in its unit and gives the range of each value ("each value: 0 to 100"); overcooling_severity tiers follows the same convention, and a test checks every tier map's default values against the range. The damper_census fixed-in-code note no longer says occupied hours are always the weekday schedule.
• The synthetic healthy chilled-water reset ran the wrong way (S2). faultlab's clean chw_plant_reset scenario raised CHWST as OAT rose (42 + 0.30 x (OAT - 55)); a healthy outdoor-air reset lowers it. It is now clip(52 - 0.30 x (OAT - 55), 42, 52). The synthetic benchmark's metrics JSON is byte-identical before and after (sha256 a0d5c8fb...); without the fix the new sign check would have raised chw_plant_reset.fpr 0 -> 1. A test fixture in tests/test_optional_role_honesty.py had the same wrong-way "working reset" and is corrected.
• A constant held while the fan is off is not a clipped sensor. clipped now needs the pile-up on the fan-on samples too for a fan-dependent point (return air, airflow, OA airflow, duct static) on a unit with a fan signal: a BAS or gap-fill that parks the return air at a round 70.0 °F while the fan is off had read as a clip at its range limit. CO2 and the outdoor and space points are still judged on every sample (a CO2 transmitter topping out in a closed room overnight is a real clip). The four clips on the catalog store all remain; the RCx golden loses its four clipped tags on the demo units' fan-off return air.
• Every recommendation names a cause. The reversed chilled-water reset recommendation (from 098-plant-chw) gets the cause "Chilled-water supply warms in hot weather (plant capacity or a reversed reset)", and a test checks that every recommender call in camber.aso sets one.
• reheat_penalty fan_heat_f documents its keyword: its range is "auto", or 0.0 to 8.0 (the keyword-or-number convention from 098-plant-chw).
• The walk-down covers actuator_stuck and the DCV fan-off cause. camber.walkdown gains a SITE_CHECKS["actuator_stuck"] entry (a "contradicted" item when any actuator's flat run contradicts the zone's demand, else the unexplained-flat item) and a fan_off_occupied item for dcv_verification (#93), so every cause the DCV recommender can lead with has its own check. The reheat walk-down cause now reads DEFAULT_PARAMS["reheat_valve_divergence_share"], the threshold the recommender uses.
• A store with migrated partitions opens on pyarrow 17-24. migrate_partitions (0.95) read each legacy part file with pyarrow.parquet.read_table(path). On pyarrow 17 through 24 that applies hive partition discovery to a single file's own path, so a part under facility_id=X/year=Y/ came back with dictionary-typed facility_id and year columns, and the rewritten month files carried them. The store then refused to open ("Unable to merge: Field facility_id has incompatible types: dictionary vs string"). Part files are now read with ParquetFile(path).read(), which returns only the stored columns on every pyarrow; the retention rollups' legacy reads use the same helper. On pyarrow 14-16 and 25 nothing changes: a migrated store is byte-identical before and after. _migrate_year calls pyarrow.compute through call_function, so mypy passes with pyarrow builds that bundle type stubs (which do not declare the generated compute functions) as well as without them.
• A constant meter finds no steps on any BLAS. detect_step_changes round 1 scaled its segmentation by the first-difference noise and fell back to the residual variance, with no rounding floor: a constant meter's exact fit leaves residuals of zero or ~1e-15 depending on the BLAS build (Accelerate gives zero, OpenBLAS does not), and on OpenBLAS round 1 segmented that rounding noise into two zero-size "steps" that the later rounds kept. Round 1 now applies the same rounding floor as the later rounds (falling back to a unit scale below it).
• Dependency floors match what works. pyproject.toml now declares numpy>=1.24.1, pandas>=2.2.1 and pyarrow>=14.0.2 (were 1.24, 2.0 and 14). CAMBER and its tests use pandas 2.2 API (Index.round, the "ME" and "min" aliases); pandas 2.2.0 has a concat regression that left a SQL source's merged index unsorted; numpy 1.24.0 breaks matplotlib's masked fill_between on time axes; pyarrow 14.0.0/14.0.1 emit pandas 2.2's BlockManager DeprecationWarning on every store read (14.0.0 also carries CVE-2023-47248). A new CI job, min-deps, runs the suite on Python 3.10 with every core floor pinned (.github/min-deps.txt; a test keeps it equal to the declared floors).
• Change-point ties break the same way on every platform. The change-point grid searches in camber.mandv.models (3PC/3PH/4P, the to-zero variants, 5P/5PZ) kept the grid point with the strictly lowest SSE. On a flat SSE surface (no data between grid points, or a 4P/5P with no second regime) several points tie up to rounding, and the BLAS build (Accelerate vs OpenBLAS) picked the winner: fitting the five kinds to 1,961 BDG2 2016 meters, daily and monthly (19,610 fits), 170 fits' change points differed between the two builds. A later grid point now replaces the best only when it lowers the objective by more than 1e-10 of it (floored at 1e-12 of the weighted sum of y² for exact fits), so ties keep the first grid point; best_model treats BICs within 1e-9 as a tie (the earlier kind wins). The two builds now agree on every fit. Default output: the five benchmarks are byte-identical before and after on both builds, and no selected kind changes. Monthly change points do move within a tie, by at most 2.2 °F (151 fits on Accelerate, 166 on OpenBLAS; 29 of them the selected model, whose SSE is unchanged to 1e-10); daily fits do not move. The shared M&V vectors are regenerated: two non-selected candidates' change points move (syn_2p monthly 4P 79.762 -> 78.64791, and the bills_bdg2_rat_public_leta_elec 5P low change point 39.71 -> 38.62). The vectors test now compares non-selected candidates' change points within SCHEMA's ±2 °F and everything else exactly, and SCHEMA.md says how ties are broken.
