"""Tests for predict_permits tool."""
import pytest
from src.tools.predict_permits import predict_permits, _extract_project_types
from src.tools.knowledge_base import get_knowledge_base
def test_extract_restaurant_keywords():
types = _extract_project_types("Convert retail to restaurant with grease trap", None)
assert "restaurant" in types
assert "change_of_use" in types
def test_extract_adu_keywords():
types = _extract_project_types("Convert detached garage to ADU, 450 sq ft", None)
assert "adu" in types
def test_extract_seismic_keywords():
types = _extract_project_types("Seismic retrofit of soft story building", None)
assert "seismic" in types
def test_extract_commercial_ti():
types = _extract_project_types("Office tenant improvement, new HVAC", None)
assert "commercial_ti" in types
def test_extract_historic():
types = _extract_project_types("Renovation of landmark building in historic district", None)
assert "historic" in types
def test_extract_override():
types = _extract_project_types("Some random text", ["restaurant", "change_of_use"])
assert types == ["restaurant", "change_of_use"]
def test_extract_default():
types = _extract_project_types("Minor work on a building", None)
assert "general_alteration" in types
@pytest.mark.asyncio
async def test_predict_restaurant_conversion():
result = await predict_permits(
project_description="Convert 1,500 sq ft retail space to restaurant in the Mission. "
"Need grease trap, hood ventilation, ADA compliance, new signage.",
estimated_cost=250000,
)
assert "restaurant" in result.lower()
assert "in_house" in result.lower() or "in-house" in result.lower()
assert "Planning" in result
assert "SFFD" in result or "Fire" in result
assert "DPH" in result or "Health" in result
@pytest.mark.asyncio
async def test_predict_kitchen_remodel():
result = await predict_permits(
project_description="Gut renovation of residential kitchen in Noe Valley, "
"removing a non-bearing wall, relocating gas line, new electrical panel. Budget $85K.",
estimated_cost=85000,
)
assert "Form 3/8" in result
assert "Confidence" in result
@pytest.mark.asyncio
async def test_predict_adu():
result = await predict_permits(
project_description="Convert detached garage to ADU in the Sunset, 450 sq ft, "
"new plumbing and electrical, fire sprinkler required.",
estimated_cost=180000,
)
assert "adu" in result.lower()
assert "in_house" in result.lower() or "in-house" in result.lower()
@pytest.mark.asyncio
async def test_predict_returns_gaps_without_cost():
result = await predict_permits(
project_description="Kitchen remodel"
)
assert "No cost estimate" in result
def test_knowledge_base_loads():
kb = get_knowledge_base()
assert len(kb._keyword_index) > 100
assert kb.decision_tree
assert kb.fee_tables
assert kb.otc_criteria
def test_semantic_concept_matching():
kb = get_knowledge_base()
concepts = kb.match_concepts("restaurant change of use grease trap")
assert len(concepts) > 0
def test_step_confidence():
kb = get_knowledge_base()
assert kb.get_step_confidence(3) in ("high", "medium", "low")
def test_concept_ranking_restaurant_first():
"""Restaurant-specific query should rank 'restaurant' highest."""
kb = get_knowledge_base()
scored = kb.match_concepts_scored("Convert retail to restaurant in the Mission")
assert len(scored) > 0
# Restaurant should be the top-ranked or at least top-3 concept
top_names = [name for name, _score in scored[:3]]
assert "restaurant" in top_names
def test_concept_ranking_scores_descending():
"""Scores should be in descending order."""
kb = get_knowledge_base()
scored = kb.match_concepts_scored("kitchen remodel with new cabinets and recessed lights")
if len(scored) >= 2:
scores = [s for _, s in scored]
assert scores == sorted(scores, reverse=True), f"Scores not descending: {scores}"
def test_concept_ranking_multiword_bonus():
"""Multi-word alias matches should score higher than single-word."""
kb = get_knowledge_base()
scored = kb.match_concepts_scored("earthquake brace bolt retrofit for my house")
assert len(scored) > 0
top_names = [name for name, _score in scored[:3]]
assert "earthquake_brace_bolt" in top_names or "seismic" in top_names
assert kb.get_step_confidence(6) == "medium" # Timeline is the gap
