PaulFenton
diff --git a/‎cloud-sql/postgres/pgvector/notebooks/pgvector_gen_ai_demo.ipynb
Copy file name to clipboardExpand all lines: cloud-sql/postgres/pgvector/notebooks/pgvector_gen_ai_demo.ipynb
+122Lines changed: 122 additions & 0 deletions b/‎cloud-sql/postgres/pgvector/notebooks/pgvector_gen_ai_demo.ipynb
Copy file name to clipboardExpand all lines: cloud-sql/postgres/pgvector/notebooks/pgvector_gen_ai_demo.ipynb
+122Lines changed: 122 additions & 0 deletions
@@ -639,6 +639,128 @@
     "await main()  # type: ignore"
    ]
   },
+  {
+    "cell_type": "markdown",
+    "metadata": {
+      "id": "Ley1ZGjnG5Fr"
+    },
+    "source": [
+      "### Create indexes for faster similarity search in pgvector\n",
+      "\n",
+      "- Vector indexes can significantly speed up similarity search operation and avoid the brute-force exact nearest neighbor search that is used by default.\n",
+      "\n",
+      "- pgvector comes with two types of indexes (as of v0.5.1): `hnsw` and `ivfflat`.\n",
+      "\n",
+      "> 💡 Click [here](https://cloud.google.com/blog/products/databases/faster-similarity-search-performance-with-pgvector-indexes) to learn more about pgvector indexes.\n",
+      "\n",
+      "Enter or modify the values of index parameters for your index of choice and run the corresponding cell:"
+    ]
+  },
+  {
+    "cell_type": "code",
+    "source": [
+      "# @markdown Create an HNSW index on the `product_embeddings` table:\n",
+      "m =  24 # @param {type:\"integer\"}\n",
+      "ef_construction = 100  # @param {type:\"integer\"}\n",
+      "operator =  \"vector_cosine_ops\"  # @param [\"vector_cosine_ops\", \"vector_l2_ops\", \"vector_ip_ops\"]\n",
+      "\n",
+      "# Quick input validations.\n",
+      "assert m, \"⚠️ Please input a valid value for m.\"\n",
+      "assert ef_construction, \"⚠️ Please input a valid value for ef_construction.\"\n",
+      "assert operator, \"⚠️ Please input a valid value for operator.\"\n",
+      "\n",
+      "from pgvector.asyncpg import register_vector\n",
+      "import asyncio\n",
+      "import asyncpg\n",
+      "from google.cloud.sql.connector import Connector\n",
+      "\n",
+      "\n",
+      "async def main():\n",
+      "    loop = asyncio.get_running_loop()\n",
+      "    async with Connector(loop=loop) as connector:\n",
+      "        # Create connection to Cloud SQL database.\n",
+      "        conn: asyncpg.Connection = await connector.connect_async(\n",
+      "            f\"{project_id}:{region}:{instance_name}\",  # Cloud SQL instance connection name\n",
+      "            \"asyncpg\",\n",
+      "            user=f\"{database_user}\",\n",
+      "            password=f\"{database_password}\",\n",
+      "            db=f\"{database_name}\",\n",
+      "        )\n",
+      "\n",
+      "        await register_vector(conn)\n",
+      "\n",
+      "        # Create an HNSW index on the `product_embeddings` table.\n",
+      "        await conn.execute(\n",
+      "            f\"\"\"CREATE INDEX ON product_embeddings\n",
+      "              USING hnsw(embedding {operator})\n",
+      "              WITH (m = {m}, ef_construction = {ef_construction})\n",
+      "            \"\"\"\n",
+      "        )\n",
+      "\n",
+      "        await conn.close()\n",
+      "\n",
+      "\n",
+      "# Run the SQL commands now.\n",
+      "await main()  # type: ignore"
+    ],
+    "metadata": {
+      "id": "EJUDntZ1KTk7",
+      "cellView": "form"
+    },
+    "execution_count": null,
+    "outputs": []
+  },
+  {
+    "cell_type": "code",
+    "source": [
+      "# @markdown Create an IVFFLAT index on the `product_embeddings` table:\n",
+      "lists =  100 # @param {type:\"integer\"}\n",
+      "operator =  \"vector_cosine_ops\"  # @param [\"vector_cosine_ops\", \"vector_l2_ops\", \"vector_ip_ops\"]\n",
+      "\n",
+      "# Quick input validations.\n",
+      "assert lists, \"⚠️ Please input a valid value for lists.\"\n",
+      "\n",
+      "from pgvector.asyncpg import register_vector\n",
+      "import asyncio\n",
+      "import asyncpg\n",
+      "from google.cloud.sql.connector import Connector\n",
+      "\n",
+      "\n",
+      "async def main():\n",
+      "    loop = asyncio.get_running_loop()\n",
+      "    async with Connector(loop=loop) as connector:\n",
+      "        # Create connection to Cloud SQL database.\n",
+      "        conn: asyncpg.Connection = await connector.connect_async(\n",
+      "            f\"{project_id}:{region}:{instance_name}\",  # Cloud SQL instance connection name\n",
+      "            \"asyncpg\",\n",
+      "            user=f\"{database_user}\",\n",
+      "            password=f\"{database_password}\",\n",
+      "            db=f\"{database_name}\",\n",
+      "        )\n",
+      "\n",
+      "        await register_vector(conn)\n",
+      "\n",
+      "        # Create an IVFFLAT index on the `product_embeddings` table.\n",
+      "        await conn.execute(\n",
+      "            f\"\"\"CREATE INDEX ON product_embeddings\n",
+      "              USING ivfflat(embedding {operator})\n",
+      "              WITH (lists = {lists})\n",
+      "            \"\"\"\n",
+      "        )\n",
+      "\n",
+      "        await conn.close()\n",
+      "\n",
+      "\n",
+      "# Run the SQL commands now.\n",
+      "await main()  # type: ignore"
+    ],
+    "metadata": {
+      "id": "7kFKBuysMk2I",
+      "cellView": "form"
+    },
+    "execution_count": null,
+    "outputs": []
+  },
   {
    "cell_type": "markdown",
    "metadata": {