feat: GoogleGenAIVectorizer — migrate Google embeddings to the supported google-genai SDK

.github/workflows/test.yml

@@ -37,6 +37,7 @@ jobs:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
           GCP_LOCATION: ${{ secrets.GCP_LOCATION }}
           GCP_PROJECT_ID: ${{ secrets.GCP_PROJECT_ID }}
+          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
           COHERE_API_KEY: ${{ secrets.COHERE_API_KEY }}
           MISTRAL_API_KEY: ${{ secrets.MISTRAL_API_KEY }}
           VOYAGE_API_KEY: ${{ secrets.VOYAGE_API_KEY }}

docs/api/vectorizer.rst

@@ -8,7 +8,8 @@ Vectorizers
    compatibility:
 
    - ``VoyageAITextVectorizer`` → Use ``VoyageAIVectorizer`` instead
-   - ``VertexAITextVectorizer`` → Use ``VertexAIVectorizer`` instead
+   - ``VertexAITextVectorizer`` → Use ``GoogleGenAIVectorizer`` instead
+     (``VertexAIVectorizer`` is itself deprecated; see below)
    - ``BedrockTextVectorizer`` → Use ``BedrockVectorizer`` instead
    - ``CustomTextVectorizer`` → Use ``CustomVectorizer`` instead
 
@@ -59,15 +60,31 @@ VertexAIVectorizer
 .. currentmodule:: redisvl.utils.vectorize.vertexai
 
 .. note::
-    For backwards compatibility, an alias ``VertexAITextVectorizer`` is available
-    in the ``redisvl.utils.vectorize.text`` module. This alias is deprecated
-    as of version 0.13.0 and will be removed in a future major release.
+    ``VertexAIVectorizer`` is **deprecated**. It uses Google's Vertex AI
+    model-garden SDK, which Google has deprecated with a scheduled removal. Use
+    :class:`~redisvl.utils.vectorize.googlegenai.GoogleGenAIVectorizer` for text
+    embeddings on the supported ``google-genai`` SDK. The alias
+    ``VertexAITextVectorizer`` (in ``redisvl.utils.vectorize.text``) is likewise
+    deprecated. Multimodal (image/video) migration is tracked in
+    `issue #620 <https://github.com/redis/redis-vl-python/issues/620>`_.
 
 .. autoclass:: VertexAIVectorizer
    :show-inheritance:
    :members:
 
 
+GoogleGenAIVectorizer
+=====================
+
+.. _googlegenaivectorizer_api:
+
+.. currentmodule:: redisvl.utils.vectorize.googlegenai
+
+.. autoclass:: GoogleGenAIVectorizer
+   :show-inheritance:
+   :members:
+
+
 CohereTextVectorizer
 ====================
 

docs/concepts/utilities.md

@@ -35,7 +35,7 @@ Vectorizers handle batching internally, breaking large batches into provider-app
 
 ### Supported Providers
 
-RedisVL includes vectorizers for OpenAI, Azure OpenAI, Cohere, HuggingFace (local), Mistral, Google Vertex AI, AWS Bedrock, VoyageAI, and others. See the {doc}`/api/vectorizer` for the complete list. You can also create custom vectorizers that wrap any embedding function.
+RedisVL includes vectorizers for OpenAI, Azure OpenAI, Cohere, HuggingFace (local), Mistral, Google (via `google-genai`, covering both Vertex AI and the Gemini Developer API), AWS Bedrock, VoyageAI, and others. See the {doc}`/api/vectorizer` for the complete list. You can also create custom vectorizers that wrap any embedding function.
 
