Drop the instruction line (line 6) from NWS wire output entirely.
Remove the _SAME_INSTRUCTION dict that was added in 503c16d.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Short, actionable instructions keyed by SAME event code (TOR -> "Seek
shelter immediately.", SVR -> "Move indoors now.", etc.). Falls back to
the CAP instruction field when no SAME code matches. Truncates at 40
UTF-8 bytes to keep wire size compact for mesh broadcast.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace single-line _render() with structured 6-line format:
L1: SAME emoji + event type + NWS office (from WMO identifier)
L2: area (first areaDesc segment, max 60 chars)
L3: hazard (from HAZARD.../TORNADO... or maxWindGust/maxHailSize params)
L4: impact (from IMPACT... in description)
L5: expires
L6: instruction (max 80 chars)
Add module-level helpers: _SAME_EMOJI, _NWS_OFFICE_SHORT, _nws_office(),
_parse_nws_description(). Emoji prefers SAME event code, falls back to
_emoji_for_event() substring match. All _render() call sites pass d=d.
Update test to match new format (coordinates removed from wire).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The enrichment pipeline writes to d._enriched, not d._enrichment.
Fix both _parse_state_511_incident and _parse_itd_511_incident.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ITD 511 sometimes sends lowercase direction strings (e.g. "east"
instead of "East"). Add lowercase and abbreviated lowercase keys
so the renderer resolves them without falling through to raw echo.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add min_magnitude dropdown (1-4), drop_non_present and
drop_zero_magnitude toggles to the TomTom Traffic adapter card.
State loads from /api/adapter-config/tomtom_incidents on mount
and saves changed keys on save, following the same pattern as
the WFIGS and fires config panels.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add tomtom_incidents.min_magnitude setting (default 4 = severe)
to adapter_config registry. Replace the hardcoded magnitude==0
drop check with a config-driven floor that silently drops any
TomTom event below the configured threshold.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Let road stay None when road_numbers is absent so the renderer
uses the from → to segment format instead of clobbering it with
the raw from string.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Line 2 now falls back to from_loc → to_loc when road is absent
(common for TomTom street-level incidents without road_numbers).
Line 3 renders length (meters from TomTom) as human-readable
distance (mi or m) alongside delay and lanes_affected.
Add length field to _parse_tomtom_incident return dict.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove the prefix parameter from _render() and all callers — the
New:/Update: labels are no longer surfaced in the multi-line format.
Add comment field extraction to _parse_state_511_incident and
_parse_itd_511_incident return dicts. Render comment as line 3b
when it provides additional context beyond lanes_affected and is
<=140 chars, skipping duplicates.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace single-line _render() with structured multi-line output:
Line 1: emoji + display name + city/state anchor
Line 2: road + full direction (Eastbound) + mile marker
Line 3: lanes affected + delay
Line 4: cause (if non-default)
Add _SUB_TYPE_DISPLAY and _DIRECTION_LONG mappings. Extend
_parse_state_511_incident and _parse_itd_511_incident return dicts
with lanes_affected, cause, description, and mile_marker fields.
Add mile_marker: None to _parse_tomtom_incident for consistency.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Events with severity=immediate skip the per-toggle cooldown check
entirely — they are already rate-controlled by source handler change
detection. Also set cooldown_seconds default to 0 (disabled).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
POST /api/debug/clear-cooldowns clears both in-memory toggle
cooldown map and SQLite dispatcher_cooldowns table.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Disable all standalone FIRMS broadcast paths (cluster, halt, spotting)
by inserting early return None. Growth events now use the shared WFIGS
_render() for consistent multi-line format with movement data, and set
_severity_override=immediate + _cooldown_suffix for per-fire cooldown.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Each WFIGS fire event now carries its irwin_id as _cooldown_suffix in
event.data. The dispatcher incorporates this suffix into the cooldown
key region field, giving each fire its own independent cooldown slot
instead of sharing one per (toggle, category, region).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove _build_prompt, _llm_render, _terse_fallback and all LLM backend
references. render_digest() now queries the fires table directly and
builds a structured multi-line wire: header with count, up to 5 fires
with acres/containment/anchor, and a +N more overflow line.
FireDigestScheduler no longer accepts or uses llm_backend. Updated the
pipeline __init__.py call site accordingly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Fire events are already change-detected by the wfigs handler so the
grouper coalescing window adds no value and causes commit callbacks
to be lost when events are replaced. Setting severity to immediate
unconditionally bypasses the grouper entirely.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Apply .title() to all town name returns in _location_anchor so
anchors render with proper capitalisation regardless of source
casing. Remove the redundant Near: prefix from the location line
in _render — the anchor text is self-describing.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Remove hamlet, suburb, and locality from _TOWN_OSM_VALUES so the
nearest_town Photon lookup only returns meaningful population centers,
avoiding misleading anchors from tiny named places.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Acres: prepend IncidentSize to _WFIGS_ACRES_RAW_KEYS so the normalizer
picks up the primary size field before falling back to DiscoveryAcres
and FinalAcres.
Location anchor: query the curated town_anchors table before falling
back to the Photon geocoder nearest_town call, giving consistent
anchor names for Idaho fires.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Normalizer: add fire_cause, agency, personnel, unique_fire_id from
WFIGS raw payload to the normalized incident dict.
Renderer: replace single-line wire format with structured multi-line
output — header, size/contained with bold deltas on updates, location
anchor, cause/discovered date, and unique fire ID. Update call sites
pass last_bcast_acres and last_bcast_contained for case-(iii) updates
to enable delta calculation and selective bolding.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
event.data can contain callables and internal _on_ callbacks that
cause json.dumps to fail with TypeError. Filter these out before
serializing to SQLite.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
WFIGS API returns both "WF" and "wildfire" as IncidentTypeCategory
values. The previous check only accepted "WF", silently dropping
wildfire-typed incidents.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When the non-WF filter returns None, the caller must check before
assigning to n[_kind]. Fixes TypeError on non-wildfire incidents.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Fire events now carry _severity_override in the data dict:
- New fires (case i/ii): priority by default
- Update fires (case iii): priority by default
- All fires: immediate if acres > 1000 OR contained_pct == 0
consumer.py checks for _severity_override before falling back to
map_severity(inner.severity). This ensures fire broadcasts are
prioritized appropriately in the dispatch queue.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add 'Central' as first tab in FAMILIES array (before Weather)
- Import Server icon from lucide-react
- Remove Central Connection card from header area
- Render Central config in its own tab panel (no adapter sub-tabs)
The Central tab now shows URL, Durable, and Region fields with the
enabled toggle, matching the previous inline card behavior.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Rework the NIFC Fire Perimeters tab to show all fire config in one place:
- Native polling settings (tick_seconds, state) shown only when native
- Broadcast Settings section (always shown):
- Broadcast on acres increase toggle
- Broadcast on containment increase toggle
- Update cooldown (hours, converted from cooldown_seconds)
- Freshness window (hours, 0 = disabled)
- Daily Digest section (always shown):
- Digest enabled toggle
- Schedule times (HH:MM list)
Changes persist to adapter_config table via PUT /api/adapter-config/{adapter}/{key}.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Fire events are always relevant regardless of age (a wildfire burning
for 9 hours is not stale — it's ongoing). The staleness filter was
designed for incidents with time_validity semantics, not persistent
fire state.
- defaults.py: add wfigs.freshness_seconds = 0 (disabled)
- dispatcher.py: for fire toggle family, read from adapter_config
instead of toggle; skip staleness check when freshness_s == 0
Fixes Blue Ridge fire being dropped after LAST_PER_SUBJECT replay.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- consumer.py: change DeliverPolicy from NEW to LAST_PER_SUBJECT to
get latest state per subject on reconnect instead of replaying backlog
- central_normalizer.py: drop RX and non-wildfire (non-WF) incident
types early in _parse_wfigs_incidents before they reach the handler
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
All three approved tiers in one commit. Reference.tsx is the deep docs
hub (8 new sections); the 10 other pages get short helper text +
tooltips that cross-reference back into Reference; 3 components get
operational-context tooltips. No new features land here -- this is the
copy that catches the GUI up to v0.6 + v0.7 system behavior.
Decisions applied per Matt's call:
- Keep both bang commands AND the LLM DM path (bangs are short on a
mesh-constrained interface; LLM is the anything-else path). Cross-
references between the two land in Reference -> Commands and
Reference -> LLM DM Queries.
- Rename "wire-string rendering" to "broadcast text" in user-facing
copy on TownAnchors.tsx, GaugeSites.tsx, and the Curation section of
Reference.tsx.
- Keep the "AND-model anti-pattern" tooltip as-is on Environment.tsx +
GaugeSites.tsx (specificity is the value for advanced users); the
OR-not-AND Reference section is its home definition that other
tooltips can link to.
Ham terminology preserved:
- Reference.tsx solar/Kp section retains "Quiet sun" / "Quiet HF
conditions" language (SFI/Kp vocabulary, not the deleted Quiet Hours
feature -- confirmed via direct grep before writing).
Tier 1: Reference.tsx (the depth doc) -- 8 new sections, ordered for
readability:
- "Fire Tracker (Fusion)": Phases 1-4 unified. Six fire-family alert
categories with example wire strings (wildfire_declared,
wildfire_growth, wildfire_halted, wildfire_spotting,
unattributed_hotspot_cluster, wildfire_incident). Attribution
mechanics (spread_radius_mi default, centroid as 24h median).
