Migrate to Gen AI SDK: semantic_matching

notebooks/vertex_genai/labs/semantic_matching_with_gemini.ipynb

@@ -7,13 +7,13 @@
     "tags": []
    },
    "source": [
-    "# Semantic Matching with the Gemini API\n",
+    "# Semantic Matching with Gemini\n",
     "\n",
     "**Learning Objectives**\n",
     "  1. Learn how to identify matching product descriptions\n",
     "  1. Learn how to design a prompt for semantic matching\n",
     "  2. Learn how to evaluate performance of a prompt for semantic matching\n",
-    "  1. Learn how to query the Vertex Gemini API\n",
+    "  1. Learn how to use Gemini with Google Gen AI SDK\n",
     "  \n",
     "Semantic matching is the problem of classifying a pair of entities $(x_1, x_2)$ as being a good match or not. So it is a classification setup that is a very flexible: Namely, it comprises general information retrieval (where the first entity can be a textual query and the second entity can be a paragraph for instance), entity resolutions, or database-record fuzzy-matching. In this notebook we will focus on matching textual descriptions of retail products. More specifically:\n",
     "  \n",
@@ -46,12 +46,10 @@
     "import random\n",
     "\n",
     "import pandas as pd\n",
-    "import vertexai\n",
-    "from vertexai.generative_models import GenerationConfig, GenerativeModel\n",
+    "from google import genai\n",
+    "from google.genai.types import GenerateContentConfig\n",
     "\n",
-    "pd.options.display.max_colwidth = 1000\n",
-    "\n",
-    "vertexai.init()"
+    "pd.options.display.max_colwidth = 1000"
    ]
   },
   {
@@ -325,7 +323,7 @@
     "id": "kPOcktSQn1Q0"
    },
    "source": [
-    "We start by instanciating our client to the `gemini-1.0-pro` version of Gemini which is a large language model (LLM) developed by Google."
+    "We start by instanciating our client using the Gen AI SDK. We'll use the `gemini-2.0-flash-001` version of Gemini which is a large language model (LLM) developed by Google."
    ]
   },
   {
@@ -337,7 +335,9 @@
    },
    "outputs": [],
    "source": [
-    "generation_model = GenerativeModel(\"gemini-1.0-pro\")"
+    "MODEL = \"gemini-2.0-flash-001\"\n",
+    "\n",
+    "client = genai.Client(vertexai=True, location=\"us-central1\")"
    ]
   },
   {
@@ -355,7 +355,7 @@
    "source": [
     "### Exercise\n",
     "\n",
-    "Complete the function below so that it queries the Gemini API using the `generation_model` client.\n",
+    "Complete the function below so that it queries Gemini using the Gen AI SDK client.\n",
     "\n",
     "**Hint:** Jump to the cell after next to see how this function is used."
    ]
@@ -372,7 +372,7 @@
     "def are_products_matching(d1, d2, prompt):\n",
     "    prompt = None  # TODO: Substitute d1 and d2 in the parametrized prompt\n",
     "    answer = None  # TODO: Call the Gemini API with the prompt\n",
-    "    return answer.text"
+    "    return answer.text.strip()"
    ]
   },
   {
@@ -767,15 +767,15 @@
    "provenance": []
   },
   "environment": {
-   "kernel": "python3",
-   "name": "tf2-gpu.2-12.m123",
+   "kernel": "conda-base-py",
+   "name": "workbench-notebooks.m128",
    "type": "gcloud",
-   "uri": "us-docker.pkg.dev/deeplearning-platform-release/gcr.io/tf2-gpu.2-12:m123"
+   "uri": "us-docker.pkg.dev/deeplearning-platform-release/gcr.io/workbench-notebooks:m128"
   },
   "kernelspec": {
-   "display_name": "Python 3 (Local)",
+   "display_name": "Python 3 (ipykernel)",
    "language": "python",
-   "name": "python3"
+   "name": "conda-base-py"
   },
   "language_info": {
    "codemirror_mode": {
@@ -787,7 +787,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.10.14"
+   "version": "3.10.16"
   }
  },
  "nbformat": 4,