 ## Rerankers
 

docs/user_guide/04_vectorizers.ipynb

@@ -6,7 +6,7 @@
       "source": [
         "# Create Embeddings with Vectorizers\n",
         "\n",
-        "This guide demonstrates how to create embeddings using RedisVL's built-in text vectorizers. RedisVL supports multiple embedding providers: OpenAI, HuggingFace, Ollama, Vertex AI, Cohere, Mistral AI, Amazon Bedrock, VoyageAI, and custom vectorizers.\n",
+        "This guide demonstrates how to create embeddings using RedisVL's built-in text vectorizers. RedisVL supports multiple embedding providers: OpenAI, HuggingFace, Ollama, Google (Vertex AI and the Gemini Developer API), Cohere, Mistral AI, Amazon Bedrock, VoyageAI, and custom vectorizers.\n",
         "\n",
         "## Prerequisites\n",
         "\n",
@@ -360,6 +360,8 @@
       "cell_type": "markdown",
       "metadata": {},
       "source": [
+        "> **⚠️ Deprecated.** `VertexAIVectorizer` uses Google's Vertex AI model-garden SDK, which Google has deprecated with a scheduled removal. For text embeddings use the **Google Gen AI** vectorizer shown in the next section (`GoogleGenAIVectorizer`) on the supported `google-genai` SDK. Multimodal (image/video) migration is tracked in [issue #620](https://github.com/redis/redis-vl-python/issues/620).\n",
+        "\n",
         "### VertexAI\n",
         "\n",
         "[VertexAI](https://cloud.google.com/vertex-ai/docs/generative-ai/embeddings/get-text-embeddings) is GCP's fully-featured AI platform including a number of pretrained LLMs. RedisVL supports using VertexAI to create embeddings from these models. To use VertexAI, you will first need to install the ``google-cloud-aiplatform`` library.\n",
@@ -402,6 +404,64 @@
         "test[:10]"
       ]
     },
+    {
+      "cell_type": "markdown",
+      "metadata": {},
+      "source": [
+        "### Google Gen AI\n",
+        "\n",
+        "The `GoogleGenAIVectorizer` uses Google's [`google-genai`](https://pypi.org/project/google-genai/) SDK — the supported replacement for the deprecated Vertex AI model-garden SDK. It reaches **both** Google embedding backends from a single client:\n",
+        "\n",
+        "- **Vertex AI / Gemini Enterprise** — GCP project auth (`project_id` + `location`, credentials via ADC).\n",
+        "- **Gemini Developer API** — a single `api_key`.\n",
+        "\n",
+        "Install it with `pip install redisvl[google-genai]`.\n",
+        "\n",
+        "**Migrating from `VertexAIVectorizer`:**\n",
+        "\n",
+        "| | `VertexAIVectorizer` (deprecated) | `GoogleGenAIVectorizer` |\n",
+        "|---|---|---|\n",
+        "| SDK | `google-cloud-aiplatform` (deprecated) | `google-genai` (supported) |\n",
+        "| Default model | `textembedding-gecko` (768 dims) | `gemini-embedding-001` (3072 dims) |\n",
+        "| Backends | Vertex AI only | Vertex AI **and** Gemini Developer API |\n",
+        "| Async | no | yes (`aembed`, `aembed_many`) |\n",
+        "\n",
+        "Because the default model and dimensions differ, embeddings are **not** interchangeable — reindex when you switch. When you request a reduced `output_dimensionality`, the vectorizer L2-normalizes the result so it stays valid for Redis COSINE / inner-product search.\n",
+        "\n",
+        "**Set one of the following:**\n",
+        "\n",
+        "```\n",
+        "# Gemini Developer API\n",
+        "GEMINI_API_KEY=<your gemini api key>\n",
+        "\n",
+        "# or Vertex AI\n",
+        "GOOGLE_CLOUD_PROJECT=<your gcp project id>\n",
+        "GOOGLE_CLOUD_LOCATION=<your gcp region>\n",
+        "GOOGLE_APPLICATION_CREDENTIALS=<path to your gcp JSON creds>\n",
+        "```\n"
+      ]
+    },
+    {
+      "cell_type": "code",
+      "execution_count": null,
+      "metadata": {},
+      "outputs": [],
+      "source": [
+        "# NBVAL_SKIP  (docs example; not executed in CI so notebook validation makes no API calls)\n",
+        "from redisvl.utils.vectorize import GoogleGenAIVectorizer\n",
+        "\n",
+        "# Auto-detects the backend: GEMINI_API_KEY -> Gemini, else GCP project/location -> Vertex AI.\n",
+        "# Guarded so this notebook still runs when Google credentials aren't configured.\n",
+        "try:\n",
+        "    genai_vectorizer = GoogleGenAIVectorizer(model=\"gemini-embedding-001\")\n",
+        "    print(f\"backend={genai_vectorizer.backend}, dims={genai_vectorizer.dims}\")\n",
+        "    genai_test = genai_vectorizer.embed(\"This is a test sentence.\")\n",
+        "    print(genai_test[:10])\n",
+        "except (ImportError, ValueError) as e:\n",
+        "    print(f\"Skipping GoogleGenAIVectorizer demo: {e}\")\n",
+        "    genai_vectorizer = None\n"
+      ]
+    },
     {
       "cell_type": "markdown",
       "metadata": {},

docs/user_guide/installation.md

@@ -27,7 +27,8 @@ $ pip install redisvl[cohere]              # Cohere embeddings and reranking
 $ pip install redisvl[mistralai]           # Mistral AI embeddings
 $ pip install redisvl[voyageai]            # Voyage AI embeddings and reranking
 $ pip install redisvl[sentence-transformers]  # HuggingFace local embeddings
-$ pip install redisvl[vertexai]            # Google Vertex AI embeddings
+$ pip install redisvl[google-genai]        # Google embeddings (Vertex AI + Gemini API)
+$ pip install redisvl[vertexai]            # Google Vertex AI embeddings (legacy; deprecated, multimodal only)
 $ pip install redisvl[bedrock]             # AWS Bedrock embeddings
 