Movement mechanics (pass_id bucketing, per-pass centroid, 8-way
bearing, mi/h drift). Spotting mechanics (convex-hull perimeter +
vertex-distance approximation + per-fire cooldown). Daily LLM digest
(twice-daily summary broadcaster). The 10 fires.* adapter_config
knobs with defaults.
- "Broadcast Types": the three prefix categories -- New: (first sight),
Update: (material change), Active: (clock-driven reminder).
- "Reminder System": cadences per adapter (WFIGS 8h, SWPC 8h, ITD 511
per-zone). The tombstone (fires.tombstoned_at) termination. The
per-adapter reminder_enabled flag.
- "LLM DM (Natural-Language Queries)": all 7 env_reporter adapter
blocks (build_fires_detail / build_alerts_detail / build_quakes_detail
/ build_traffic_detail / build_gauges_detail / build_swpc_detail /
build_drop_audit) with example questions that hit each one. The
grounding clause behavior ("No active X right now" when an adapter
block is empty -- the v0.7-fire-tracker-4-final clamp). The
include_in_llm_context per-adapter toggle.
- "OR-not-AND Architecture": the per-adapter Central vs native
contract. Mutually exclusive. The AND-mode anti-pattern definition
(referenced by the Environment + GaugeSites tooltips). The Spokane
fix context.
- "Adapter Config & the CODE Rule": the GUI knob hub. The CONFIG-vs-
CODE split (thresholds in CONFIG, sentence templates / emoji /
translation maps in CODE). Restart-required vs live keys. The
include_in_llm_context toggle.
- "Curation: Gauges & Towns": Gauge Sites (NWS-AHPS thresholds, USGS
lookup, Action/Minor/Moderate/Major). Town Anchors (broadcast text
suffix lookup chain: Photon -> this table -> landclass -> county
-> coords). Example output "3 mi N of Almo".
- "Schema Migrations": light touch. v11-v16 schema additions tagged
with the phase they shipped under.
Tier 2: per-page tooltips and cross-references (10 pages):
- AdapterConfig.tsx: header paragraph extended with the CODE rule
pointer + LLM context toggle explanation.
- Alerts.tsx: !subscribe blurb extended with the three broadcast types
and links to Reference -> Broadcast Types + Reminder System.
- Config.tsx: environmental section description updated to point at
Environment.tsx for adapter knobs + Reference -> OR-not-AND for the
architecture.
- Dashboard.tsx: RF Propagation title carries SWPC R/S/G + Kp legend
tooltip; LOCAL badge defines what counts as local.
- Environment.tsx: Central region-token helper now references the
OR-not-AND section; tick_seconds defined inline as the native-mode
poll interval.
- GaugeSites.tsx: page description rewritten -- replaces "envelope
time" jargon with operational language, explains USGS lookup
mechanics, points at Reference -> OR-not-AND for the central-feed
disable.
- Mesh.tsx: Topology + Geographic buttons get tooltips defining the
rendering model.
- Notifications.tsx: band-conditions block extended with the daily
fire digest pointer + Reference -> Fire Tracker + Broadcast Types
cross-refs.
- TownAnchors.tsx: page description rewritten -- "wire-string
rendering" -> "broadcast text", chain fallback explained ("Photon
-> this table -> landclass -> county/state -> coords"), example
output included.
Tier 3: component tooltip polish (3 components):
- NodeTable.tsx: Battery + Last Heard column headers get title-bearing
spans with the voltage chart + offline-threshold legend.
- NodeDetail.tsx: SNR quality bands documented as a comment in the
neighbor render block (the legend lives next to where the colored
quality dots are computed).
- RestartBanner.tsx: banner copy extended with the restart-required
catalog (Config -> environmental, LLM backend swap, dispatcher
cold-start grace) so operators know what touched it.
Build verification:
- tsc + vite build green (one warning about chunk size > 500kB --
pre-existing).
- All 8 new TOPICS ids resolve in the served bundle:
adapter-config, broadcast-types, curation, fire-tracker,
llm-dm, or-not-and, reminders, schema.
- Distinctive new strings present in the bundle ("3 mi N of Almo",
"Photon nearest-town", "AND-mode anti-pattern", "R (Radio Blackouts").
- "Quiet sun" preserved (the ham SFI/Kp vocabulary in the Solar
section, not the deleted Quiet Hours feature).
- Container Up healthy, 0 tracebacks in 2 min post-rebuild.
Changelog: v0.7-docs-rewrite.md (per-page strip / rewrite / add table).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Follow-up to v0.7-fire-tracker-4-revised. 7-path verify identified 4
paths failing for 2 root causes: (A) _ENV_KEYWORDS_TO_SUBTYPE missing
"traffic" + drop-audit phrases so build_traffic_detail and
build_drop_audit never ran; (B) LLM hallucinated specific numbers when
env blocks were empty (fabricated 144 earthquakes from a blank
quake_events table). This commit widens the keyword catch (phrase-match
for multi-word triggers to avoid false positives) and adds a positive-
framed anti-hallucination instruction to the env-context system prompt.
Re-verified all 7 paths against real Gemini in the prod container;
verdicts in v0.7-firetracker-phase4.md.
Class A -- routing miss fix:
- _ENV_KEYWORDS_TO_SUBTYPE gains "traffic"/"commute"/"highway"
mapped to the existing "traffic" subtype.
- New _ENV_PHRASES_TO_SUBTYPE dict for multi-word triggers, matched
as whole-phrase substrings (NOT single-word membership). Drop-audit
phrases: "why didn't"/"why didnt"/"why am i not"/"why am i missing"
/"what was filtered"/"drop audit"/"filtered out" all map to a new
"drop_audit" subtype. Phrase-match keeps "why" alone from
false-positing every "why is X" question.
- _detect_env_subtype now checks phrases first, then falls back to
single-word tokenized match.
Class B -- positive-framed anti-hallucination clause:
- New module-level ENV_GROUNDING_CLAUSE constant. Appended to the
system prompt whenever env scope is detected (after env_block +
drop_block injection).
- Per Matt's mitigation guidance: positive ("answer from the blocks")
not negative ("do not hallucinate"). Wording:
"ENVIRONMENTAL CONTEXT GROUNDING:
Answer only from the environmental context blocks above. If a
block is empty or missing for an adapter the user asked about
(e.g. no NWS alerts in the block), say something like 'No active
<category> right now' -- never invent specific numbers, place
names, or counts. If you do not have a relevant block for the
question, say so briefly."
7-path verification, post-fix (real Gemini, prod container):
| # | query | method | verdict |
|---|-------------------------------------------------|----------------------|---------|
| 1 | "are there any fires near me?" | build_fires_detail | PASS |
| 2 | "any weather alerts?" | build_alerts_detail | PASS |
| 3 | "any earthquakes nearby?" | build_quakes_detail | PASS |
| 4 | "how's traffic on I-84?" | build_traffic_detail | PASS |
| 5 | "what's the snake river level?" | build_gauges_detail | PASS |
| 6 | "what are the band conditions?" | build_swpc_detail | PASS |
| 7 | "why didn't I hear about anything today?" | build_drop_audit | PASS |
Hallucination evidence (pre vs post on the quakes path):
pre-fix: "There have been 144 earthquakes of magnitude 1.5 or
greater in the past 24 hours worldwide. Some of the most
recent earthquakes reported include: A magnitude 2.1
earthquake in Pahala, Hawaii..." (fabricated)
post-fix: "I haven't observed any information about earthquakes
in the mesh data." (grounded)
Routing-miss evidence (pre vs post on the traffic path):
pre-fix: _detect_env_subtype("how's traffic on I-84?") -> None
-> env scope NOT triggered, build_traffic_detail never
called, LLM fabricated I-84 conditions in OR
post-fix: _detect_env_subtype("how's traffic on I-84?") -> "traffic"
-> env scope triggers, build_traffic_detail returns
185 chars of real Ada-county incident data, LLM
grounds on it: "I haven't observed any active
traffic incidents on I-84 within the last two hours.
The current active incidents are on North 9th Street
/ South 9th Street and SH-21, both in Ada."
Tests:
- 7 phase4 tests pass (no new tests needed; verification is the LLM
DM path itself).
- Full suite: 56 passed in 3.80s across phase1+phase2+phase3+phase4
+or-arch+include-roundtrip.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Matt review caught a scope error: ?status was a hypothetical sketch
in the design doc ("a node could ping ?status cache peak") treated as
authorization without asking. Ripping the structured-command path
entirely. The LLM DM path with env_reporter injection is the natural-
language interface; ?status was redundant infrastructure parallel to
the path the design depends on.
What landed:
- router.py: _maybe_rewrite_status_query + _lookup_fire_fuzzy +
_build_fire_status_context removed. route() restored to:
bang -> IGNORE-empty -> LLM with verbatim query.
- tests/test_fire_tracker_phase4.py: 5 ?status tests removed; replaced
with two regression guards:
test_natural_language_fire_question_routes_to_llm -- "how's the
cache peak fire?" returns RouteType.LLM with the verbatim query
(no in-router rewriting).
test_status_helpers_removed_from_router -- hard-block on
_maybe_rewrite_status_query / _lookup_fire_fuzzy / "?status"
appearing anywhere in router.py source. If anyone adds a
structured-command path for fires, this test fails and the
author has to talk to Matt first.
- 56 passed in 3.80s across phase1+phase2+phase3+phase4+or-arch+
include-roundtrip.
What stays (NOT ripped):
- Daily fire digest -- scheduled broadcaster, not a command. Its 4
adapter_config rows (fires.digest_enabled / digest_schedule /
digest_timezone / digest_max_chars) stay GUI-editable.
