unleash the caching on the subclass vectorizers

Author: tylerhutcherson

poetry.lock

@@ -6,7 +6,5 @@
 """
 
 from redisvl.extensions.cache.base import BaseCache
-from redisvl.extensions.cache.embeddings import EmbeddingsCache
-from redisvl.extensions.cache.llm import SemanticCache
 
-__all__ = ["BaseCache", "EmbeddingsCache", "SemanticCache"]
+__all__ = ["BaseCache"]

redisvl/extensions/cache/__init__.py

@@ -9,6 +9,8 @@
 from redis import Redis
 from redis.asyncio import Redis as AsyncRedis
 
+from redisvl.redis.connection import RedisConnectionFactory
+
 
 class BaseCache:
     """Base abstract cache interface for all RedisVL caches.
@@ -121,10 +123,15 @@ async def _get_async_redis_client(self) -> AsyncRedis:
             AsyncRedis: An async Redis client instance.
         """
         if not hasattr(self, "_async_redis_client") or self._async_redis_client is None:
-            # Create new async Redis client
-            url = self.redis_kwargs["redis_url"]
-            kwargs = self.redis_kwargs["connection_kwargs"]
-            self._async_redis_client = AsyncRedis.from_url(url, **kwargs)  # type: ignore
+            client = self.redis_kwargs.get("redis_client")
+            if isinstance(client, Redis):
+                self._async_redis_client = RedisConnectionFactory.sync_to_async_redis(
+                    client
+                )
+            else:
+                url = self.redis_kwargs["redis_url"]
+                kwargs = self.redis_kwargs["connection_kwargs"]
+                self._async_redis_client = RedisConnectionFactory.get_async_redis_connection(url, **kwargs)  # type: ignore
         return self._async_redis_client
 
     def expire(self, key: str, ttl: Optional[int] = None) -> None:

redisvl/extensions/cache/base.py

@@ -22,7 +22,6 @@
 from redisvl.index import AsyncSearchIndex, SearchIndex
 from redisvl.query import VectorRangeQuery
 from redisvl.query.filter import FilterExpression
-from redisvl.redis.connection import RedisConnectionFactory
 from redisvl.redis.utils import hashify
 from redisvl.utils.log import get_logger
 from redisvl.utils.utils import (
@@ -31,7 +30,7 @@
     serialize,
     validate_vector_dims,
 )
-from redisvl.utils.vectorize import BaseVectorizer, HFTextVectorizer
+from redisvl.utils.vectorize.base import BaseVectorizer
 
 logger = get_logger("[RedisVL]")
 
@@ -105,6 +104,8 @@ def __init__(
                 )
             self._vectorizer = vectorizer
         else:
+            from redisvl.utils.vectorize.text.huggingface import HFTextVectorizer
+
             # Create a default vectorizer
             vectorizer_kwargs = kwargs
             if dtype:
@@ -183,11 +184,8 @@ def _modify_schema(
     async def _get_async_index(self) -> AsyncSearchIndex:
         """Lazily construct the async search index class."""
         # Construct async index if necessary
-        async_client = None
         if self._aindex is None:
-            client = self.redis_kwargs.get("redis_client")
-            if isinstance(client, Redis):
-                async_client = RedisConnectionFactory.sync_to_async_redis(client)
+            async_client = await self._get_async_redis_client()
             self._aindex = AsyncSearchIndex(
                 schema=self._index.schema,
                 redis_client=async_client,

redisvl/extensions/cache/llm/semantic.py

@@ -62,7 +62,7 @@ class RoutingConfig(BaseModel):
     """Configuration for routing behavior."""
 
     """The maximum number of top matches to return."""
-    max_k: Annotated[int, Field(strict=True, default=1, gt=0)] = 1
+    max_k: Annotated[int, Field(strict=True, gt=0)] = 1
     """Aggregation method to use to classify queries."""
     aggregation_method: DistanceAggregationMethod = Field(
         default=DistanceAggregationMethod.avg

redisvl/extensions/router/schema.py

@@ -21,11 +21,8 @@
 from redisvl.redis.utils import convert_bytes, hashify, make_dict
 from redisvl.utils.log import get_logger
 from redisvl.utils.utils import deprecated_argument, model_to_dict
-from redisvl.utils.vectorize import (
-    BaseVectorizer,
-    HFTextVectorizer,
-    vectorizer_from_dict,
-)
+from redisvl.utils.vectorize.base import BaseVectorizer
+from redisvl.utils.vectorize.text.huggingface import HFTextVectorizer
 
 logger = get_logger(__name__)
 
@@ -483,6 +480,8 @@ def from_dict(
             }
             router = SemanticRouter.from_dict(router_data)
         """
+        from redisvl.utils.vectorize import vectorizer_from_dict
+
         try:
             name = data["name"]
             routes_data = data["routes"]

redisvl/extensions/router/semantic.py

@@ -20,7 +20,7 @@
 from redisvl.query import FilterQuery, RangeQuery
 from redisvl.query.filter import Tag
 from redisvl.utils.utils import deprecated_argument, validate_vector_dims
-from redisvl.utils.vectorize import BaseVectorizer, HFTextVectorizer
+from redisvl.utils.vectorize.base import BaseVectorizer
 
 
 class SemanticSessionManager(BaseSessionManager):
@@ -82,6 +82,8 @@ def __init__(
                     f"Provided dtype {dtype} does not match vectorizer dtype {vectorizer.dtype}"
                 )
         else:
+            from redisvl.utils.vectorize.text.huggingface import HFTextVectorizer
+
             vectorizer_kwargs = kwargs
 
             if dtype: