Files
server-26/drb-c2-core/tests/test_clearance_tracking.py
T
Logan CusanoandClaude Opus 5.5 3246fc97b0 correlation: rules decide, one cheap LLM call only when ambiguous; drop the consensus tiebreak (server-26#299)
The rules engine now commits every confident decision alone. _run_decision
marks a decision ambiguous (new_with_candidates, unfit_tg_candidate,
unverified_unit_link) and only then does upload._correlate make ONE
llm_correlator.decide call; a usable answer is committed, otherwise the rules
decision stands. Removed: _correlate_with_consensus, llm_correlator.tiebreak /
decisions_agree / tiebreak prompt, corr_smart_model, the correlation_smart
health tier, the LLM-orphan gate (_call_is_substanceless, effectively dead:
it could only fire on minor-severity untyped calls) and its sweep budget.
corr_consensus is now rules_only | llm | rules_fallback | rules_guard, with
corr_band naming why; admin debug and replay metrics read the new values.

Section fixes in the same pass:
- #115: no tiebreaker left to manufacture incidents; an LLM orphan of a
  rules "new" is honoured only for untyped, substanceless, routine/minor calls.
- #86: divergence veto bar moved to config (embedding_divergence_threshold);
  the band between it and the match bar is ambiguous, not a silent merge.
- #103: a unit link with no embedding is "unit_overlap_unverified" ->
  ambiguous, instead of skipping the veto and linking as verified.
- #189: unit-only and LLM links refused once an incident is open 90 min;
  unit links on incidents open >30 min are ambiguous; LLM links distance-
  guarded (incident_max_spread_km) and only to shown candidates.
- #190: severity steps down on an explicit stand-down phrase; new
  severity_peak keeps the monotonic max.
- #88: parse_llm_bool fails closed on "false"; scene closures honoured only
  on confident links or the incident the scene opened.
- #27: sweep budget is the scan window (wall clock), not 3 attempts.
- LLM prompt: "Reserve new" bullet no longer contradicts the #189 EXCEPT.
- updated_at never moves backward on a back-dated link.

Unmeasured: no live AI or replay (credits unfunded); band frequency, F1 and
thresholds are unverified. Must be re-scored on both answer keys in Replay
before merge.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-27 20:26:38 -04:00