- Bug A fix (UnboundLocalError at router.py:745) -- independent of
?status. Confirmed still in effect.
LLM DM 7-path verification result -- 3 of 7 pass, INCOMPLETE:
| # | query | env_reporter | verdict |
|---|-----------------------------------------------|----------------------|---------|
| 1 | "are there any fires near me?" | build_fires_detail | PASS |
| 2 | "any weather alerts?" | build_alerts_detail | FAIL |
| 3 | "any earthquakes nearby?" | build_quakes_detail | FAIL |
| 4 | "how's traffic on I-84?" | build_traffic_detail | FAIL |
| 5 | "what's the snake river level?" | build_gauges_detail | PASS |
| 6 | "what are the band conditions?" | build_swpc_detail | PASS |
| 7 | "why didn't I hear about anything today?" | build_drop_audit | FAIL |
Two distinct failure classes:
Class A -- routing miss (#4 traffic, #7 drop):
_ENV_KEYWORDS_TO_SUBTYPE lacks "traffic" (only road/jam/crash/
closure/511/incident map to "traffic"), so a query literally
mentioning "traffic" never triggers env scope -> build_traffic_detail
never runs even though traffic_events has 9 rows on disk. The LLM
fell back to training data and hallucinated I-84 conditions.
build_drop_audit has no natural-language trigger phrase at all;
"why didn't I hear about anything today?" has no env keyword.
Class B -- empty data + LLM hallucination (#2 alerts, #3 quakes):
Env scope IS detected, build_alerts_detail and build_quakes_detail
DO run, but return empty because nws_alerts has 0 rows and
quake_events 24h-window has 0 rows (legitimate empty state). The
LLM has no env block to ground on and hallucinated "144 earthquakes
worldwide" -- sounds authoritative, is fabricated.
Not fixed in this commit -- needs Matt's call on:
(a) keyword additions to _ENV_KEYWORDS_TO_SUBTYPE for traffic +
drop_audit triggers (risk: false-positive env-scope triggers
for unrelated phrases).
(b) anti-hallucination prompt clamp: "If a topic's env block is
missing/empty, say you don't have live data instead of
answering from general knowledge." (risk: bot apologizes
every other message.)
Per the "STOP if any path fails" instruction, this commit does NOT
claim verification done; the report at
v0.7-firetracker-phase4.md has the full table + per-row mesh-receiver
wire + per-failure root cause analysis.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 4 of FIRMS+WFIGS fusion. Foundation: every direct LLM DM
mentioning a fire/weather/quake/avalanche/flood/etc. keyword was
failing silently in prod with UnboundLocalError because router.py
referenced scope_type before assigning it. With that path restored,
two new features land: a twice-daily fire-digest scheduled broadcast
(LLM-rendered) and a ?status <fire_name> on-demand mesh-DM intent.
BUG-FIX ROOT CAUSE (Job Zero):
router.py:745 ("if should_inject_mesh and scope_type == 'env'") read
`scope_type` -- a local variable bound only at line 761 inside an
unrelated `if self.source_manager and self.mesh_reporter` block.
Python's lexical scoping made scope_type a local of the whole
generate_llm_response function, so reading it before the assignment
raised UnboundLocalError on every env-keyword DM. The exception
propagated to main.py's outer except, no response went out, bot
appeared dead on fire/weather/quake/avalanche/flood queries.
Evidence (synthetic in-process trace against the live container's
config + GoogleBackend):
"are there any fires near me?" -> UnboundLocalError (pre-fix)
-> real LLM answer (post-fix)
"Yes, there are a few active
fires reported in the region.
Salmon River: 4,200 acres, 78%
contained. Cache Peak: 1,847
acres, 23% contained. ..."
"what's the weather?" -> UnboundLocalError (pre-fix)
-> "I do not have current weather
information. I can tell you
about active fires, stream gauge
levels, space weather, or band
conditions if you'd like." (post-fix)
"hi there" -> normal LLM answer in both cases
Fix: hoist `scope_type, scope_value = self._detect_mesh_scope(query)`
to right after `should_inject_mesh` is computed; remove the
now-duplicate detection inside the source_manager block.
Secondary mitigation: tightened the "do not invent commands" prompt
with an explicit "if no list appears above, you have NO commands"
clause. The prior prompt told the LLM "answer based on the command
list provided below" without always providing one, so the LLM
hallucinated plausible-sounding !commands (the "use ! commands"
canned-looking response Matt was seeing on non-env queries).
PHASE 4 FEATURES:
1. Fire-digest scheduler (meshai/notifications/scheduled/fire_digest.py).
Modeled after BandConditionsScheduler. Runs in the pipeline's
start_pipeline coroutine alongside band_conditions + reminders.
On each slot (default 06:00 + 18:00 America/Boise):
- Queries active fires (tombstoned_at IS NULL) + last 24h passes.
- Builds a prompt asking for a single mesh-wire summary <= 200
chars.
- Calls the LLM (Google/Anthropic/OpenAI per config).
- Falls back to a terse "Fires today (N): Cache Peak 1847 ac;
Twin Peaks 320 ac; +N more" line when the LLM is unavailable.
- Dispatches via dispatcher.dispatch_scheduled_broadcast (same
path band_conditions uses).
Idempotency: v16.sql adds fire_digest_broadcasts(slot_epoch PK,
sent_at, summary, source). INSERT OR IGNORE pattern blocks the same
slot firing twice (matters when container restarts mid-day).
2. ?status <fire_name> on-demand intent (router.py).
Before falling through to the LLM, route() now checks for a leading
"?status" / "status:" sigil or natural-language triggers like
"how is X fire?". On match:
- _lookup_fire_fuzzy walks fires by exact -> startswith ->
contains -> word-overlap (skipping a trailing " fire" word so
"cache peak fire" matches "Cache Peak"). Active fires rank
above tombstoned ones.
- _build_fire_status_context composes a small context block
(name, acres, containment, county/state, last 3 passes with
drift).
- The query is REWRITTEN into an LLM prompt with that context
inlined; the rest of the normal LLM path (chunking, history,
summary persistence) runs unchanged.
Live verification: "?status Cache Peak" -> "The Cache Peak fire is
1,847 acres and 23% contained. It's located in Probe / ID.";
"?status Salmon" -> word-overlap matches "Salmon River" ->
"The Salmon River fire is 4,200 acres and 78% contained, located
in Probe / ID."
3. adapter_config rows (GUI-editable per CONFIG-vs-CODE rule):
fires.digest_enabled = true (master toggle)
fires.digest_schedule = ["06:00", "18:00"]
fires.digest_timezone = "America/Boise"
fires.digest_max_chars = 200
Schema (v16.sql):
- fire_digest_broadcasts(slot_epoch INTEGER PK, sent_at, summary,
source) with source in {'llm', 'fallback_terse', 'skipped_no_fires'}.
- Index on sent_at for ops queries.
Tests (tests/test_fire_tracker_phase4.py, 10 cases all green):
- Regression guard: scope_type appears as an assignment BEFORE the
env_reporter check (prevents the UnboundLocalError from coming back).
- adapter_config seeds all 4 digest keys with expected defaults.
- render_digest returns ('', 'no_fires') when no active fires.
- render_digest falls back to terse line when LLM is None; wire fits cap.
- render_digest with a stub LLM returns ('<llm text>', 'llm').
- _lookup_fire_fuzzy: exact, "X fire" trim, word-overlap, no-match.
- _maybe_rewrite_status_query: builds context-bearing prompt; returns
None on non-status queries.
Combined suite: 60 passed in 3.81s across phase1+phase2+phase3+phase4
+or-arch+include-roundtrip.
Live verification on CT108 after rebuild:
- v16 migration applied (schema_meta=16, no Traceback in 3 min).
- FireDigestScheduler started: enabled=True schedule=['06:00','18:00']
tz=America/Boise.
- LLM DM probe (real Gemini) returns real answers on env queries
(Bug A fixed end-to-end).
- ?status Cache Peak + ?status Salmon return fire-specific summaries.
- render_digest with real LLM returns source=llm + non-empty wire.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 3 of FIRMS+WFIGS fusion. v15.sql adds perimeter_geojson to
fire_passes + last_spotting_broadcast_at to fires. FIRMS handler
computes convex hull of each pass on pass-boundary close; attributed
pixels >= 1.5 mi (configurable) from previous-pass perimeter emit
wildfire_spotting broadcast. Cooldown 1h between spotting broadcasts
per fire so rapid embers do not spam. wildfire_spotting category at
immediate severity -- spotting is the highest-actionable fire signal
(spread beyond perimeter). All thresholds GUI-editable. Phase 4 (LLM
summaries + on-demand queries) deferred.
Schema (v15.sql):
- fire_passes gains perimeter_geojson TEXT (nullable; populated by
_close_prev_perimeter at boundary). GeoJSON Polygon, single outer
ring in (lon, lat) order per RFC 7946, closed (first == last).
- fires gains last_spotting_broadcast_at REAL (per-fire cooldown
latch). Index (irwin_id, last_spotting_broadcast_at) for the
cooldown probe.
adapter_config (defaults.py REGISTRY):
- fires.spotting_distance_threshold_mi = 1.5 (float). Matches design
doc Phase 3 spec; design doc open question #6 lists this as TBD
pending real spotting observation data.
- fires.spotting_cooldown_seconds = 3600 (int, 1h). Suppresses
rapid-ember spam from a single satellite pass.
