langchain[minor]: allow CacheBackedEmbeddings to cache queries (#20073)

Add optional caching of queries to cache backed embeddings

docs/docs/how_to/caching_embeddings.ipynb

@@ -18,11 +18,12 @@
     "- document_embedding_cache: Any [`ByteStore`](/docs/integrations/stores/) for caching document embeddings.\n",
     "- batch_size: (optional, defaults to `None`) The number of documents to embed between store updates.\n",
     "- namespace: (optional, defaults to `\"\"`) The namespace to use for document cache. This namespace is used to avoid collisions with other caches. For example, set it to the name of the embedding model used.\n",
+    "- query_embedding_cache: (optional, defaults to `None` or not caching) A [`ByteStore`](/docs/integrations/stores/) for caching query embeddings, or `True` to use the same store as `document_embedding_cache`.\n",
     "\n",
     "**Attention**:\n",
     "\n",
     "- Be sure to set the `namespace` parameter to avoid collisions of the same text embedded using different embeddings models.\n",
-    "- Currently `CacheBackedEmbeddings` does not cache embedding created with `embed_query()` `aembed_query()` methods."
+    "- `CacheBackedEmbeddings` does not cache query embeddings by default. To enable query caching, one need to specify a `query_embedding_cache`."
    ]
   },
   {

libs/langchain/langchain/embeddings/cache.py

@@ -6,6 +6,7 @@ embeddings for the same text.
 
 The text is hashed and the hash is used as the key in the cache.
 """
+
 from __future__ import annotations
 
 import hashlib
@@ -59,6 +60,9 @@ class CacheBackedEmbeddings(Embeddings):
     If need be, the interface can be extended to accept other implementations
     of the value serializer and deserializer, as well as the key encoder.
 
+    Note that by default only document embeddings are cached. To cache query
+    embeddings too, pass in a query_embedding_store to constructor.
+
     Examples:
 
         .. code-block: python
@@ -87,6 +91,7 @@ class CacheBackedEmbeddings(Embeddings):
         document_embedding_store: BaseStore[str, List[float]],
         *,
         batch_size: Optional[int] = None,
+        query_embedding_store: Optional[BaseStore[str, List[float]]] = None,
     ) -> None:
         """Initialize the embedder.
 
@@ -94,9 +99,12 @@ class CacheBackedEmbeddings(Embeddings):
             underlying_embeddings: the embedder to use for computing embeddings.
             document_embedding_store: The store to use for caching document embeddings.
             batch_size: The number of documents to embed between store updates.
+            query_embedding_store: The store to use for caching query embeddings.
+                If None, query embeddings are not cached.
         """
         super().__init__()
         self.document_embedding_store = document_embedding_store
+        self.query_embedding_store = query_embedding_store
         self.underlying_embeddings = underlying_embeddings
         self.batch_size = batch_size
 
@@ -173,14 +181,8 @@ class CacheBackedEmbeddings(Embeddings):
     def embed_query(self, text: str) -> List[float]:
         """Embed query text.
 
-        This method does not support caching at the moment.
-
-        Support for caching queries is easily to implement, but might make
-        sense to hold off to see the most common patterns.
-
-        If the cache has an eviction policy, we may need to be a bit more careful
-        about sharing the cache between documents and queries. Generally,
-        one is OK evicting query caches, but document caches should be kept.
+        By default, this method does not cache queries. To enable caching, set the
+        `cache_query` parameter to `True` when initializing the embedder.
 
         Args:
             text: The text to embed.
@@ -188,19 +190,22 @@ class CacheBackedEmbeddings(Embeddings):
         Returns:
             The embedding for the given text.
         """
+        if not self.query_embedding_store:
             return self.underlying_embeddings.embed_query(text)
 
+        (cached,) = self.query_embedding_store.mget([text])
+        if cached is not None:
+            return cached
+
+        vector = self.underlying_embeddings.embed_query(text)
+        self.query_embedding_store.mset([(text, vector)])
+        return vector
+
     async def aembed_query(self, text: str) -> List[float]:
         """Embed query text.
 
-        This method does not support caching at the moment.
-
-        Support for caching queries is easily to implement, but might make
-        sense to hold off to see the most common patterns.
-
-        If the cache has an eviction policy, we may need to be a bit more careful
-        about sharing the cache between documents and queries. Generally,
-        one is OK evicting query caches, but document caches should be kept.
+        By default, this method does not cache queries. To enable caching, set the
+        `cache_query` parameter to `True` when initializing the embedder.
 
