Logan CusanoandClaude Sonnet 5
ef1e3d7f9d
The last leg of the #80/#95 scene-context leak. llm_correlator._call_block read call_doc's whole-call transcript for every scene, so on a multi-scene call every scene's cheap-tier and tiebreaker decision was made against text that also contained the other scenes. - intelligence.py: each processed[] scene now carries its own "transcript" — transcript_corrected, else this scene's segments joined, else (single scene) the whole transcript. - _build_context / preview_correlation / correlate_call: take a `transcript` param; _build_context resolves ctx["scene_transcript"] from it, falling back to the call doc (sweep, single-scene, tests) — the fallback is kept here, unlike embedding/severity, because a scene always has real text. - upload.py: both scene loops pass scene["transcript"]. - llm_correlator._call_block: reads ctx["scene_transcript"] (call-doc fallback retained for test-built ctx). - recorrelation_sweep: passes the call doc's text explicitly. - +1 regression test. Full c2-core suite green (296 passed, sandboxed venv). NOT for merge until the running correlation measurement window closes and its dump is analysed — deploying a correlator change mid-window would mix old and new behaviour in the sample. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
483 lines
21 KiB
Python
483 lines
21 KiB
Python
"""
|
|
An incident must not lie about what it is or where it is — server-26#23 / #26.
|
|
|
|
Both defects come from the 2026-08-20 production dump
|
|
(CORRELATION_REVIEW_0820.md) and both are the same shape: a field of the
|
|
incident header re-derived from whichever call linked most recently.
|
|
|
|
* `location` (the label) and `location_coords` (the map pin) were two
|
|
independent last-write-wins fields. A call could move one and not the
|
|
other, so they drifted: 5 of 6 incidents in the dump were pinned somewhere
|
|
other than the place they were labelled. `b9b4f392` said "100 South
|
|
Mosher" and pinned `Westmed`.
|
|
* `location` was never validated, so "Flames from 49" put `location: "49"`
|
|
on `9d376ffe` and the summarizer wrote "reported at location 49".
|
|
* `title` was re-derived from every classified call, so `b9b4f392` — opened
|
|
on a suspect search at 80 Grasslands Road — was named after its third
|
|
call, and `f5190670` after the thirteenth of its thirteen events.
|
|
|
|
The rules under test:
|
|
1. label and pin are ONE value, written together on every path;
|
|
2. a pin is only ever kept next to the label it was geocoded from;
|
|
3. a string with no word in it is not a place;
|
|
4. the title names the founding event and only ever escalates.
|
|
"""
|
|
import pytest
|
|
from datetime import datetime, timedelta, timezone
|
|
from unittest.mock import AsyncMock, patch
|
|
|
|
from app.internal.incident_correlator import (
|
|
_build_context, _create_incident, _update_incident,
|
|
_resolve_location_pair, _verified_pin, clean_location, location_is_unit,
|
|
)
|
|
|
|
NOW = datetime(2026, 8, 20, 7, 25, 0, tzinfo=timezone.utc)
|
|
|
|
# TG 383 from the dump: "Ch 1 (Patched with 155.310)" — a shared dispatch
|
|
# backbone, which is where every one of these chains happened.
|
|
DISPATCH_TG = "Ch 1 (Patched with 155.310)"
|
|
|
|
GRASSLANDS = {"lat": 41.0891, "lng": -73.8010}
|
|
WESTMED = {"lat": 41.0348, "lng": -73.7629}
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Harness — incidents are stored with merge=True, so folding each write back
|
|
# into the dict is exactly what Firestore does between calls.
|
|
# ---------------------------------------------------------------------------
|
|
|
|
async def _create(**call) -> dict:
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_set = AsyncMock()
|
|
await _create_incident(
|
|
call.get("call_id", "call-0"), "org-1",
|
|
call.get("incident_type", "police"), 383, DISPATCH_TG, "sys-1",
|
|
call.get("tags", []), call.get("location"), call.get("coords"),
|
|
call.get("units", []), [], None,
|
|
call.get("severity", "routine"), call.get("now", NOW),
|
|
)
|
|
return dict(mock_fstore.doc_set.await_args.args[2])
|
|
|
|
|
|
async def _link(inc: dict, **call) -> dict:
|
|
"""Run one call through _update_incident, fold the write back, return it."""