ALERT_CATEGORIES (notifications/categories.py):
- wildfire_spotting: immediate / fire. Highest fire severity --
spotting represents fire spread BEYOND the existing perimeter, the
most actionable detection signal.
FIRMS handler (central/firms_handler.py):
- _handle_pass_boundary now closes the prior pass's perimeter (convex
hull of fire_pixels via Andrew's monotone chain) on the first
boundary; subsequent in-pass pixels reuse the stored hull.
- _check_spotting runs for every attributed pixel: looks up the most
recent CLOSED pass (perimeter_geojson NOT NULL AND pass_id !=
current), point-in-polygon test, vertex-distance approximation
per design doc Q (sparse pixels make edge projection overkill at
VIIRS 375 m resolution), per-fire cooldown gate.
- Priority order: spotting (immediate) > growth (priority) > cluster
(priority) > halt (routine). Spotting preempts growth at the same
pixel because immediate > priority.
- Helpers: _convex_hull (Andrew's monotone chain), _hull_to_geojson
(RFC 7946 Polygon), _point_in_polygon (ray casting),
_close_prev_perimeter, _check_spotting, _prev_has_perimeter.
Wire string:
- wildfire_spotting: "🔥 Possible spotting <dist:.1f> mi <dir> of
<incident_name> perimeter" -- direction is 8-way bearing from the
previous pass's centroid to the spotting pixel.
Tests (tests/test_fire_tracker_phase3.py, 11 cases all green):
- Pass close stamps perimeter_geojson as a closed Polygon (6 hex
vertices -> 7-entry closed ring).
- Pixel 2 mi NE of perimeter fires spotting with distance in the
1.0..2.5 mi band (vertex-distance approximation) and direction NE.
- Pixel inside perimeter -> NO spotting wire.
- Second spotting candidate within 1h cooldown -> suppressed.
- Past-cooldown spotting fires again.
- Convex hull / point-in-polygon / GeoJSON round-trip helper tests.
- adapter_config seed for both new fires.* keys.
- wildfire_spotting category registered with immediate severity.
- 49 tests green across phase1/phase2/phase3/or-arch/include-roundtrip.
Live verification on CT108 after rebuild:
- v15 migration applied (schema_meta=15, no Traceback in 3 min).
- Container healthy.
Synthetic 25-pixel probe (PROBE-V07P3-*, cleaned up after):
- Pass A: 20 pixels in a ~0.3 mi circle. Perimeter stored on boundary.
- Pass B: 5 pixels at distances 0.5/1.0/2.0/5.0/7.0 mi from center.
Observed wires:
"🔥 Possible spotting 1.7 mi NE of Probe Spotting Fire perimeter"
"🔥 Possible spotting 4.7 mi NW of Probe Spotting Fire perimeter"
(Plus a Phase 2 growth wire on the first pass B pixel -- documented
side effect: single-pixel pass B centroid shows 0.5 mi drift from
pass A.)
- 7.0 mi E pixel: outside 5 mi spread, no broadcast (cluster check
found no co-located unattributed pixels). Cleanup confirmed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 2 of FIRMS+WFIGS fusion. v14.sql adds fire_passes table for
per-satellite-pass centroid tracking + drift computation. FIRMS handler
now detects pass boundaries (satellite + time bucket), computes pass
centroid (median of pass pixels), Haversine drift from previous pass,
bearing to 8-way direction, mi/h speed. Drift >= 0.5 mi (configurable)
emits wildfire_growth broadcast with wire including movement vector
and nearest-town context. Halt detection: fire with no new pixels for
>=12h (configurable) emits wildfire_halted broadcast (routine). Two
new ALERT_CATEGORIES: wildfire_growth (priority), wildfire_halted
(routine). All thresholds GUI-editable via adapter_config.fires.*.
Phase 3 (spotting) and Phase 4 (LLM summaries) deferred to subsequent
commits.
Schema (v14.sql):
- fire_passes table (irwin_id FK CASCADE, pass_id, pass_centroid_lat/lon,
pixel_count, total_frp, pass_started_at, pass_ended_at,
drift_mi_from_prev, drift_direction, drift_mi_per_hour). PRIMARY KEY
(irwin_id, pass_id) so the UPSERT path is cheap; secondary index on
(irwin_id, pass_ended_at) for the prev-pass lookup + halt counter.
- fires gains last_pass_id, last_pass_at, halt_broadcast_at columns.
halt_broadcast_at is latched per halt event; the detector filter
(halt_broadcast_at IS NULL OR halt_broadcast_at < last_pass_at)
reopens eligibility automatically when an idle fire receives a new
attributed pixel that advances last_pass_at.
adapter_config (defaults.py REGISTRY):
- fires.growth_drift_threshold_mi = 0.5 (float). Per-pass centroid drift
at or above this fires wildfire_growth. 0.5 mi matches the design
doc Phase 2 spec and is roughly 2x the VIIRS 375m pixel size (i.e.,
detectable as more than centroid jitter).
- fires.halt_passes_threshold = 2 (int). Documented intent; the
operational rule uses halt_minimum_seconds below as the time gate
because per-satellite pass-count enforcement would require modeling
the global VIIRS schedule per satellite. The 12h gate subsumes it
(4 passes/day in Idaho).
- fires.halt_minimum_seconds = 43200 (int, 12h).
ALERT_CATEGORIES (notifications/categories.py):
- wildfire_growth: priority/fire. FIRMS handler tags data["category"]
+ data["severity"] on the pass-boundary path when drift >= threshold.
- wildfire_halted: routine/fire. Halt detector tags data["category"]
+ data["severity"] when a fire transitions to idle for >=12h.
FIRMS handler (central/firms_handler.py):
- The Phase 1 attribution branch now passes through
_handle_pass_boundary(): UPSERT fire_passes row for the current
(irwin_id, pass_id) with median centroid + pixel count + total FRP
+ min/max acq_time; lookup the prior pass; compute drift mi +
8-way direction + mi/h speed and write them into the current pass
row (only the FIRST boundary fills these; subsequent in-pass pixels
COALESCE keep them stable). Update fires cursor (last_pass_id,
last_pass_at) and current_centroid_lat/lon to the latest pass
centroid -- this overrides Phase 1's 24h all-pixels median for
fires that have pass data.
- Growth wire emitted ONLY at the boundary (last_pass_id != current,
prev pass exists, drift >= threshold). Subsequent in-pass pixels
stay silent because pass_id == last_pass_id.
- _maybe_emit_halt runs as a final fallback when neither growth nor
cluster has fired. SELECT one fire matching the halt criteria,
stamp halt_broadcast_at, return the wire. The fallback ordering is
growth > cluster > halt so a busy fire's growth broadcast doesn't
starve a quiet fire's halt.
- New helpers: _bearing() (great-circle initial bearing, deg CW from N),
_direction_8() (compass 8-way mapping with +/-22.5 deg sectors).
Wire strings:
- wildfire_growth: `🔥 <incident_name> moving <dir> <speed:.1f> mi/h
~<dist_to_nearest_town:.1f> mi from <nearest_town>`. nearest_town
via meshai.central_normalizer.nearest_town (same Photon-backed
cache that wfigs_handler uses); failure falls back to bare
"moving <dir> <speed> mi/h".
- wildfire_halted: `🔥 <incident_name> no growth in <hours>h`.
Tests (tests/test_fire_tracker_phase2.py, 10 cases all green):
- 2-pass attribution with pass2 1.0 mi N of pass1 -> drift=1.0,
direction='N', mi/h computed, growth wire returned, data tagged.
- Drift below threshold (0.3 mi) -> NO growth broadcast; pass row
still records the (sub-threshold) drift for ops visibility.
- Halt detector: last_pass_at 14h ago -> fires once, halt_broadcast_at
stamped.
- Re-run halt detector with halt latched -> NO second broadcast.
- Halt re-eligibility: halt_broadcast_at < last_pass_at -> eligible
again (a resurrected then re-idled fire).
- Bearing + direction round-trip across all 8 cardinals.
- Direction sector boundary (22.5/67.5 deg) correctness.
- adapter_config seed for 3 new fires.* keys.
- Two new ALERT_CATEGORIES registered.
- 5-pixel single-pass aggregate (pixel_count, total_frp sum, median
centroid, started/ended_at min/max).
Phase 1 test fix:
- tests/test_fire_tracker_phase1.py::test_centroid_recomputes_as_median_across_passes
retimed to 12:00/12:10/12:20 so all 3 pixels land in one
N20 bucket. Phase 2 makes current_centroid_* the per-pass median
(latest pass overrides Phase 1's 24h median); the same-pass shape
preserves the original median-computation intent. 39 total tests
green across phase1/phase2/or-arch/include-roundtrip.
Live verification on CT108 after rebuild:
- v14 migration applied (schema_meta version=14, no Traceback in 3 min).
- adapter_config.fires.growth_drift_threshold_mi = 0.5
- adapter_config.fires.halt_passes_threshold = 2
- adapter_config.fires.halt_minimum_seconds = 43200
- Container healthy.
Synthetic 100-pixel probe inside prod container (PROBE-V07P2-*,
cleaned up after):
- Pass A (50 pixels @ 12:00-12:25, N20 bucket 329768): centroid
(44.30000, -115.50000), pixel_count=50, total_frp=975.0, drift=NULL
(first pass).
- Pass B (50 pixels @ 18:00-18:25, N20 bucket 329772, centered 1.2 mi
NE of A): centroid (44.31230, -115.48282), pixel_count=50,
total_frp=975.0, drift_mi_from_prev=1.1703 (~design target 1.2 mi
with -0.03 mi rounding), drift_direction="NE",
drift_mi_per_hour=0.209 (1.17 mi over 5.5h between pass ends).