 # Other optional features

pyproject.toml

@@ -44,10 +44,19 @@ cohere = ["cohere>=4.44"]
 voyageai = ["voyageai>=0.2.2"]
 sentence-transformers = ["sentence-transformers>=3.4.0,<4"]
 langcache = ["langcache>=0.11.0"]
+# Legacy Google backend for the deprecated VertexAIVectorizer. The <2.0.0 cap is
+# intentional: Google's model-garden modules (vertexai.language_models /
+# vertexai.vision_models) that this vectorizer imports are deprecated with a
+# scheduled removal. They still ship in the 1.x line (latest 1.162.0); the 2.0.0
+# release that drops them is yanked on PyPI. Keep the cap so the deprecated class
+# (and its multimodal path, which has no google-genai replacement yet) keeps
+# importing. New text embeddings should use the `google-genai` extra below.
 vertexai = [
     "google-cloud-aiplatform>=1.26,<2.0.0",
     "protobuf>=5.28.0,<6.0.0",
 ]
+# Modern Google backend (Vertex AI + Gemini Developer API) for GoogleGenAIVectorizer.
+google-genai = ["google-genai>=1.0.0"]
 bedrock = [
     "boto3>=1.36.0,<2",
     "urllib3<2.2.0",
@@ -67,6 +76,7 @@ all = [
     "langcache>=0.11.0",
     "google-cloud-aiplatform>=1.26,<2.0.0",
     "protobuf>=5.28.0,<6.0.0",
+    "google-genai>=1.0.0",
     "boto3>=1.36.0,<2",
     "urllib3<2.2.0",
     "pillow>=11.3.0",

redisvl/utils/utils.py

@@ -203,7 +203,15 @@ def decorator(cls):
 
         @wraps(original_init)
         def new_init(self, *args, **kwargs):
-            warn(warning_message, category=DeprecationWarning, stacklevel=2)
+            # Emit only once per instance. When a deprecated subclass wraps a
+            # deprecated parent, both __init__ wrappers run via super().__init__;
+            # the sentinel keeps that to a single warning.
+            if not getattr(self, "_deprecation_warned", False):
+                warn(warning_message, category=DeprecationWarning, stacklevel=2)
+                try:
+                    object.__setattr__(self, "_deprecation_warned", True)
+                except Exception:
+                    pass
             original_init(self, *args, **kwargs)
 
         cls.__init__ = new_init

redisvl/utils/vectorize/__init__.py

@@ -2,6 +2,7 @@
 from redisvl.utils.vectorize.base import BaseVectorizer, Vectorizers
 from redisvl.utils.vectorize.bedrock import BedrockVectorizer
 from redisvl.utils.vectorize.custom import CustomVectorizer
+from redisvl.utils.vectorize.googlegenai import GoogleGenAIVectorizer
 from redisvl.utils.vectorize.text.azureopenai import AzureOpenAITextVectorizer
 from redisvl.utils.vectorize.text.bedrock import BedrockTextVectorizer
 from redisvl.utils.vectorize.text.cohere import CohereTextVectorizer
@@ -20,6 +21,7 @@
     "CohereTextVectorizer",
     "HFTextVectorizer",
     "OpenAITextVectorizer",
+    "GoogleGenAIVectorizer",
     "VertexAIVectorizer",
     "VertexAITextVectorizer",
     "AzureOpenAITextVectorizer",
@@ -61,6 +63,8 @@ def vectorizer_from_dict(
         return OllamaTextVectorizer(**args)
     elif vectorizer_type == Vectorizers.vertexai:
         return VertexAIVectorizer(**args)
+    elif vectorizer_type == Vectorizers.google_genai:
+        return GoogleGenAIVectorizer(**args)
     elif vectorizer_type == Vectorizers.voyageai:
         return VoyageAIVectorizer(**args)
     else:

redisvl/utils/vectorize/base.py

@@ -28,6 +28,7 @@ class Vectorizers(Enum):
     mistral = "mistral"
     ollama = "ollama"
     vertexai = "vertexai"
+    google_genai = "google_genai"
     hf = "hf"
     voyageai = "voyageai"