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_set = AsyncMock()
|
|
await _update_incident(
|
|
inc, call.get("call_id", "call-n"), 383, "sys-1",
|
|
call.get("tags", []), call.get("location"), call.get("coords"),
|
|
call.get("units", []), [], None, call.get("now", NOW),
|
|
talkgroup_name=DISPATCH_TG,
|
|
incident_type=call.get("incident_type"),
|
|
call_severity=call.get("severity", "routine"),
|
|
)
|
|
updates = dict(mock_fstore.doc_set.await_args.args[2])
|
|
inc.update(updates)
|
|
return updates
|
|
|
|
|
|
def _assert_pin_matches_label(doc: dict):
|
|
"""The invariant: a pin exists only alongside the label it was geocoded from."""
|
|
if doc.get("location_coords") is not None:
|
|
assert doc.get("location"), "pin with no label"
|
|
assert doc.get("location_coords_source") == doc["location"], (
|
|
f"pin sourced from {doc.get('location_coords_source')!r} "
|
|
f"but incident is labelled {doc['location']!r}"
|
|
)
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# server-26#23 — the label and the pin are one value
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_label_and_pin_stay_consistent_across_a_chain_of_calls():
|
|
"""
|
|
Replays `b9b4f392` exactly: a suspect search at 80 Grasslands Road, then an
|
|
EMS transport to Westmed, then a brand-new open-911 dispatch at 100 South
|
|
Mosher. Production ended up labelled "100 South Mosher" and pinned at
|
|
Westmed — the label from one call, the pin from another.
|
|
"""
|
|
inc = await _create(
|
|
tags=["suspect-search"], location="80 Grasslands Road", coords=GRASSLANDS,
|
|
severity="moderate", incident_type="police",
|
|
)
|
|
_assert_pin_matches_label(inc)
|
|
|
|
await _link( # 07:41 — "one female to Westmed"
|
|
inc, call_id="call-ems", tags=["ems-transport"], location="Westmed",
|
|
coords=WESTMED, incident_type="ems", now=NOW + timedelta(minutes=16),
|
|
)
|
|
_assert_pin_matches_label(inc)
|
|
|
|
await _link( # 07:55 — "open 911 line", a different job entirely
|
|
inc, call_id="call-911", tags=["open-911"], location="100 South Mosher",
|
|
coords=None, incident_type="police", now=NOW + timedelta(minutes=30),
|
|
)
|
|
_assert_pin_matches_label(inc)
|
|
|
|
assert inc["location"] == "80 Grasslands Road"
|
|
assert inc["location_coords"] == GRASSLANDS
|
|
# Every place anyone named is still recorded — that is what the map path
|
|
# is drawn from; it just isn't the incident's own location.
|
|
assert inc["location_mentions"] == [
|
|
"80 Grasslands Road", "Westmed", "100 South Mosher",
|
|
]
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_later_call_never_moves_the_pin_without_the_label():
|
|
"""The direct mechanism: coords updating on their own."""
|
|
inc = await _create(tags=["suspect-search"], location="80 Grasslands Road",
|
|
coords=None, incident_type="police")
|
|
assert inc["location_coords"] is None
|
|
|
|
updates = await _link(inc, tags=["ems-transport"], location="Westmed",
|
|
coords=WESTMED, incident_type="ems")
|
|
|
|
assert updates["location"] == "80 Grasslands Road"
|
|
assert updates["location_coords"] is None
|
|
_assert_pin_matches_label(inc)
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_pin_can_still_be_filled_in_for_the_same_place():
|
|
"""
|
|
The one permitted change. Geocoding is not deterministic in practice — it
|
|
needs the node's position, an API quota and a response — so the same
|
|
address can fail on one call and resolve on the next. Filling in a pin the
|
|
incident never had is not a move; matching is deliberately by exact label,
|
|
so it can never quietly re-point at a different street.
|
|
"""
|
|
inc = await _create(tags=["welfare-check"], location="55 Hyman Hills Road",
|
|
coords=None, incident_type="police")
|
|
assert inc["location_coords"] is None
|
|
|
|
await _link(inc, tags=["welfare-check"], location="55 Hyman Hills Road",
|
|
coords=GRASSLANDS, incident_type="police")
|
|
|
|
assert inc["location"] == "55 Hyman Hills Road"
|
|
assert inc["location_coords"] == GRASSLANDS
|
|
_assert_pin_matches_label(inc)
|
|
|
|
|
|
def test_a_pin_that_cannot_be_tied_to_the_label_is_not_shown():
|
|
"""
|
|
Every incident written before this change carries a pin with no record of
|
|
where it came from — and the dump says 5 of 6 of those are wrong. An
|
|
unverifiable pin is dropped, not displayed: a missing pin reads as missing
|
|
data, a wrong one reads as fact.