- Growth wire: "🔥 Probe Movement Fire moving NE 0.2 mi/h, ~13.0 mi
from Long Creek Summit Home" (Photon nearest-town anchor populated
successfully).
- Exactly ONE growth broadcast (first pixel of pass B); 99 other
pixels stayed silent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 1 of the FIRMS+WFIGS fusion design doc. v13.sql adds fire_pixels
table for per-fire pixel history + spread_radius_mi/current_centroid_*/
last_hotspot_at on fires. FIRMS handler now attributes incoming pixels
to fires via point-in-circle within configurable radius (default 5 mi),
updating per-fire centroid as median of recent pixels. Unattributed
pixels go through a cluster detector: 3+ pixels within 1 mi within 60
min triggers a single unattributed_hotspot_cluster broadcast (Possible
new fire). Two new ALERT_CATEGORIES: wildfire_declared (priority,
WFIGS first-sight) and unattributed_hotspot_cluster (priority, FIRMS
cluster). All thresholds GUI-editable via adapter_config.fires.* and
adapter_config.firms.*. Phases 2-4 (movement analysis, spotting, LLM
summaries) deferred to subsequent commits.
Schema (v13.sql):
- fire_pixels table (irwin_id FK CASCADE, acq_time, lat/lon, frp,
satellite, pass_id, attributed_at). Indexed on (irwin_id, acq_time)
for centroid queries + on (acq_time) for Phase 2.
- fires gains spread_radius_mi (nullable; NULL => use global default),
current_centroid_lat/lon (median of last 24h pixels, distinct from
the WFIGS-declared anchor lat/lon), last_hotspot_at (Phase 2 halt
detector).
- firms_pixels gains attributed_at + cluster_broadcast_at + compound
index on (attributed_at, cluster_broadcast_at, acq_time) for the
cluster query.
adapter_config (defaults.py REGISTRY + ADAPTER_META):
- fires.spread_radius_mi_default = 5.0 (float)
- firms.cluster_min_pixels = 3 (int)
- firms.cluster_max_radius_mi = 1.0 (float)
- firms.cluster_time_window_minutes = 60 (int)
- ADAPTER_META["fires"] meta block (display_name + description).
ALERT_CATEGORIES (notifications/categories.py):
- wildfire_declared: priority/fire. WFIGS handler tags data["category"]
on cases (i)+(ii) [INSERT or row-exists-but-never-broadcast]; case
(iii) Update keeps the existing wildfire_incident category.
- unattributed_hotspot_cluster: priority/fire. FIRMS handler tags
data["category"] + data["severity"] when emitting the cluster wire.
FIRMS handler (central/firms_handler.py):
- Unchanged storage path: filter, INSERT OR IGNORE into firms_pixels.
- New _attribute_or_cluster() runs on every newly-stored pixel (dedup
hits skip -- the original insert had its shot already).
- Attribution: bbox prefilter on fires.tombstoned_at IS NULL, then
exact Haversine to fires(current_centroid_lat ?? lat,
current_centroid_lon ?? lon) inside spread_radius_mi (per-fire ?? global
default). Multi-match resolves to nearest (design doc Q2). On match:
INSERT fire_pixels, UPDATE firms_pixels.attributed_at,
recompute centroid as median of last 24h pixels for this fire.
- Cluster: on attribution miss, query firms_pixels WHERE attributed_at
IS NULL AND cluster_broadcast_at IS NULL AND acq_time > NOW-window.
If count >= cluster_min_pixels, fire the cluster wire and stamp
cluster_broadcast_at on every member so a 4th arrival cannot re-fire.
WFIGS handler (central/wfigs_handler.py): the existing prefix=New
branches (i)+(ii) now set data["category"]="wildfire_declared".
Existing _render() unchanged.
Wire strings:
- wildfire_declared: re-uses _render(prefix="New") -- emoji + name +
type + anchor + acres + containment + coords.
- unattributed_hotspot_cluster: _render_cluster_wire() emits
"Possible new fire: <N> hotspots within <r> mi @ <lat>,<lon>
(combined <total_frp> MW)".
Tests (tests/test_fire_tracker_phase1.py, 10 cases all green):
- Pixel within radius -> attribution + centroid + last_hotspot_at.
- Centroid recomputes as median across multiple passes.
- Pixel outside radius -> NO attribution + stays unattributed.
- 3 unattributed within 1 mi within 60 min -> cluster broadcast fires
exactly once, all 3 stamped cluster_broadcast_at.
- 4th pixel in the same footprint -> NO second broadcast (existing
3 are stamped so SQL filter excludes them).
- 5th-7th pixels 2h later -> form a NEW cluster (window prune fires).
- WFIGS first-sight tags data["category"]="wildfire_declared".
- WFIGS Update branch does NOT retag wildfire_declared.
- New adapter_config rows seeded on init_db.
- ALERT_CATEGORIES contains both new entries with correct toggle/severity.
Live verification on CT108 after rebuild:
- v13 migration applied (schema_meta version=13, no Traceback).
- adapter_config.fires.spread_radius_mi_default = 5.0
- adapter_config.firms.cluster_min_pixels = 3
- adapter_config.firms.cluster_max_radius_mi = 1.0
- adapter_config.firms.cluster_time_window_minutes = 60
- fires gains 4 new columns; firms_pixels gains 2 new columns; fire_pixels
table created.
- Container healthy, FIRMS pixels continue arriving (126 pre-deploy).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Pre-existing issue surfaced by v0.6-tail-3: prod config at
/data/config/config.yaml:82 uses !include to compose from separate
files, but the loader had no constructor registered so PUT
/api/config/<section> returned 500 with could-not-determine-constructor
when the section save path round-tripped YAML. This adds the !include
constructor (read path) + preserves the include structure on write so
multi-file config layouts work end-to-end via the GUI. The runtime
config behavior is unchanged; this only fixes the PUT-and-round-trip
case.
Implementation note: the read-only runtime path
(_load_yaml_with_includes) already had a working !include constructor
that recursively substitutes content. The bug was specifically in
save_section() -- it used plain yaml.safe_load() to re-read the target
file off disk for secret-ref preservation and for in-place section
updates. When target_file == "config.yaml" that file contains !include
directives for OTHER sections, and safe_load died on them.
Adding a third constructor that substitutes !include on save would
have flattened the multi-file layout to a single file the first time
anyone PUT an inline section. Instead this commit adds a preserve-mode
loader/dumper pair:
- _load_yaml_preserve() returns an Include("path") sentinel for each
!include node instead of recursing into the referenced file.
- _dump_yaml_preserve() re-emits Include("path") back to disk as
`!include path`. (PyYAML auto-quotes when the scalar contains a
period, so the on-disk form is `!include 'foo.yaml'`; both forms
are equivalent at parse time.)
- save_section()'s three yaml-touching sites (the secret-ref raw
read, the existing-target read, and the final dump) now use these
helpers. Local.yaml stays on yaml.safe_load/dump because local.yaml
never contains !include.
The runtime loader is untouched, so boot-time config still substitutes
includes and Config dataclasses see real values. Only the GUI's
section-save path round-trips through the preserve helpers.
Tests (tests/test_include_roundtrip.py, 8 cases):
- Runtime loader still substitutes !include content (regression guard)
- Preserve loader returns Include() sentinels
- Preserve dumper re-emits `!include path` (tolerant of PyYAML
auto-quoting)
- Read -> write -> read identity through the preserve helpers
- save_section('bot', ...) on a config.yaml that uses !include for
sibling sections succeeds AND leaves the includes intact on disk
(this is the exact prod PUT 500 case from v0.6-tail-3)
- After save_section, the runtime loader re-resolves all !include
files AND sees the saved change to the inline section
- save_section on a dedicated file (env_feeds.yaml) writes only that
file; config.yaml's !include directives are untouched
- Runtime cycle detection still trips on A!include->B!include->A
Live verification on CT108 after rebuild:
PUT /api/config/bot {"name":"AIDA","owner":"Malice","respond_to_dms":true,"filter_bbs_protocols":true}
-> HTTP 200 {"saved":true,"restart_required":false,"changed_keys":[]}
/data/config/config.yaml retains all 7 !include directives
(meshtastic, mesh_sources, mesh_intelligence, environmental,
notifications, llm, dashboard)
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Gap 1 -- env_routes.lookup_usgs_site no longer creates a temporary
USGSStreamsAdapter to hit USGS.gov directly. When the env_store has no
native usgs adapter (because usgs.feed_source != native), the endpoint
returns HTTP 404 with a body that says "site lookup unavailable in
central-feed mode; values must be entered manually or sourced from
Central". This closes the AND-mode anti-pattern Central's v0.10.2
report flagged: meshai was in central-feed mode for usgs but the
lookup helper would still call USGS.gov directly the first time the
dashboard opened the Add-Gauge form.
Gap 2 -- config_routes.RESTART_REQUIRED_SECTIONS gains "environmental"
and the PUT handler now diffs the section before/after, returning
{saved, restart_required, changed_keys}. restart_required is true only
when there are actual changes AND the section is in the restart-required
set, so a no-op PUT to environmental never raises a false alarm.
Frontend wiring:
- New RestartBanner component (yellow top-of-main banner) listens to a
meshai:restart-required CustomEvent + cross-tab storage event,
persists across navigations via localStorage, shows changed_keys
preview + Restart-now button (POSTs /api/system/restart) + dismiss.