• g36_afdd: shut valves alone are not free cooling (#94). The G36 operating-state classifier read every fan-on interval with both coil valves shut as OS#2 (free cooling), even with the outdoor-air damper shut. Following the G36 §5.16.14 operating-state definitions, OS#2 now also needs the OA damper open beyond its minimum position plus oa_damper_tol (5 points). At or below that, the interval is OS#5, where only the state-independent FC1-FC4 apply. These are deadband hours and unoccupied recirculation runs. OS#5 is used rather than "unclassified" because the definitions place an interval that fits none of OS#1-#4 there, and it keeps FC1-FC4. The minimum position is oa_damper_min. By default it is learned as the median damper command over fan-on mechanical-cooling intervals below econ_damper_open (the OS#4 position), or taken as 0 % with a caveat when there are fewer than 24 such intervals. A missing damper reading with both valves shut is unclassified. A frame without an OA damper point keeps the valves-only reading and gets a caveat. New finding metrics: oa_damper_min, oa_damper_min_source, idle_at_min_oa_hours, occupancy_gate and unoccupied_hours. New trailing G36Result fields: oa_damper_min, oa_damper_min_source, n_idle_at_min_oa and n_unoccupied. run_g36_afdd gains oa_damper_min=, oa_damper_tol= and occupied=, and classify_os gains oa_damper_min= and oa_damper_tol=.
• Occupancy. Unoccupied operation is still evaluated by default, since G36 suspends AFDD only while the AHU is not operating. A new occupancy_gate parameter ("off" by default, or "trended") limits the evaluation to the occupied hours of a trended occupancy point, with no assumed-schedule fallback. Once OS#2 is fixed it changes no verdict on lbnl-sdahu or lbnl-ddahu.
• lbnl-sdahu, full subset, at defaults.
  • The fault-free run goes from warn to ok: FC9 drops from 18.65 % to 0 %. All 323 FC9 hours were unoccupied hours with the fan at full speed and the damper at 0 %, and those 483 hours are now OS#5. The learned minimum is 10 %, the unit's documented fixed minimum.
  • onset_damper_stuck_025: FC9 drops from 11.65 % to 0 % (the same unoccupied pattern). The run stays fault on FC10 and FC11.
  • damper_stuck_075 and damper_stuck_100_short go from fault to ok. FC8 drops from 30.9 % and 26.8 % to 0 %, and FC12 from 8.75 % and 3.98 % to 0 %. Those hours had the damper commanded to its 10 % minimum with both valves shut, while the stuck damper let in 68-100 % outdoor air. That is OS#5 by command, so the free-cooling tests no longer apply, and their earlier hits came from hours that were wrongly read as free cooling. The G36 test for this fault is FC6 (outdoor-air fraction vs the minimum). With the template's own min_oa_pct 1.6, FC6 flags both runs (27.8 % and 28.1 %, fault), while the fault-free run reads 1.18 %.
  • coi_stuck_050: 7 hours move to OS#5, and FC14 goes from 13.06 % to 12.68 % (still fault on FC13).
  • The other ten runs are unchanged.
• Other datasets.
  • lbnl-ddahu DMPRStuck_OA_0: 186 hours with the damper at or below its learned 28 % minimum (176 of them at 0 %) move to OS#5, and FC8 goes from 16.2 % to 39.7 % (still fault on FC10). The fault-free and DMPRStuck_OA_100 runs are unchanged.
  • nuig-ahu101 is unchanged, and irish-ahu declines before and after (it has no fan signal).
• The lbnl-sdahu template runs g36_afdd with FC6 enabled. It passes the unit's documented minimum, min_oa_pct 1.6 (the 10 % fixed damper minimum measured as an OA fraction, the value outdoor_air_fraction already uses), with that provenance in the template comment and its basis map. The rule's own defaults are unchanged. From the template, damper_stuck_075 and damper_stuck_100_short read fault again on FC6 (27.8 % and 28.1 %), and the fault-free run reads ok (FC6 1.18 %). No other run's verdict changes; FC6 reads 0-1.5 % on the other runs.
  • Templates left unchanged. lbnl-ddahu states a seasonal minimum (31.8 %, 11.9 % in Jun-Aug). No single min_oa_pct works there: FC6 faults the fault-free run at 20.5 % with 31.8 and at 47.1 % with 11.9. g36_afdd declines irish-ahu (no fan signal) and lbnl-fcu (a fan-coil unit), so they are left as they were. No exercise config runs g36_afdd. camber datasets score lbnl-sdahu reads only the declared targets, so its score is unchanged.
• Unchanged. No gated benchmark key moves (the synthetic, fleet, LBNL, BDG2 and BDG2 savings benchmarks are all stable), and no workbook answer changes (no exercise runs g36_afdd).