         Args:
             text: The text to embed.
@@ -208,8 +213,17 @@ class CacheBackedEmbeddings(Embeddings):
         Returns:
             The embedding for the given text.
         """
+        if not self.query_embedding_store:
             return await self.underlying_embeddings.aembed_query(text)
 
+        (cached,) = await self.query_embedding_store.amget([text])
+        if cached is not None:
+            return cached
+
+        vector = await self.underlying_embeddings.aembed_query(text)
+        await self.query_embedding_store.amset([(text, vector)])
+        return vector
+
     @classmethod
     def from_bytes_store(
         cls,
@@ -218,6 +232,7 @@ class CacheBackedEmbeddings(Embeddings):
         *,
         namespace: str = "",
         batch_size: Optional[int] = None,
+        query_embedding_cache: Union[bool, ByteStore] = False,
     ) -> CacheBackedEmbeddings:
         """On-ramp that adds the necessary serialization and encoding to the store.
 
@@ -229,13 +244,33 @@ class CacheBackedEmbeddings(Embeddings):
                        This namespace is used to avoid collisions with other caches.
                        For example, set it to the name of the embedding model used.
             batch_size: The number of documents to embed between store updates.
+            query_embedding_cache: The cache to use for storing query embeddings.
+                True to use the same cache as document embeddings.
+                False to not cache query embeddings.
         """
         namespace = namespace
         key_encoder = _create_key_encoder(namespace)
-        encoder_backed_store = EncoderBackedStore[str, List[float]](
+        document_embedding_store = EncoderBackedStore[str, List[float]](
             document_embedding_cache,
             key_encoder,
             _value_serializer,
             _value_deserializer,
         )
-        return cls(underlying_embeddings, encoder_backed_store, batch_size=batch_size)
+        if query_embedding_cache is True:
+            query_embedding_store = document_embedding_store
+        elif query_embedding_cache is False:
+            query_embedding_store = None
+        else:
+            query_embedding_store = EncoderBackedStore[str, List[float]](
+                query_embedding_cache,
+                key_encoder,
+                _value_serializer,
+                _value_deserializer,
+            )
+
+        return cls(
+            underlying_embeddings,
+            document_embedding_store,
+            batch_size=batch_size,
+            query_embedding_store=query_embedding_store,
+        )

libs/langchain/tests/unit_tests/embeddings/test_caching.py

@@ -43,6 +43,20 @@ def cache_embeddings_batch() -> CacheBackedEmbeddings:
     )
 
 
+@pytest.fixture
+def cache_embeddings_with_query() -> CacheBackedEmbeddings:
+    """Create a cache backed embeddings with query caching."""
+    doc_store = InMemoryStore()
+    query_store = InMemoryStore()
+    embeddings = MockEmbeddings()
+    return CacheBackedEmbeddings.from_bytes_store(
+        embeddings,
+        document_embedding_cache=doc_store,
+        namespace="test_namespace",
+        query_embedding_cache=query_store,
+    )
+
+
 def test_embed_documents(cache_embeddings: CacheBackedEmbeddings) -> None:
     texts = ["1", "22", "a", "333"]
     vectors = cache_embeddings.embed_documents(texts)
@@ -73,6 +87,17 @@ def test_embed_query(cache_embeddings: CacheBackedEmbeddings) -> None:
     vector = cache_embeddings.embed_query(text)
     expected_vector = [5.0, 6.0]
     assert vector == expected_vector
+    assert cache_embeddings.query_embedding_store is None
+
+
+def test_embed_cached_query(cache_embeddings_with_query: CacheBackedEmbeddings) -> None:
+    text = "query_text"
+    vector = cache_embeddings_with_query.embed_query(text)
+    expected_vector = [5.0, 6.0]
+    assert vector == expected_vector
+    keys = list(cache_embeddings_with_query.query_embedding_store.yield_keys())  # type: ignore[union-attr]
+    assert len(keys) == 1
+    assert keys[0] == "test_namespace89ec3dae-a4d9-5636-a62e-ff3b56cdfa15"
 
 
 async def test_aembed_documents(cache_embeddings: CacheBackedEmbeddings) -> None:
@@ -112,3 +137,13 @@ async def test_aembed_query(cache_embeddings: CacheBackedEmbeddings) -> None:
     vector = await cache_embeddings.aembed_query(text)
     expected_vector = [5.0, 6.0]
     assert vector == expected_vector
+
+
+async def test_aembed_query_cached(
+    cache_embeddings_with_query: CacheBackedEmbeddings,
+) -> None:
+    text = "query_text"
+    await cache_embeddings_with_query.aembed_query(text)
+    keys = list(cache_embeddings_with_query.query_embedding_store.yield_keys())  # type: ignore[union-attr]
+    assert len(keys) == 1
+    assert keys[0] == "test_namespace89ec3dae-a4d9-5636-a62e-ff3b56cdfa15"