- Layout.tsx mounts <RestartBanner /> above {children} so it surfaces
on every page.
- Config.tsx saveSection() now calls notifyRestartRequired(changed_keys)
alongside its existing setRestartRequired(true) when the API flags
the section.
- GaugeSites.tsx probes /api/config/environmental at mount and shows a
"USGS lookup" button next to the site_id input. The button is
disabled with an explanatory tooltip when usgs.feed_source != native,
and gracefully renders the 404 detail when the API returns 404 in
central-feed mode -- enter-manually UX, no silent fallback.
Tests -- tests/test_or_arch_continuous.py (11 cases, all passing):
- USGS lookup 404 with no env_store / no native usgs adapter
- 502 on native-adapter exception
- 200 + payload on native-adapter happy path
- environmental in RESTART_REQUIRED_SECTIONS
- PUT environmental with changed feed_source -> restart_required:true
+ changed_keys list including foo.feed_source dotted path
- PUT bot (non-restart section) -> restart_required:false
- No-op PUT to bot / environmental -> restart_required:false, empty
changed_keys
- _diff_keys helper unit tests (nested dicts, list-element changes)
Why this matters: per the Spokane post-mortem and Central's v0.10.2
response, both sides need belt-and-suspenders against transient
AND-modes. meshai's static OR enforcement at env_store boot is the
runtime guard; this commit makes the GUI honor it continuously --
the lookup helper can't sneak past it any more, and the user is told
explicitly that an environmental config change does not take effect
until the container restarts.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Phase 2 (b948ed7) stripped Notifications.tsx + backend config + pipeline
checks but missed Reference.tsx, which has live documentation describing
Quiet Hours as if it still exists. Per feedback-quiet-hours-trash.md the
intent is removal, not refactor -- the Reference page tells users about
a feature that does not exist. Strip the SectionHeader Quiet Hours block
(lines 952-956) and the quiet-hours clause from the rules-evaluation
checklist (line 922). Rebuild dashboard bundle so the JS no longer
contains the strings.
Reference.tsx changes:
- Line 922 rules-evaluation checklist: drop the 'Are we in quiet hours?'
clause from the 'MeshAI checks your rules' bullet.
- Lines 952-956 deleted: the SectionHeader 'Quiet Hours' + the two
paragraphs explaining 10pm-6am suppression and the 'Override Quiet
Hours' rule flag.
Other 'Quiet' matches in Reference.tsx are SFI/Kp solar terminology
('Quiet sun', 'Quiet. Best HF conditions.') and remain untouched --
unrelated to quiet hours.
Dashboard bundle rebuilt via vite. New artefacts:
meshai/dashboard/static/assets/index-WV9oBF1j.js (2,145.58 kB)
meshai/dashboard/static/assets/index-j88L17ja.css (45.08 kB)
Verification:
grep -rln 'quiet hours|quiet_hours|quietHours|override_quiet|Override Quiet' dashboard-frontend/src
-> 0 hits
grep -ic 'quiet hours|override.quiet|quiet_hours|quietHours' index-WV9oBF1j.js
-> 0 hits
Side artefact (not in commit): independent verification of Central
v0.10.2 Spokane-class leak fix completed today; all four meshai-side
checks pass (config audit, JetStream subscribe, event_log breakdown,
mesh_broadcasts_out scan). Report at OneDrive:
v0.6-tail-2-spokane-verify.md. No meshai-side code change needed.
(1) Auto-call refresh-toggles on PUT /api/config/notifications
meshai/dashboard/api/config_routes.py adds register_config_routes_hooks(app)
which registers a FastAPI HTTP middleware: on any 2xx PUT whose path
matches /api/config/notifications or /api/config, the middleware
invokes _refresh_toggle_filter(app) which reaches into app.state.bus._
pipeline_components["toggle_filter"] and calls .refresh(app.state.config).
The dashboard no longer has to remember to ping POST /api/notifications/
refresh-toggles after a toggle change. The explicit endpoint stays for
backwards-compat.
(2) env_reporter block-size cap moved to adapter_config
New registry row pipeline.env_reporter_block_chars (int, default 3000).
meshai/notifications/env_reporter.py replaces the hardcoded
_BLOCK_MAX_CHARS = 3000 with _DEFAULT_BLOCK_MAX_CHARS (the fallback) +
a _block_cap() helper that reads from adapter_config on every slice.
Mutating the row via PUT /api/adapter-config takes effect on the next
env_reporter call -- no restart.
(3) Bulk-import endpoint for gauge_sites
meshai/dashboard/api/gauge_sites_import.py adds
POST /api/gauge-sites/import with two paths:
format=csv -- expects "data" (CSV text with header row matching
gauge_sites columns: site_id, gauge_name, lat, lon,
and optionally action_ft/flood_minor_ft/
flood_moderate_ft/flood_major_ft/enabled). UPSERT
via ON CONFLICT(site_id) DO UPDATE. Returns
{inserted, updated, skipped}.
format=nws-ahps -- expects "wfo" (list of WFO codes). Fetches
water.weather.gov/ahps2/index.php?wfo=<WFO> for each,
regex-parses gauge links, then fetches up to 50
gauge detail pages per request and regex-parses
lat/lon + four threshold values. Best-effort; rows
stored under "AHPS-<gauge_id>" so they dont collide
with USGS-* ids. Returns the same shape plus
detail_fetched + errors list.
Frontend (dashboard-frontend/src/pages/GaugeSites.tsx) gains a
Import button + modal with two tabs (Paste CSV / Scrape NWS-AHPS)
rendered via an ImportModal component. CSV tab has a 48-row textarea
with the column-header hint inline; AHPS tab has a comma-separated WFO
input defaulting to BOI. Both submit via fetch() and show the JSON
response inline. Invalidates the curation cache server-side on any
successful insert/update so nwis_handler sees the new gauges on its
next call.
(4) WFIGS tombstone column -- CORRECTNESS
v12.sql adds fires.tombstoned_at REAL (nullable) + idx_fires_tombstoned_at.
meshai/central/wfigs_handler.py: the tombstone branch
(kind=="wfigs_tombstone") UPDATE fires SET tombstoned_at=COALESCE(
tombstoned_at, ?) so the first tombstone-time wins (idempotent against
repeated tombstone envelopes).
meshai/notifications/reminders/__init__.py: the wfigs tombstone
termination condition now checks row["tombstoned_at"] IS NOT NULL.
Reminders correctly STOP for closed fires -- before this change the
8h cadence would have kept Active: broadcasts going indefinitely past
a WFIGS removal.
SCHEMA_VERSION 11 -> 12.
(5) Delete INCIDENT_BROADCAST_HEARTBEAT_S
meshai/central/incident_handler.py: removed the dead constant
(v0.5.9 REVISED dropped the heartbeat path but left the constant
imported-but-never-read).
tests/test_incident_handler.py: removed the orphan
test_i_8h_heartbeat_triggers_update test (asserted None, used the
deleted constant for time arithmetic) and the stray import line.
Tests (tests/test_tail_followups.py, 16 cases):
- middleware fires refresh on PUT /api/config/notifications (200), does
NOT fire on PUT /api/config/llm
- env_reporter _block_cap() default 3000; mutate via PUT, invalidate,
next read returns the new cap
- CSV import inserts new rows, updates existing, skips bad rows,
rejects missing required columns, rejects bad format
- AHPS index parser extracts (gauge_id, name) from realistic HTML
- AHPS detail parser extracts lat/lon + four thresholds from realistic
HTML
- fires has tombstoned_at column after migrations
- wfigs tombstone branch stamps tombstoned_at
- ReminderScheduler skips a fire whose tombstoned_at is NOT NULL
- ReminderScheduler still fires for a fire whose tombstoned_at IS NULL
- INCIDENT_BROADCAST_HEARTBEAT_S no longer importable
Foundation/API test counts bumped:
REGISTRY 58 -> 59 (+ env_reporter_block_chars)
schema_meta v11 -> v12
Test count: 844 -> 859 (+16 new, -1 deleted dead test). 0 regressions.
Third broadcast type Active: clock-driven re-broadcasts of still-live
events at human-scale cadences. WFIGS fires 8h, itd_511 work zones daily
8 AM Mountain, SWPC G-storms 8h. NWS is NOT a clock reminder -- instead
the per-CAP-id dedup is relaxed to allow re-broadcast if >3h since last.
Schema split first_broadcast_at + last_broadcast_at on all reminder-
eligible tables. Wire prefix logic: New (first sight), Update (WFIGS
material change), Active (clock reminder). All cadences, channels, day-
of-week patterns, timezones, and termination conditions GUI-editable
from day one via the existing adapter_config editor. Termination:
tombstone OR containment_100 OR end_date_passed (no max-count). Quiet
hours not respected -- ripped out in Phase 2.
Schema (v11.sql):
- ALTER TABLE fires|nws_alerts|traffic_events|quake_events|swpc_events|
gauge_readings ADD COLUMN first_broadcast_at REAL
- Backfill: UPDATE ... SET first_broadcast_at = last_broadcast_at
WHERE last_broadcast_at IS NOT NULL
- ALTER TABLE adapter_meta ADD COLUMN reminder_enabled INTEGER NOT NULL
DEFAULT 0
- UPDATE adapter_meta SET reminder_enabled=1 WHERE adapter IN
('wfigs', 'swpc') -- itd_511_work_zone is a new meta row seeded
with reminder_enabled=1
- SCHEMA_VERSION 10 -> 11
Handler commit-callbacks (wfigs/nws/quake/swpc/incident):
- UPDATE ... SET last_broadcast_at=?, first_broadcast_at=COALESCE(
first_broadcast_at, ?) -- first_broadcast_at stamped once, never
overwritten
NWS handler (meshai/central/nws_handler.py):
- _render() gains a prefix kwarg
- After-first-broadcast branch: when (now - last_broadcast_at) >=
adapter_config.nws.duplicate_allowed_after_seconds (default 10800
= 3h), allow the re-broadcast with prefix=Active. Under the
window, suppress as before. The commit callback continues to
update last_broadcast_at.
