diff --git a/semantic_router/layer.py b/semantic_router/layer.py
index 222a5797a323171110820ce941df9d5803555cd6..4b7aa70d12f765a6c50030ba53f04fdaea88d215 100644
--- a/semantic_router/layer.py
+++ b/semantic_router/layer.py
@@ -204,10 +204,13 @@ class RouteLayer:
if text is None:
raise ValueError("Either text or vector must be provided")
vector = self._encode(text=text)
+ else:
+ vector = np.array(vector)
# get relevant utterances
results = self._retrieve(xq=vector)
# decide most relevant routes
top_class, top_class_scores = self._semantic_classify(results)
+ # TODO do we need this check?
route = self.check_for_matching_routes(top_class)
if route is None:
return RouteChoice()
@@ -218,6 +221,10 @@ class RouteLayer:
)
passed = self._pass_threshold(top_class_scores, threshold)
if passed:
+ if route.function_schema and text is None:
+ raise ValueError(
+ "Route has a function schema, but no text was provided."
+ )
if route.function_schema and not isinstance(route.llm, BaseLLM):
if not self.llm:
logger.warning(
@@ -228,10 +235,6 @@ class RouteLayer:
self.llm = OpenAILLM()
route.llm = self.llm
- elif text is None:
- raise ValueError(
- "Text must be provided to use dynamic route with function_schema"
- )
else:
route.llm = self.llm
return route(text)
@@ -382,7 +385,7 @@ class RouteLayer:
config = self.to_config()
config.to_file(file_path)
- def get_route_thresholds(self) -> Dict[str, float]:
+ def get_thresholds(self) -> Dict[str, float]:
# TODO: float() below is hacky fix for lint, fix this with new type?
thresholds = {
route.name: float(route.score_threshold or self.score_threshold)
@@ -400,7 +403,7 @@ class RouteLayer:
Xq: Any = np.array(self.encoder(X))
# initial eval (we will iterate from here)
best_acc = self._vec_evaluate(Xq=Xq, y=y)
- best_thresholds = self.get_route_thresholds()
+ best_thresholds = self.get_thresholds()
# begin fit
for _ in (pbar := tqdm(range(max_iter))):
pbar.set_postfix({"acc": round(best_acc, 2)})
@@ -447,7 +450,7 @@ def threshold_random_search(
) -> Dict[str, float]:
"""Performs a random search iteration given a route layer and a search range."""
# extract the route names
- routes = route_layer.get_route_thresholds()
+ routes = route_layer.get_thresholds()
route_names = list(routes.keys())
route_thresholds = list(routes.values())
# generate search range for each
diff --git a/tests/unit/test_layer.py b/tests/unit/test_layer.py
index 6e2ac317000a20f10d45cd671227e4831057e2da..c3c47e3def19d00ad1e85a6b78512a01974f197e 100644
--- a/tests/unit/test_layer.py
+++ b/tests/unit/test_layer.py
@@ -95,8 +95,18 @@ def routes():
@pytest.fixture
def dynamic_routes():
return [
- Route(name="Route 1", utterances=["Hello", "Hi"], function_schema="test"),
- Route(name="Route 2", utterances=["Goodbye", "Bye", "Au revoir"]),
+ Route(name="Route 1", utterances=["Hello", "Hi"], function_schema={"name": "test"}),
+ Route(name="Route 2", utterances=["Goodbye", "Bye", "Au revoir"], function_schema={"name": "test"}),
+ ]
+
+@pytest.fixture
+def test_data():
+ return [
+ ("What's your opinion on the current government?", "politics"),
+ ("what's the weather like today?", "chitchat"),
+ ("what is the Pythagorean theorem?", "mathematics"),
+ ("what is photosynthesis?", "biology"),
+ ("tell me an interesting fact", None)
]
@@ -124,10 +134,10 @@ class TestRouteLayer:
route_layer_none = RouteLayer(encoder=None)
assert route_layer_none.score_threshold == openai_encoder.score_threshold
- def test_initialization_dynamic_route(self, cohere_encoder, openai_encoder):
- route_layer_cohere = RouteLayer(encoder=cohere_encoder)
+ def test_initialization_dynamic_route(self, cohere_encoder, openai_encoder, dynamic_routes):
+ route_layer_cohere = RouteLayer(encoder=cohere_encoder, routes=dynamic_routes)
assert route_layer_cohere.score_threshold == 0.3
- route_layer_openai = RouteLayer(encoder=openai_encoder)
+ route_layer_openai = RouteLayer(encoder=openai_encoder, routes=dynamic_routes)
assert openai_encoder.score_threshold == 0.82
assert route_layer_openai.score_threshold == 0.82
@@ -157,12 +167,23 @@ class TestRouteLayer:
def test_query_and_classification(self, openai_encoder, routes):
route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
- query_result = route_layer("Hello").name
+ query_result = route_layer(text="Hello").name
assert query_result in ["Route 1", "Route 2"]
def test_query_with_no_index(self, openai_encoder):
route_layer = RouteLayer(encoder=openai_encoder)
- assert route_layer("Anything").name is None
+ assert route_layer(text="Anything").name is None
+
+ def test_query_with_vector(self, openai_encoder, routes):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
+ vector = [0.1, 0.2, 0.3]
+ query_result = route_layer(vector=vector).name
+ assert query_result in ["Route 1", "Route 2"]
+
+ def test_query_with_no_text_or_vector(self, openai_encoder, routes):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
+ with pytest.raises(ValueError):
+ route_layer()
def test_semantic_classify(self, openai_encoder, routes):
route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
@@ -186,6 +207,12 @@ class TestRouteLayer:
)
assert classification == "Route 1"
assert score == [0.9, 0.8]
+
+ def test_query_no_text_dynamic_route(self, openai_encoder, dynamic_routes):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=dynamic_routes)
+ vector = [0.1, 0.2, 0.3]
+ with pytest.raises(ValueError):
+ route_layer(vector=vector)
def test_pass_threshold(self, openai_encoder):
route_layer = RouteLayer(encoder=openai_encoder)
@@ -234,6 +261,24 @@ class TestRouteLayer:
assert (route_layer_from_config.categories == route_layer.categories).all()
assert route_layer_from_config.score_threshold == route_layer.score_threshold
+ def test_get_thresholds(self, openai_encoder, routes):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
+ assert route_layer.get_thresholds() == {'Route 1': 0.82, 'Route 2': 0.82}
+
+
+class TestLayerFit:
+ def test_eval(self, openai_encoder, routes, test_data):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
+ # unpack test data
+ X, y = zip(*test_data)
+ # evaluate
+ route_layer.evaluate(X=X, y=y)
+
+ def test_fit(self, openai_encoder, routes, test_data):
+ route_layer = RouteLayer(encoder=openai_encoder, routes=routes)
+ # unpack test data
+ X, y = zip(*test_data)
+ route_layer.fit(X=X, y=y)
# Add more tests for edge cases and error handling as needed.