|
|
"""
|
|
legacy = {"location": "100 South Mosher", "location_coords": WESTMED}
|
|
assert _verified_pin(legacy) is None
|
|
|
|
resolved = _resolve_location_pair(legacy, None, None)
|
|
assert resolved["location"] == "100 South Mosher"
|
|
assert resolved["location_coords"] is None
|
|
assert resolved["location_coords_source"] is None
|
|
|
|
tagged = {"location": "Westmed", "location_coords": WESTMED,
|
|
"location_coords_source": "westmed"}
|
|
assert _verified_pin(tagged) == WESTMED
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# server-26#23 — "49" is not a place
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.parametrize("junk", [
|
|
"49", # 9d376ffe, from "Fire received. Flames from 49."
|
|
"10-24", # a ten-code
|
|
"5-5-2", # a unit designator
|
|
" ",
|
|
"",
|
|
None,
|
|
"1",
|
|
])
|
|
def test_bare_numbers_are_rejected_as_locations(junk):
|
|
assert clean_location(junk) is None
|
|
|
|
|
|
@pytest.mark.parametrize("place", [
|
|
"80 Grasslands Road",
|
|
"Westmed",
|
|
"Rt 9",
|
|
"226 East Main Street, apartment number 1",
|
|
])
|
|
def test_real_place_names_survive(place):
|
|
assert clean_location(place) == place
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_bare_number_never_reaches_the_correlator():
|
|
"""
|
|
Rejected at the context boundary, so it is not a location in the fit tests,
|
|
the thin-call test, the LLM prompt or the incident — and its coordinates go
|
|
with it, because they were geocoded from that very string.
|
|
"""
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_get = AsyncMock(return_value={})
|
|
mock_fstore.collection_list = AsyncMock(return_value=[])
|
|
ctx = await _build_context(
|
|
call_id="call-49", units=None, vehicles=None, cleared_units=None,
|
|
location_coords={"lat": 41.0, "lng": -73.8}, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="fire", location="49",
|
|
reassignment=False, create_if_new=True,
|
|
)
|
|
assert ctx["location"] is None
|
|
assert ctx["location_coords"] is None
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_scene_with_no_location_does_not_inherit_the_call_docs_pin():
|
|
"""
|
|
server-26#87. One call can be split into several scenes, and only the
|
|
primary scene's geocode is written to the call doc. A non-primary scene
|
|
that passes no location of its own must not inherit that pin — doing so
|
|
fabricates location_proximity, the strongest accept signal, for a scene
|
|
that has none, and drives it into the primary scene's incident.
|
|
"""
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_get = AsyncMock(
|
|
return_value={"location_coords": GRASSLANDS}
|
|
)
|
|
mock_fstore.collection_list = AsyncMock(return_value=[])
|
|
ctx = await _build_context(
|
|
call_id="call-scene-2", units=None, vehicles=None, cleared_units=None,
|
|
location_coords=None, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="police", location=None,
|
|
reassignment=False, create_if_new=True,
|
|
)
|
|
assert ctx["coords"] is None
|
|
assert ctx["is_thin_call"] is True
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_scene_does_not_inherit_the_call_docs_embedding_or_severity():
|
|
"""
|
|
server-26#80 / #95. Same shape as the #87 coords leak above:
|
|
intelligence.py writes only the PRIMARY scene's embedding and severity to
|
|
calls/{id}. A non-primary scene being correlated must be judged on its own
|
|
embedding (or none) and its own severity — not the call doc's — or a scene
|
|
about a different event scores against the wrong incident on the embedding
|
|
path and can inherit a minor/moderate/major rung it never had, clearing the
|
|
creation gate on borrowed weight.
|
|
"""
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_get = AsyncMock(
|
|
return_value={"embedding": [0.1] * 1536, "severity": "major"}
|
|
)
|
|
mock_fstore.collection_list = AsyncMock(return_value=[])
|
|
ctx = await _build_context(
|
|
call_id="call-scene-2", units=None, vehicles=None, cleared_units=None,
|
|
location_coords=None, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="police", location=None,
|
|
reassignment=False, create_if_new=True,
|
|
embedding=None, severity=None,
|
|
)
|
|
assert ctx["call_embedding"] is None
|
|
assert ctx["call_severity"] == "routine"
|
|
assert ctx["is_thin_call"] is True
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_scene_is_judged_on_its_own_embedding_and_severity():
|
|
"""The other half of #80/#95: the scene's own values are what land in ctx."""