ReminderScheduler (meshai/notifications/reminders/__init__.py):
- Async loop, ticks every 60s
- Each tick: SELECT adapter FROM adapter_meta WHERE reminder_enabled=1
- Per adapter, load reminders_<adapter> config from adapter_config
(cadence_kind, cadence_value, channels, terminate_when, dow_mask,
timezone)
- Interval cadence: rows where last_broadcast_at <= now - cadence_value
- Clock cadence: localizes now to configured tz, finds slots that
just passed in the last tick window, gated by dow_mask
- Termination conditions checked per adapter:
wfigs.containment_100 -> current_contained_pct >= 100
wfigs.last_event_age_24h -> last_event_at older than 24h
swpc.end_date_passed -> payload_json end_time in past
itd_511_work_zone.end_date_passed -> traffic_events.end_at in past
- Active: prefix on every emitted wire; dispatcher.dispatch_scheduled_
broadcast() honors cold-start grace, bypasses toggle path
- On success, last_broadcast_at = now; first_broadcast_at preserved
Launched from notifications/pipeline/__init__.py:start_pipeline()
alongside BandConditionsScheduler.
adapter_config registry (+15 new keys, 43 -> 58):
- reminders_wfigs.cadence_kind/cadence_value/channels/terminate_when
- reminders_swpc.cadence_kind/cadence_value/channels/terminate_when
- reminders_itd_511_work_zone.cadence_kind/cadence_value/channels/
dow_mask/timezone/terminate_when
- nws.duplicate_allowed_after_seconds
adapter_meta (+4 rows, 15 -> 19):
- reminders_wfigs, reminders_swpc, reminders_itd_511_work_zone
(pseudo-adapters carrying the reminder config)
- itd_511_work_zone (reminder target row; reminder_enabled=1)
- reminder_enabled flag added to wfigs/swpc (existing rows updated by
v11.sql) and to itd_511_work_zone seed.
Tests (tests/test_reminders.py, 10 cases):
- wfigs reminder fires past 8h cadence, stamps last_broadcast_at,
preserves first_broadcast_at
- reminder skipped within cadence
- reminder skipped when containment_100, last_event_age_24h
- swpc reminder fires (interval)
- work_zone clock reminder fires at 08:00 Mountain on enabled DOW
- work_zone reminder skipped when end_date_passed
- work_zone reminder skipped outside slot window
- reminder_enabled=0 suppresses all reminders for that adapter
tests/test_nws_dedup_relaxation.py (5 cases):
- First sighting renders without Active: prefix
- Re-broadcast within 3h suppressed
- Re-broadcast after 3h allowed with Active: prefix
- adapter_config.nws.duplicate_allowed_after_seconds override takes
effect (1h window verified)
- First sighting stamps first_broadcast_at=committed_at,
last_broadcast_at=committed_at; 4h later broadcast stamps
last_broadcast_at only, first_broadcast_at preserved
Test count: 829 -> 844 (+15 new, 0 regressions). Foundation tests
updated for new counts (REGISTRY=58, ADAPTER_META=19, schema=v11).
Closes audit doc section A.9 + finding #5. The last Phase-1 pipeline
state that lived only in instance memory now writes through to SQLite,
and ToggleFilter changes propagate without a container restart.
Schema:
v10.sql adds inhibit_state(key PK, rank, expires_at, updated_at) and
grouper_held(group_key PK, event_json, hold_until_at, updated_at).
Indexes on expires_at / hold_until_at support the prune sweeps.
SCHEMA_VERSION 9 -> 10.
Migration runner:
Fixed the alphabetical-vs-numeric sort bug v10 surfaced -- the runner
now sorts pending migrations by their integer version, not by
filename, so v10.sql correctly applies AFTER v9.sql (was applying
after v1 alphabetically, which made schema_meta stick at 9).
Inhibitor (meshai/notifications/pipeline/inhibitor.py):
- __init__ restores non-expired keys from inhibit_state on construct.
- handle() write-throughs every (key, rank, expires_at) tuple.
- _prune_expired DELETEs the same expired keys from disk.
- clear() (test path) drops the table.
Grouper (meshai/notifications/pipeline/grouper.py):
- __init__ restores non-expired held events from grouper_held; the
Event is rebuilt via Event.from_dict(json.loads(event_json)).
- handle() write-throughs (group_key, event_json, hold_until_at).
- tick() and flush_all() DELETE on emit.
ToggleFilter (meshai/notifications/pipeline/toggle_filter.py):
- new refresh(config) method re-reads config.notifications.toggles and
rebuilds the enabled set.
Live wiring:
- meshai/dashboard/api/config_routes.py adds a POST
/api/notifications/refresh-toggles endpoint that reaches into
app.state.bus._pipeline_components["toggle_filter"] and calls
refresh(app.state.config). The frontend pings this after PUT
/api/config/notifications so toggles take effect on the next event.
- meshai/main.py stashes self.event_bus on the dashboard FastAPI
app.state after build_pipeline so the route can reach it.
- Inhibitor.ttl_seconds and Grouper.window_seconds already read from
adapter_config.pipeline.{inhibitor_ttl_seconds, grouper_window_seconds}
via the v0.6-3b None-default wiring (rows seeded in v0.6-3a.1).
Tests (tests/test_pipeline_persistence.py, 11 cases):
- v10 tables present
- Inhibitor: state persists across simulated restart; expired rows
not restored; prune removes from disk; clear() wipes both.
- Grouper: state persists across restart; tick() clears disk;
expired rows not restored.
- ToggleFilter: refresh() picks up new enabled set; refresh(None)
is a no-op; disabling a family in config + refresh drops it.
Test count: 819 -> 830 (+11 pipeline persistence cases + schema test
bump).
Closes Section A.5 (gauge_sites) and A.12 (town_anchors) of the audit
doc by lifting both Python-dict curation lists into editable SQLite
tables. Operators can add/edit/disable rows from the dashboard without
a deploy; runtime reads go through cached accessors that invalidate
when the REST API mutates state.
Schema:
v8.sql adds gauge_sites(site_id PK, gauge_name, lat, lon, action_ft,
flood_minor_ft, flood_moderate_ft, flood_major_ft, enabled, updated_at).
v9.sql adds town_anchors(anchor_id AUTOINC PK, name UNIQUE, lat, lon,
state, enabled, updated_at).
SCHEMA_VERSION 7 -> 9.
Seed (meshai/persistence/curation.py):
_GAUGE_SITES_SEED carries the original 9 Idaho rows from
IDAHO_CURATED_SITES verbatim.
_TOWN_ANCHORS_SEED carries the 29 Idaho-and-neighbor towns from
_TOWN_COORDS verbatim.
seed_gauge_sites() / seed_town_anchors() INSERT OR IGNORE -- safe to
re-run; never overwrites user edits.
Handler integration:
- meshai/central/idaho_gauge_sites.py: IDAHO_CURATED_SITES dict deleted.
lookup_site() now calls meshai.persistence.curation.lookup_gauge_site()
which reads the table. THRESHOLD_RANK, normalize_site_id, and
compute_threshold_state remain in this module (CODE per Matt s rule).
- meshai/central/nwis_handler.py drops IDAHO_CURATED_SITES from its
import list; the table-backed lookup_site() is API-compatible.
- meshai/central_normalizer.py: _TOWN_COORDS dict deleted.
_compute_distance_bearing() now calls
meshai.persistence.curation.lookup_town_anchor() with the same
lowercased-name semantics it always used.
REST API (meshai/dashboard/api/curation_routes.py):
/api/gauge-sites GET list, GET one, POST add, PUT update, DELETE
/api/town-anchors GET list, GET one, POST add, PUT update, DELETE
Every mutation calls invalidate_curation_cache() so handler reads see
the new state on the next call -- no container restart.
Dashboard (dashboard-frontend/src/pages/):
- GaugeSites.tsx: table view with Add row / Edit row inline / Delete
confirm + per-row enabled toggle. 8 columns mirror the schema.
- TownAnchors.tsx: same pattern, 5 columns. Name is lowercased on
save to match the lookup key.
- Left-nav entries "Gauge Sites" (Droplets icon) and "Town Anchors"
(MapPin icon) added to Layout.tsx; routes added to App.tsx.
Tests (tests/test_curation.py, 18 cases):
- v8/v9 tables exist
- Seed lands every row from both dicts
- Seed idempotent; never overwrites user edits
- lookup_gauge_site hits/miss, disabled rows are invisible
- lookup_town_anchor case-insensitive
- REST API: GET list, GET one, GET 404, POST add, PUT update, DELETE,
POST missing-field 400; both gauge_sites + town_anchors
- Accessor reflects API mutations after invalidate_curation_cache()
tests/test_nwis_handler.py back-compat: IDAHO_CURATED_SITES dict alias
points at _GAUGE_SITES_SEED so the existing assertion suite still passes.
tests/test_adapter_config_foundation.py schema_meta v7 -> v9 bump.
