Add tests for routing and heuristics module
Browse files- tests/test_router.py +101 -0
tests/test_router.py
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
"""Tests for the routing module."""
|
| 2 |
+
|
| 3 |
+
import pytest
|
| 4 |
+
from cascade.router.heuristics import classify_by_heuristics, estimate_token_count
|
| 5 |
+
from cascade.router.routing_engine import RoutingEngine, RoutingDecision
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
class TestHeuristics:
|
| 9 |
+
"""Tests for heuristic-based classification."""
|
| 10 |
+
|
| 11 |
+
def test_simple_greetings(self, sample_queries):
|
| 12 |
+
"""Simple greetings should be classified as simple."""
|
| 13 |
+
for query in sample_queries["simple"]:
|
| 14 |
+
score, label = classify_by_heuristics(query)
|
| 15 |
+
assert label == "simple" or score < 0.5, f"Failed for: {query}"
|
| 16 |
+
|
| 17 |
+
def test_complex_coding_queries(self, sample_queries):
|
| 18 |
+
"""Coding queries should be classified as complex."""
|
| 19 |
+
for query in sample_queries["complex"]:
|
| 20 |
+
score, label = classify_by_heuristics(query)
|
| 21 |
+
assert label == "complex" or score > 0.7, f"Failed for: {query}"
|
| 22 |
+
|
| 23 |
+
def test_code_block_detection(self):
|
| 24 |
+
"""Queries with code blocks should be complex."""
|
| 25 |
+
query = "Can you fix this?\n```python\ndef foo():\n pass\n```"
|
| 26 |
+
score, label = classify_by_heuristics(query)
|
| 27 |
+
assert label == "complex"
|
| 28 |
+
assert score >= 0.85
|
| 29 |
+
|
| 30 |
+
def test_keyword_detection(self):
|
| 31 |
+
"""Keywords should trigger appropriate classification."""
|
| 32 |
+
# Complex keywords
|
| 33 |
+
assert classify_by_heuristics("write a function")[1] == "complex"
|
| 34 |
+
assert classify_by_heuristics("implement an algorithm")[1] == "complex"
|
| 35 |
+
|
| 36 |
+
# Simple keywords
|
| 37 |
+
assert classify_by_heuristics("hello there")[1] == "simple"
|
| 38 |
+
assert classify_by_heuristics("what is Python")[1] == "simple"
|
| 39 |
+
|
| 40 |
+
def test_length_based_classification(self):
|
| 41 |
+
"""Very short queries should be simple."""
|
| 42 |
+
score, label = classify_by_heuristics("Hi")
|
| 43 |
+
assert label == "simple"
|
| 44 |
+
assert score < 0.3
|
| 45 |
+
|
| 46 |
+
def test_estimate_token_count(self):
|
| 47 |
+
"""Token estimation should be reasonable."""
|
| 48 |
+
text = "Hello world" # ~3 tokens
|
| 49 |
+
estimate = estimate_token_count(text)
|
| 50 |
+
assert 2 <= estimate <= 5
|
| 51 |
+
|
| 52 |
+
|
| 53 |
+
class TestRoutingEngine:
|
| 54 |
+
"""Tests for the routing engine."""
|
| 55 |
+
|
| 56 |
+
def test_routing_decision_creation(self):
|
| 57 |
+
"""RoutingDecision should be created correctly."""
|
| 58 |
+
decision = RoutingDecision(
|
| 59 |
+
complexity_score=0.8,
|
| 60 |
+
complexity_label="complex",
|
| 61 |
+
recommended_model="gpt-4o",
|
| 62 |
+
routing_reason="High complexity query",
|
| 63 |
+
)
|
| 64 |
+
assert decision.complexity_score == 0.8
|
| 65 |
+
assert decision.complexity_label == "complex"
|
| 66 |
+
assert decision.recommended_model == "gpt-4o"
|
| 67 |
+
|
| 68 |
+
def test_model_selection_by_threshold(self):
|
| 69 |
+
"""Models should be selected based on complexity thresholds."""
|
| 70 |
+
engine = RoutingEngine()
|
| 71 |
+
|
| 72 |
+
# Simple -> local model
|
| 73 |
+
assert engine._select_model(0.2) == "llama3.2"
|
| 74 |
+
|
| 75 |
+
# Medium -> mini model
|
| 76 |
+
assert engine._select_model(0.5) == "gpt-4o-mini"
|
| 77 |
+
|
| 78 |
+
# Complex -> full model
|
| 79 |
+
assert engine._select_model(0.85) == "gpt-4o"
|
| 80 |
+
|
| 81 |
+
def test_threshold_boundaries(self):
|
| 82 |
+
"""Test exact threshold boundaries."""
|
| 83 |
+
engine = RoutingEngine()
|
| 84 |
+
|
| 85 |
+
# At lower boundary
|
| 86 |
+
assert engine._select_model(0.35) == "gpt-4o-mini"
|
| 87 |
+
|
| 88 |
+
# At upper boundary
|
| 89 |
+
assert engine._select_model(0.70) == "gpt-4o"
|
| 90 |
+
|
| 91 |
+
@pytest.mark.asyncio
|
| 92 |
+
async def test_route_query_returns_decision(self):
|
| 93 |
+
"""route_query should return a valid RoutingDecision."""
|
| 94 |
+
from cascade.router import route_query
|
| 95 |
+
|
| 96 |
+
decision = await route_query("Hello world")
|
| 97 |
+
|
| 98 |
+
assert isinstance(decision, RoutingDecision)
|
| 99 |
+
assert 0 <= decision.complexity_score <= 1
|
| 100 |
+
assert decision.complexity_label in ["simple", "medium", "complex"]
|
| 101 |
+
assert decision.recommended_model in ["llama3.2", "gpt-4o-mini", "gpt-4o"]
|