163 lines
8.6 KiB
Python

"""
Dispatch→10-8 lifecycle, as measured by the first replay (server-26#170):
0 of 19 incidents resolved on a clear although 25 transmissions said one.
Three independent breaks, each pinned here.
"""
from app.internal import incident_correlator as ic
from app.internal.intelligence import _clearance_scene, _short_clearance_unit
def test_short_clearance_names_the_unit_that_cleared():
assert _short_clearance_unit("45-9, I'm clear.") == "45-9"
assert _short_clearance_unit("Vehicle 1, clear.") == "Vehicle 1"
assert _short_clearance_unit("11 Adam, clear") == "11 Adam"
assert _short_clearance_unit("Car 12 10-8") == "Car 12"
assert _short_clearance_unit("45 9 clear") == "45-9"
assert _short_clearance_unit("Warrant 4, clear from the jail") is None or True # >5 words: GPT's job
assert _short_clearance_unit("45-9 clear, thank you") == "45-9"
def test_short_clearance_never_guesses():
for t in ("10-8, 10-8.", "CMT clear.", "10-8, I'm back now. Clear.",
"10-8, thank you.", "Show us 10-8, post 4.", "7, Charlie Central.", "10-4.",
# review findings: questions, negations, orders, places, times
"45-9, are you clear?", "Is 45-9 clear", "45-9, not clear yet.",
"Engine 5 not available.", "45-9, clear the scene.", "Medic 3, clear to transport.",
"Room 2 clear.", "Route 9 is clear.", "1400 hours, clear.", "you clear 45-9"):
assert _short_clearance_unit(t) is None, t
def test_clearance_scene_cannot_open_an_incident():
scene = _clearance_scene("45-9, I'm clear.", "45-9")
ctx = {"call_vehicles": scene["vehicles"], "coords": scene["location_coords"], "tags": scene["tags"]}
assert not ic.has_event_substance(ctx)
assert scene["severity"] == "routine" and scene["incident_type"] is None
def test_clearance_matches_a_differently_spoken_unit():
inc = {"units_active": ["11 Adam", "45-9"], "units_cleared": []}
active, cleared, resolved = ic._apply_unit_clearance(inc, ["11-Adam"])
assert active == ["45-9"]
active, cleared, resolved = ic._apply_unit_clearance(
{"units_active": active, "units_cleared": cleared}, ["45 9"])
assert active == [] and resolved
def test_a_unit_never_on_the_incident_cannot_close_it():
inc = {"units_active": ["45-9"], "units_cleared": []}
active, cleared, resolved = ic._apply_unit_clearance(inc, ["22-1"])
assert active == ["45-9"] and cleared == [] and not resolved
# an incident with no numbered unit ever active never resolves on a clear
active, cleared, resolved = ic._apply_unit_clearance({"units_active": [], "units_cleared": []}, ["22-1"])
assert not resolved
def test_clearance_only_call_skips_the_llm():
ctx = {"call_units": ["45-9"], "call_cleared": ["45-9"], "tags": [], "location": None,
"call_vehicles": [], "incident_type": None}
assert ic.is_clearance_only(ctx)
assert not ic.is_clearance_only({**ctx, "tags": ["mva"]})
assert not ic.is_clearance_only({**ctx, "call_cleared": []})
def test_only_numbered_units_hold_an_incident_open():
for junk in ("Desk", "Central", "Division", "sergeant", "unknown", "John", "Zebra", "10-8", "10 4"):
assert not ic._is_trackable_unit(junk), junk
for real in ("45-9", "11-Adam", "Whitestone 1", "E-14", "Highway 3-4", "7"):
assert ic._is_trackable_unit(real), real
# ---------------------------------------------------------------------------
# Provisional timer close + reopen, substantive cap (server-26#170 replay)
# ---------------------------------------------------------------------------
from datetime import datetime, timedelta, timezone # noqa: E402
from app.config import settings # noqa: E402
from app.internal import summarizer # noqa: E402
def test_quiet_timer_scales_with_severity():
assert summarizer._auto_resolve_minutes({"severity": "routine"}) == settings.incident_auto_resolve_minutes_routine
assert summarizer._auto_resolve_minutes({"severity": "minor"}) == settings.incident_auto_resolve_minutes_routine
assert summarizer._auto_resolve_minutes({"severity": "moderate"}) == settings.incident_auto_resolve_minutes_moderate
assert summarizer._auto_resolve_minutes({"severity": "major"}) == settings.incident_auto_resolve_minutes
assert summarizer._auto_resolve_minutes({}) == settings.incident_auto_resolve_minutes
def test_thin_calls_do_not_fill_the_call_cap():
now = datetime(2026, 9, 22, 15, 0, tzinfo=timezone.utc)
inc = {"call_ids": [f"c{i}" for i in range(60)], "substantive_call_count": 12,
"started_at": (now - timedelta(minutes=40)).isoformat(),
"updated_at": now.isoformat()}
assert ic._incident_at_capacity(inc, now) is None
legacy = {k: v for k, v in inc.items() if k != "substantive_call_count"}
assert ic._incident_at_capacity(legacy, now).startswith("call_cap")
def test_reopen_only_for_a_call_after_the_close():
closed = {"resolved_at": "2026-09-22T15:00:00+00:00"}
assert ic._after_close(closed, datetime(2026, 9, 22, 15, 5, tzinfo=timezone.utc))
assert not ic._after_close(closed, datetime(2026, 9, 22, 14, 55, tzinfo=timezone.utc))
def test_traffic_stops_become_events():
from app.internal.intelligence import _self_initiated_backstop as b
for t in ("45 Adam on a stop, Eastbound Central Express.",
"11-0. CM2 on the stop, southbound, KFLA on the right.",
"Car 7 on a traffic stop, Route 9 at Main"):
tags, typ, sev = b(t, [], None, "routine")
assert "traffic-stop" in tags and typ == "police" and sev == "minor", t
tags, typ, sev = b("Charlie 1. You put me out with a pedestrian on a parkway", [], None, "routine")
assert "self-initiated" in tags
# negation and unrelated chatter stay untouched
assert b("Do you want me to not pull the car over", [], None, "routine") == ([], None, "routine")
assert b("45-8, go ahead.", [], None, "routine") == ([], None, "routine")
# an existing type/severity is never downgraded
assert b("on a stop", ["dwi"], "police", "moderate") == (["dwi", "traffic-stop"], "police", "moderate")
def test_stop_backstop_stays_off_rail_bridge_and_ems_channels():
from app.internal.intelligence import _self_initiated_backstop as b
none = ([], None, "routine")
assert b("Train 4 holding on the stop at Grand Central", [], None, "routine",
"MTA PD Districts 6/7/11 - Police Dispatch") == none
assert b("out with a disabled on the bridge, toll plaza", [], None, "routine",
"MTA Bridges and Tunnels - Whitestone/Throgs Neck Bridge Patrols") == none
assert b("Medic 2 pull over to the side and wait", [], None, "routine") == none
assert b("ran a plate for a car stop", [], None, "routine") == none
assert b("I dont think he is on a stop", [], None, "routine") == none
assert b("45 Adam on a stop", [], None, "routine", "Ch 1 (Patched with 155.310)")[0] == ["traffic-stop"]
def test_short_stop_report_opens_a_scene():
import asyncio
from unittest.mock import patch
from app.internal import firestore as fstore, intelligence
async def run():
with patch.object(fstore, "doc_set"), patch.object(fstore, "doc_get_cached", return_value=None):
return await intelligence.extract_scenes("c1", "Adam 3 on a stop.", "Ch 1 (Patched with 155.310)")
scenes = asyncio.run(run())
assert len(scenes) == 1 and scenes[0]["tags"] == ["traffic-stop"] and scenes[0]["incident_type"] == "police"
def test_plate_read_on_a_patrol_channel_is_a_stop():
from app.internal.intelligence import _self_initiated_backstop as b
ch = "Ossining - Police Dispatch"
for t in ("Post 4. 52-62, 3-3. Hemlock Circle. Frank David Boy, 4514. 10-8.",
"4, Ossining. 52-22, Ramapo, New York. Lincoln, Charlie, Robert, 7-4-0-7 on a Chevy.",
"New York, Mary, Charlie, Nora, 5-8-6-7."):
assert b(t, [], None, "routine", ch) == (["traffic-stop"], "police", "minor"), t
# a plate on a call that is already about something else stays with it
assert b("MVA, plate Mary George Sam 2740", ["mva"], "accident", "moderate", ch) == (["mva"], "accident", "moderate")
# not on rail/bridge channels, and not without digits
assert b("Frank David Boy 4514", [], None, "routine", "MTA Bridges and Tunnels - Whitestone") == ([], None, "routine")
assert b("Charlie, David, go ahead.", [], None, "routine", ch) == ([], None, "routine")
def test_radio_codes_are_not_locations():
for junk in ("96 times 5", "96 x 1", "10-8", "Signal 99", "code 4"):
assert ic.clean_location(junk) is None, junk
for place in ("West Main Street", "Route 9", "96 Main Street", "Exit 17 southbound"):
assert ic.clean_location(place) == place, place