|
|
scene_vec = [0.9] * 1536
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_get = AsyncMock(
|
|
return_value={"embedding": [0.1] * 1536, "severity": "routine"}
|
|
)
|
|
mock_fstore.collection_list = AsyncMock(return_value=[])
|
|
ctx = await _build_context(
|
|
call_id="call-scene-2", units=None, vehicles=None, cleared_units=None,
|
|
location_coords=None, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="police", location=None,
|
|
reassignment=False, create_if_new=True,
|
|
embedding=scene_vec, severity="major",
|
|
)
|
|
assert ctx["call_embedding"] == scene_vec
|
|
assert ctx["call_severity"] == "major"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_the_llm_tier_reads_the_scene_transcript_not_the_whole_call():
|
|
"""
|
|
server-26#102. intelligence.py writes only the primary scene's corrected
|
|
text to calls/{id}. _call_block (the LLM correlation prompt) must reason
|
|
over the SCENE being correlated, not a whole-call transcript that also
|
|
contains the other scenes. _build_context threads the scene's text in;
|
|
with no scene text it falls back to the call doc (sweep / single-scene).
|
|
"""
|
|
with patch("app.internal.incident_correlator.fstore") as mock_fstore:
|
|
mock_fstore.doc_get = AsyncMock(return_value={
|
|
"transcript": "scene one about a fire. scene two about a traffic stop.",
|
|
})
|
|
mock_fstore.collection_list = AsyncMock(return_value=[])
|
|
scene = await _build_context(
|
|
call_id="call-1", units=None, vehicles=None, cleared_units=None,
|
|
location_coords=None, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="police", location=None,
|
|
reassignment=False, create_if_new=True,
|
|
transcript="scene two about a traffic stop.",
|
|
)
|
|
fallback = await _build_context(
|
|
call_id="call-1", units=None, vehicles=None, cleared_units=None,
|
|
location_coords=None, reference_time=NOW,
|
|
system_id="sys-1", talkgroup_id=383, talkgroup_name=DISPATCH_TG,
|
|
tags=[], incident_type="police", location=None,
|
|
reassignment=False, create_if_new=True,
|
|
)
|
|
assert scene["scene_transcript"] == "scene two about a traffic stop."
|
|
assert fallback["scene_transcript"] == "scene one about a fire. scene two about a traffic stop."
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_bare_number_never_becomes_an_incident_location_or_title():
|
|
inc = await _create(tags=["flames"], location="49", coords=None,
|
|
incident_type="fire")
|
|
assert inc["location"] is None
|
|
assert inc["location_coords"] is None
|
|
assert "49" not in inc["title"]
|
|
assert inc["location_mentions"] == []
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# server-26#26 — the title names the founding event
|
|
# ---------------------------------------------------------------------------
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_an_unrelated_later_call_does_not_rename_the_incident():
|
|
"""`b9b4f392` again, from the title's side."""
|
|
inc = await _create(tags=["suspect-search"], location="80 Grasslands Road",
|
|
coords=GRASSLANDS, severity="moderate", incident_type="police")
|
|
assert inc["title"] == "Suspect Search at 80 Grasslands Road"
|
|
|
|
await _link(inc, tags=["ems-transport"], location="Westmed", coords=WESTMED,
|
|
incident_type="ems", severity="routine")
|
|
await _link(inc, tags=["open-911"], location="100 South Mosher",
|
|
incident_type="police", severity="moderate")
|
|
|
|
assert inc["title"] == "Suspect Search at 80 Grasslands Road"
|
|
assert inc["title_tag"] == "Suspect Search"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_routine_status_traffic_never_touches_the_title():
|
|
inc = await _create(tags=["welfare-check"], location="55 Hyman Hills Road",
|
|
incident_type="police")
|
|
updates = await _link(inc, tags=[], incident_type=None, units=["6-Adam"])
|
|
assert "title" not in updates
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_worse_event_takes_the_title_over():
|
|
"""
|
|
The one case where the header must change: a check-condition that turns
|
|
into a structure fire is a structure fire. Monotonic like _max_severity —
|
|
a calmer later call can never take it back.
