unit tests for lance (#3)

* untested lance implrementation

* Update lancedb.py

* 2nd untested implementation

* Update lancedb.py

* added open_table workaround

* Update requirements.txt

* unit tests for lance

superagi/vector_store/lancedb.py

@@ -87,7 +87,7 @@ def get_matching_text(self, query: str, top_k: int = 5, **kwargs: Any) -> List[D
         Returns:
             The list of documents most similar to the query
         """
-        namespace = kwargs.get("namespace", self.namespace)
+        namespace = kwargs.get("namespace", "None")
 
         for table in self.db.table_names():
             if table == namespace:
@@ -96,17 +96,18 @@ def get_matching_text(self, query: str, top_k: int = 5, **kwargs: Any) -> List[D
         try:
             tbl
         except:
-            raise ValueError(namespace + " Table not found in LanceDB.")
+            raise ValueError(namespace + " Table not found in LanceDB. Please call this function with a valid table name.")
 
         embed_text = self.embedding_model.get_embedding(query)
         res = tbl.search(embed_text).limit(top_k).to_df()
+        print(res)
 
         documents = []
 
         for i in range(len(res)):
             meta = {}
             for col in res:
-                if col != 'vector' and col != 'id':
+                if col != 'vector' and col != 'id' and col != 'score':
                     meta[col] = res[col][i]
 
             documents.append(

tests/integration_tests/vector_store/test_lancedb.py

@@ -0,0 +1,106 @@
+import numpy as np
+import shutil
+import pytest
+
+import lancedb
+from superagi.vector_store.lancedb import LanceDB
+from superagi.vector_store.document import Document
+from superagi.vector_store.embedding.openai import OpenAiEmbedding
+
+
+@pytest.fixture
+def client():
+    db = lancedb.connect(".test_lancedb")
+    yield db
+    shutil.rmtree(".test_lancedb")
+
+
+@pytest.fixture
+def mock_openai_embedding(monkeypatch):
+    monkeypatch.setattr(
+        OpenAiEmbedding,
+        "get_embedding",
+        lambda self, text: np.random.random(3).tolist(),
+    )
+
+
+@pytest.fixture
+def store(client, mock_openai_embedding):
+    yield LanceDB(client, OpenAiEmbedding(api_key="test_api_key"), "text")
+
+
+@pytest.fixture
+def dataset():
+    book_titles = [
+        "The Great Gatsby",
+        "To Kill a Mockingbird",
+        "1984",
+        "Pride and Prejudice",
+        "The Catcher in the Rye",
+    ]
+
+    documents = []
+    for i, title in enumerate(book_titles):
+        author = f"Author {i}"
+        description = f"A summary of {title}"
+        text_content = f"This is the text for {title}"
+        metadata = {"author": author, "description": description}
+        document = Document(text_content=text_content, metadata=metadata)
+
+        documents.append(document)
+
+    return documents
+
+
+@pytest.fixture
+def dataset_no_metadata():
+    book_titles = [
+        "The Lord of the Rings",
+        "The Hobbit",
+        "The Chronicles of Narnia",
+    ]
+
+    documents = []
+    for title in book_titles:
+        text_content = f"This is the text for {title}"
+        document = Document(text_content=text_content)
+        documents.append(document)
+
+    return documents
+
+
+@pytest.mark.parametrize(
+    "data, results, table_name",
+    [
+        ("dataset", (5, 2), "test_table"),
+        ("dataset_no_metadata", (3, 0), "test_table_no_metadata"),
+    ],
+)
+def test_add_texts(store, client, data, results, table_name, request):
+    dataset = request.getfixturevalue(data)
+    count, meta_count = results
+    ids = store.add_documents(dataset, table_name=table_name)
+    assert len(ids) == count
+
+    tbl = client.open_table(table_name)
+    assert len(tbl.to_pandas().columns) - 3 == meta_count
+    # Subtracting 3 because of the id, vector, and text columns. The rest
+    # should be metadata columns.
+
+
+@pytest.mark.parametrize(
+    "data, search_text, table_name, index",
+    [
+        ("dataset", "The Great Gatsby", "test_table", 0),
+        ("dataset", "1984", "test_table2", 2),
+        ("dataset_no_metadata", "The Hobbit", "test_table_no_metadata", 1),
+    ],
+)
+def test_get_matching_text(store, data, search_text, table_name, index, request):
+    print("SEARCHING FOR " + search_text)
+    dataset = request.getfixturevalue(data)
+    store.add_documents(dataset, table_name=table_name)
+    results = store.get_matching_text(search_text, top_k=2, namespace=table_name)
+    print(results[0])
+    assert len(results) == 2
+    assert results[0] == dataset[index]