Use new lazy_import function

Author: abrookins

redisvl/cli/index.py

@@ -2,14 +2,13 @@
 import sys
 from argparse import Namespace
 
-from tabulate import tabulate
-
 from redisvl.cli.utils import add_index_parsing_options, create_redis_url
 from redisvl.index import SearchIndex
 from redisvl.redis.connection import RedisConnectionFactory
 from redisvl.redis.utils import convert_bytes, make_dict
 from redisvl.schema.schema import IndexSchema
 from redisvl.utils.log import get_logger
+from redisvl.utils.utils import lazy_import
 
 logger = get_logger("[RedisVL]")
 
@@ -127,6 +126,8 @@ def _connect_to_index(self, args: Namespace) -> SearchIndex:
 
 
 def _display_in_table(index_info, output_format="rounded_outline"):
+    tabulate = lazy_import("tabulate")
+
     print("\n")
     attributes = index_info.get("attributes", [])
     definition = make_dict(index_info.get("index_definition"))

redisvl/cli/stats.py

@@ -2,12 +2,11 @@
 import sys
 from argparse import Namespace
 
-from tabulate import tabulate
-
 from redisvl.cli.utils import add_index_parsing_options, create_redis_url
 from redisvl.index import SearchIndex
 from redisvl.schema.schema import IndexSchema
 from redisvl.utils.log import get_logger
+from redisvl.utils.utils import lazy_import
 
 logger = get_logger("[RedisVL]")
 
@@ -87,6 +86,8 @@ def _connect_to_index(self, args: Namespace) -> SearchIndex:
 
 
 def _display_stats(index_info, output_format="rounded_outline"):
+    tabulate = lazy_import("tabulate")
+
     # Extracting the statistics
     stats_data = [(key, str(index_info.get(key))) for key in STATS_KEYS]
 

redisvl/query/aggregate.py

@@ -164,8 +164,10 @@ def _set_stopwords(self, stopwords: Optional[Union[str, Set[str]]] = "english"):
         elif isinstance(stopwords, str):
             # Lazy import because nltk is an optional dependency
             try:
-                import nltk
-                from nltk.corpus import stopwords as nltk_stopwords
+                from redisvl.utils.utils import lazy_import
+
+                nltk = lazy_import("nltk")
+                nltk_stopwords = lazy_import("nltk.corpus.stopwords")
             except ImportError:
                 raise ValueError(
                     f"Loading stopwords for {stopwords} failed: nltk is not installed."

redisvl/query/query.py

@@ -895,8 +895,10 @@ def _set_stopwords(self, stopwords: Optional[Union[str, Set[str]]] = "english"):
         elif isinstance(stopwords, str):
             # Lazy import because nltk is an optional dependency
             try:
-                import nltk
-                from nltk.corpus import stopwords as nltk_stopwords
+                from redisvl.utils.utils import lazy_import
+
+                nltk = lazy_import("nltk")
+                nltk_stopwords = lazy_import("nltk.corpus.stopwords")
             except ImportError:
                 raise ValueError(
                     f"Loading stopwords for {stopwords} failed: nltk is not installed."

redisvl/redis/utils.py

@@ -1,8 +1,12 @@
 import hashlib
 from typing import Any, Dict, List, Optional
 
-import numpy as np
-from ml_dtypes import bfloat16
+from redisvl.utils.utils import lazy_import
+
+# Lazy import numpy
+np = lazy_import("numpy")
+# Lazy import ml_dtypes
+bfloat16 = lazy_import("ml_dtypes.bfloat16")
 
 from redisvl.schema.fields import VectorDataType
 
@@ -41,6 +45,14 @@ def array_to_buffer(array: List[float], dtype: str) -> bytes:
         raise ValueError(
             f"Invalid data type: {dtype}. Supported types are: {[t.lower() for t in VectorDataType]}"
         )
+
+    # Special handling for bfloat16 which requires explicit import from ml_dtypes
+    if dtype.lower() == "bfloat16":
+        # Import ml_dtypes.bfloat16 directly to ensure it's registered with numpy
+        from ml_dtypes import bfloat16 as bf16
+
+        return np.array(array, dtype=bf16).tobytes()
+
     return np.array(array, dtype=dtype.lower()).tobytes()
 
 
@@ -52,6 +64,14 @@ def buffer_to_array(buffer: bytes, dtype: str) -> List[Any]:
         raise ValueError(
             f"Invalid data type: {dtype}. Supported types are: {[t.lower() for t in VectorDataType]}"
         )
+
+    # Special handling for bfloat16 which requires explicit import from ml_dtypes
+    if dtype.lower() == "bfloat16":
+        # Import ml_dtypes.bfloat16 directly to ensure it's registered with numpy
+        from ml_dtypes import bfloat16 as bf16
+
+        return np.frombuffer(buffer, dtype=bf16).tolist()  # type: ignore[return-value]
+
     return np.frombuffer(buffer, dtype=dtype.lower()).tolist()  # type: ignore[return-value]
 
 

redisvl/utils/optimize/cache.py

@@ -1,6 +1,8 @@
 from typing import Any, Callable, Dict, List
 
-import numpy as np
+from redisvl.utils.utils import lazy_import
+
+np = lazy_import("numpy")
 from ranx import Qrels, Run, evaluate
 
 from redisvl.extensions.cache.llm.semantic import SemanticCache

redisvl/utils/optimize/router.py

@@ -1,7 +1,9 @@
 import random
 from typing import Any, Callable, Dict, List
 
-import numpy as np
+from redisvl.utils.utils import lazy_import
+
+np = lazy_import("numpy")
 from ranx import Qrels, Run, evaluate
 
 from redisvl.extensions.router.semantic import SemanticRouter

redisvl/utils/optimize/utils.py

@@ -1,6 +1,8 @@
 from typing import List
 
-import numpy as np
+from redisvl.utils.utils import lazy_import
+
+np = lazy_import("numpy")
 from ranx import Qrels
 
 from redisvl.utils.optimize.schema import LabeledData

tests/unit/test_utils.py

@@ -165,7 +165,15 @@ def test_empty_list_to_bytes():
 def test_conversion_with_various_dtypes(dtype):
     """Test conversion of a list of floats to bytes with various dtypes"""
     array = [1.0, -2.0, 3.5]
-    expected = np.array(array, dtype=dtype).tobytes()
+
+    # Special handling for bfloat16 which requires explicit import from ml_dtypes
+    if dtype == "bfloat16":
+        from ml_dtypes import bfloat16 as bf16
+
+        expected = np.array(array, dtype=bf16).tobytes()
+    else:
+        expected = np.array(array, dtype=dtype).tobytes()
+
     assert array_to_buffer(array, dtype=dtype) == expected