|
|
"""
|
|
inc = await _create(tags=["check-condition"], location="226 East Main Street",
|
|
severity="minor", incident_type="police")
|
|
assert inc["title"] == "Check Condition at 226 East Main Street"
|
|
|
|
await _link(inc, tags=["structure-fire"], incident_type="fire", severity="major")
|
|
assert inc["title"] == "Structure Fire at 226 East Main Street"
|
|
assert inc["severity"] == "major"
|
|
|
|
await _link(inc, tags=["ems-transport"], incident_type="ems", severity="routine")
|
|
assert inc["title"] == "Structure Fire at 226 East Main Street"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_placeholder_title_is_filled_in_not_overwritten():
|
|
"""
|
|
An incident that opened on a call with no content tags is named after its
|
|
type ("Police — <talkgroup>"). That is a placeholder, not an event name,
|
|
so the first classified call may name it — and only the first.
|
|
"""
|
|
inc = await _create(tags=[], location=None, incident_type="police")
|
|
assert inc["title"] == f"Police — {DISPATCH_TG}"
|
|
assert inc["title_tag"] is None
|
|
|
|
await _link(inc, tags=["vehicle-accident"], location="Airport Road",
|
|
incident_type="police", severity="minor")
|
|
assert inc["title"] == "Vehicle Accident at Airport Road"
|
|
|
|
await _link(inc, tags=["disabled-vehicle"], location="Yonkers Avenue",
|
|
incident_type="police", severity="minor")
|
|
assert inc["title"] == "Vehicle Accident at Airport Road"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_the_title_picks_up_an_address_learned_later():
|
|
"""
|
|
Same event, new information — not a rename. The founding call classified
|
|
the event but named no place; a later call on the same event does.
|
|
"""
|
|
inc = await _create(tags=["welfare-check"], location=None, incident_type="police")
|
|
assert inc["title"] == f"Welfare Check — {DISPATCH_TG}"
|
|
|
|
await _link(inc, tags=["welfare-check"], location="55 Hyman Hills Road",
|
|
incident_type="police")
|
|
assert inc["title"] == "Welfare Check at 55 Hyman Hills Road"
|
|
assert inc["location"] == "55 Hyman Hills Road"
|
|
|
|
|
|
@pytest.mark.asyncio
|
|
async def test_a_legacy_incidents_title_is_not_claimed_by_the_next_call():
|
|
"""
|
|
Incidents created before this change have no `title_tag`, so their founding
|
|
event is unrecoverable. Their existing title is treated as the founding
|
|
one rather than handed to whichever call links next.
|
|
"""
|
|
legacy = {
|
|
"incident_id": "b9b4f392",
|
|
"title": "Suspect Search at 80 Grasslands Road",
|
|
"location": "80 Grasslands Road",
|
|
"call_ids": ["call-0"],
|
|
"started_at": NOW.isoformat(),
|
|
"updated_at": NOW.isoformat(),
|
|
}
|
|
updates = await _link(legacy, tags=["open-911"], location="100 South Mosher",
|
|
incident_type="police", severity="routine")
|
|
assert "title" not in updates
|
|
assert updates["location"] == "80 Grasslands Road"
|
|
|
|
|
|
# ── server-26#52: a unit call-sign must never become a map pin ────────────────
|
|
#
|
|
# "Post 1-2" passed clean_location (it has a word in it), geocoded against the
|
|
# Ossining anchor and produced a confident pin in the right town for an event
|
|
# with no known location. It was in the same incident's `units` all along.
|
|
|
|
@pytest.mark.parametrize("location,units", [
|
|
("Post 1-2", ["1-2", "Lincoln", "Post 1-2"]), # the dump's actual incident
|
|
("post 1-2", ["Post 1-2"]), # case-blind
|
|
("Post 1-2.", ["Post 1-2"]), # punctuation-blind
|
|
("Engine 4", ["Engine 4", "Ladder 1"]),
|
|
])
|
|
def test_location_matching_a_unit_is_rejected(location, units):
|
|
assert location_is_unit(location, units) is True
|
|
|
|
|
|
@pytest.mark.parametrize("location,units", [
|
|
("Water Street", ["1-2", "Post 1-2"]), # a real place, same incident
|
|
("South High", []), # no units extracted
|
|
("Riverdale Station", ["Lincoln"]),
|
|
("", ["Post 1-2"]), # nothing to compare
|
|
(None, ["Post 1-2"]),
|
|
])
|
|
def test_real_places_survive_the_unit_check(location, units):
|
|
assert location_is_unit(location, units) is False
|
|
|
|
|
|
def test_unit_check_does_not_match_on_substrings():
|
|
"""`1-2` is a unit; "1-2 Main Street" is an address that contains it."""
|
|
assert location_is_unit("1-2 Main Street", ["1-2"]) is False
|