diff --git a/docs/embeddings/python-api.md b/docs/embeddings/python-api.md
index 67d49e32..1f9e8c9f 100644
--- a/docs/embeddings/python-api.md
+++ b/docs/embeddings/python-api.md
@@ -30,13 +30,17 @@ To work with embeddings in this way you will need an instance of a [sqlite-utils
 import sqlite_utils
 import llm
 
+# This collection will use an in-memory database that will be
+# discarded when the Python process exits
+collection = llm.Collection("entries", model_id="ada-002")
+
+# Or you can persist the database to disk like this:
 db = sqlite_utils.Database("my-embeddings.db")
-# Pass model_id= to specify a model for the collection
-collection = llm.Collection(db, "entries", model_id="ada-002")
+collection = llm.Collection("entries", db, model_id="ada-002")
 
-# Or you can pass a model directly using model=
+# You can pass a model directly using model= instead of model_id=
 embedding_model = llm.get_embedding_model("ada-002")
-collection = llm.Collection(db, "entries", model=embedding_model)
+collection = llm.Collection("entries", db, model=embedding_model)
 ```
 If the collection already exists in the database you can omit the `model` or `model_id` argument - the model ID will be read from the `collections` table.
 
diff --git a/llm/cli.py b/llm/cli.py
index 05c6c3e4..298671b3 100644
--- a/llm/cli.py
+++ b/llm/cli.py
@@ -910,7 +910,7 @@ def get_db():
     model_obj = None
     if collection:
         db = get_db()
-        collection_obj = Collection(db, collection, model_id=model)
+        collection_obj = Collection(collection, db, model_id=model)
         model_obj = collection_obj.model()
 
     if model_obj is None:
@@ -995,7 +995,7 @@ def similar(collection, id, input, content, number, database):
         raise click.ClickException("No embeddings table found in database")
 
     try:
-        collection_obj = Collection(db, collection, create=False)
+        collection_obj = Collection(collection, db, create=False)
     except Collection.DoesNotExist:
         raise click.ClickException("Collection does not exist")
 
diff --git a/llm/embeddings.py b/llm/embeddings.py
index 997a61c6..2ea8e1c8 100644
--- a/llm/embeddings.py
+++ b/llm/embeddings.py
@@ -24,8 +24,8 @@ class DoesNotExist(Exception):
 
     def __init__(
         self,
-        db: Database,
         name: str,
+        db: Optional[Database] = None,
         *,
         model: Optional[EmbeddingModel] = None,
         model_id: Optional[str] = None,
@@ -48,7 +48,7 @@ def __init__(
         """
         import llm
 
-        self.db = db
+        self.db = db or Database(memory=True)
         self.name = name
         self._model = model
 
diff --git a/tests/conftest.py b/tests/conftest.py
index 0b7dc504..1b8443ec 100644
--- a/tests/conftest.py
+++ b/tests/conftest.py
@@ -21,7 +21,7 @@ def user_path(tmpdir):
 def user_path_with_embeddings(user_path):
     path = str(user_path / "embeddings.db")
     db = sqlite_utils.Database(path)
-    collection = llm.Collection(db, "demo", model_id="embed-demo")
+    collection = llm.Collection("demo", db, model_id="embed-demo")
     collection.embed("1", "hello world")
     collection.embed("2", "goodbye world")
 
diff --git a/tests/test_embed.py b/tests/test_embed.py
index 19b60f6e..d7c34423 100644
--- a/tests/test_embed.py
+++ b/tests/test_embed.py
@@ -7,8 +7,7 @@
 
 @pytest.fixture
 def collection():
-    db = sqlite_utils.Database(memory=True)
-    collection = llm.Collection(db, "test", model_id="embed-demo")
+    collection = llm.Collection("test", model_id="embed-demo")
     collection.embed(1, "hello world")
     collection.embed(2, "goodbye world")
     return collection
@@ -95,7 +94,7 @@ def test_similar_by_id(collection):
 @pytest.mark.parametrize("with_metadata", (False, True))
 def test_embed_multi(with_metadata):
     db = sqlite_utils.Database(memory=True)
-    collection = llm.Collection(db, "test", model_id="embed-demo")
+    collection = llm.Collection("test", db, model_id="embed-demo")
     ids_and_texts = ((str(i), "hello {}".format(i)) for i in range(1000))
     if with_metadata:
         ids_and_texts = ((id, text, {"meta": id}) for id, text in ids_and_texts)