Test count: 797 -> 819 (+18 curation cases + 4 maintenance updates).
Closes audit doc Section C. The LLM can now answer "any fires near me?",
"how are band conditions?", "why didnt I hear about that quake?"
without any tool-use / MCP / SQL pass-through -- via the same prompt-
injection contract mesh_reporter uses.
env_reporter (meshai/notifications/env_reporter.py):
- EnvReporter class with build_env_summary / build_fires_detail /
build_alerts_detail / build_quakes_detail / build_traffic_detail /
build_gauges_detail / build_swpc_detail / build_drop_audit / build_all
- Reads from fires + firms_pixels + nws_alerts + quake_events +
traffic_events + gauge_readings + swpc_events +
band_conditions_broadcasts + event_log + dispatcher_state
- Each build_*_detail() checks adapter_meta.include_in_llm_context for
the relevant adapter(s) before reading; turning the meta off via
/api/adapter-meta drops that adapters block out of the LLM prompt
- Defensive: missing meta row defaults to True (include); DB-unavailable
returns empty string; per-block 3000-char cap
- Module-level env_reporter singleton for the router
Router wiring (meshai/router.py):
- Extended _MESH_KEYWORDS dispatcher with _ENV_KEYWORDS_TO_SUBTYPE
mapping (fire/quake/flood/warning/storm/road/swpc/etc -> coarse
subtype). "flood" intentionally precedes "warning" so
"river flood warning" routes to gauges, not alerts
- _detect_env_subtype helper at module level (also test-importable)
- _is_mesh_question now also fires for env keywords -- single detector
per Matt s spec
- _detect_mesh_scope returns ("env", subtype) when an env keyword
matches, taking precedence over the node/region branches
- generate_llm_response: when scope_type == "env", appends
env_reporter.build_all() + env_reporter.build_drop_audit(hours=1)
to the system prompt. Wrapped in try/except so a reporter fault
never blocks the LLM call
Tests:
- tests/test_env_reporter.py (18 cases): meta gate, every build_*
method shape, build_all combines blocks, all-off produces empty
- tests/test_router_env_scope.py (18 cases): parametrized subtype
detection across fires/quakes/alerts/gauges/traffic/swpc, word-
boundary check (firearm != fire), synthetic-probe end-to-end
(seed fires table -> env_reporter emits a fires block with the
seeded row)
Test count: 761 -> 797 (+36 new, 0 regressions).
Closes the audit-doc Section A keystone (the GUI editor). Together with
v0.6-3a foundation, v0.6-3a.1 trim, and v0.6-3b handler wiring, every
Rule-17 CONFIG knob from the audit is now editable in the dashboard
without a container restart.
API (meshai/dashboard/api/adapter_config_routes.py):
GET /api/adapter-config -- {adapter: [{key, value, default,
type, description}]}
GET /api/adapter-config/<adapter> -- one adapter list
GET /api/adapter-config/<adapter>/<key> -- single row
PUT /api/adapter-config/<adapter>/<key> body {value} -- typed validation
int: int or whole-number float; rejects bool, fractional float, str
float: int or float; rejects bool
str: str only
bool: bool only
json: any JSON-serializable value
Every PUT calls invalidate_cache() so the next handler accessor
read sees the new value -- no container restart needed.
POST /api/adapter-config/<adapter>/<key>/reset -- value_json = default_json,
cache invalidated
GET /api/adapter-meta -- {adapter: {display_name,
include_in_llm_context, description}}
PUT /api/adapter-meta/<adapter> partial-update body, fields:
include_in_llm_context: bool, display_name: non-empty str
Dashboard (dashboard-frontend/src/pages/AdapterConfig.tsx):
- Per-adapter cards. Header row shows display_name, the include_in_llm_context
toggle, and an expand chevron. Adapters with zero config keys (e.g. itd_511)
still render so users can toggle their LLM-context inclusion.
- Expanded body lists each key with a type-aware widget:
bool -> checkbox, commit-on-change
int/float -> number input, commit-on-blur (or Enter)
str -> text input, commit-on-blur
json -> textarea, commit-on-blur (JSON.parse with inline error)
Each row shows the key name, type tag, description, "edited" badge when
value != default, a per-key Reset button, and a save badge (spinner,
check, error tooltip, or a small amber dot for unsaved local changes).
- Auto-save semantics: every blur/change/reset triggers PUT immediately;
no explicit Save button needed. Reset is one-click per key.
Wiring:
- meshai/dashboard/server.py registers the new router with prefix /api.
- dashboard-frontend/src/App.tsx adds the /adapter-config route.
- dashboard-frontend/src/components/Layout.tsx adds the left-nav entry
(Sliders icon, label "Adapter Config", after Reference).
- Vite build produces a fresh meshai/dashboard/static bundle. The
Dockerfile copies meshai/ so the new bundle ships with the container
image at next rebuild.
Tests (tests/test_adapter_config_api.py, 30 cases):
- GET grouped, per-adapter, single key
- GET per-adapter returns [] for adapters with zero keys (itd_511)
- PUT updates value, GET shows new value, accessor returns new value
(proves cache invalidation propagates to the in-process accessor)
- PUT type validation per (int, float, str, bool, json) incl. edge cases:
int rejects str + fractional float + bool but accepts whole-number float;
float accepts int + float, rejects bool; bool rejects int; str rejects
other types; json accepts list / dict / None
- PUT 404 on unknown key, 400 on missing value field
- POST reset restores default + invalidates cache
- GET /api/adapter-meta: include_in_llm_context defaults match registry
(central / geocoder false, rest true)
- PUT meta partial update: only provided fields change
- PUT meta rejects non-bool include_in_llm_context, empty display_name,
unknown adapter
Test count: 731 -> 761 (+30 API cases, 0 regressions).
Refs audit doc v0.6-phase1-audit.md Section A keystone + finding #4.
Replaces module-level magic numbers in 12 handlers with reads via the
v0.6-3a.1 typed accessor. Every default matches the prior hardcoded
value exactly, so first-deploy behavior is unchanged.
Handlers wired (43 keys across the 43-row registry):
wfigs cooldown_seconds, anchor_max_mi, broadcast_on_acres,
broadcast_on_contained
nws broadcast_severities, tombstone_msgtypes,
warning_suffix_promotes
usgs_quake regional_centroid, regional_radius_mi,
broadcast_pager_alerts, global_mag_floor,
regional_mag_floor, escalate_mag_floor
swpc geomag_kp_floor (extends G-scale down to Kp 5 when
lowered), flare_class_floor (R-scale
extended to M-class when lowered),
proton_pfu_floor
usgs_nwis parameter_codes, broadcast_on_recede
incident freshness_seconds, broadcast_on_update (Update path
re-implemented when toggled True:
magnitude step-up / delay doubling /
icon_category change)
tomtom_incidents drop_zero_magnitude, drop_non_present
state_511_atis skipped_states (case-insensitive match against both
state_code and primary_region suffix)
central severity_thresholds (immediate_min check ordered before
priority_max so the +inf clamp still
works)
dispatcher dedup_lru_max, cooldown_prune_size,
cooldown_prune_multiplier, dedup_db_retention_days
band_conditions swpc_freshness_seconds, hamqsl_url, hamqsl_timeout_s
geocoder (photon_url/timeout/radius/limit/town_osm_values/
h3_cache_max -- module-level constants kept as
backward-compat aliases; runtime reads via accessor)
pipeline Inhibitor.ttl_seconds + Grouper.window_seconds now
default to None, falling back to
adapter_config.pipeline.{inhibitor_ttl_seconds,
grouper_window_seconds}. Explicit constructor values
still win (test fixtures unchanged).
firms confidence_floor, frp_floor, bbox, dedup_distance_m
Schema:
v7.sql adds firms_pixels.dedup_key column + drops the old hardcoded
round(lat,5) UNIQUE INDEX, replaces with UNIQUE (dedup_key, acq_time,
satellite). The firms_handler quantizes lat/lon to
(dedup_distance_m / 111000) degrees at INSERT time -- meters-based
precision per Matt s spec, tunable via the GUI without schema changes.
SCHEMA_VERSION 6 -> 7. firms_pixels has 0 rows in production so no
backfill needed.
CODE preserved (Matt s rule): sentence templates, emoji literals, the
TomTom icon_map / ITD sub_type_map / Central adapter_map / category_map
translation tables, the band_conditions Kp/SFI -> Good/Fair/Poor
heuristic, anchor-priority ordering, expires-bucket boundaries, the
NOAA G/R/S scale tables. None of these reach the GUI.
Hot-path performance: every accessor read hits the in-memory cache after
the first call; cache hit is one dict get. Per-event reads (e.g. WFIGS
cooldown_seconds on every WFIGS poll-cycle) add a single dict lookup
to existing pipelines.
Backward-compat aliases retained for module-level imports that exist in
test code: WFIGS_BROADCAST_COOLDOWN_S, FIRMS_CONFIDENCE_FLOOR,
FIRMS_FRP_FLOOR, FIRMS_BBOX_OPTIONAL, INCIDENT_FRESHNESS_MAX_S,
PHOTON_BASE_URL/TIMEOUT_S/RADIUS_KM/LIMIT. Handler code reads via
adapter_config; tests can either monkeypatch the module attribute (firms)
or mutate adapter_config DB values.
Test count: 731 -> 731 (no new tests in 3b -- handler wiring is a pure
refactor; coverage comes from the existing handler test suites passing
unchanged).
Refs audit doc v0.6-phase1-audit.md Section A + Matt s CONFIG-vs-CODE rule.
