Added twelvelabs container runtime nb

samples/notebooks/container_runtime/dash_container_runtime_twelvelabs_app.ipynb

@@ -0,0 +1,58 @@
+{
+ "metadata": {
+  "kernelspec": {
+   "display_name": "Streamlit Notebook",
+   "name": "streamlit"
+  }
+ },
+ "nbformat_minor": 5,
+ "nbformat": 4,
+ "cells": [
+  {
+   "cell_type": "markdown",
+   "id": "f9ab3d0d-8bca-422e-ace6-ae19aded5148",
+   "metadata": {
+    "name": "cell1",
+    "collapsed": false
+   },
+   "source": "# Multimodal AI app for videos using Twelve Labs\n\nIn three easy steps:\n\n1) pip install twelvelabs\n2) Create twelvelabs client to generate video embeddings\n3) Use Snowflake's VECTOR_COSINE_SIMILARITY function to find (similar) videos based on text query\n\nPrerequisites: \n> *Create [network rule](https://docs.snowflake.com/en/sql-reference/sql/create-network-rule) and [external access integration](https://docs.snowflake.com/en/sql-reference/sql/create-external-access-integration) (EAI) for installing `twelvelabs` Python packages* and [enable](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-external-access#enable-existing-external-access-integrations-eai) it for this notebook.\n\n> *Obtain a [Twelve Labs API key](https://api.twelvelabs.io).*\n\n\n## What is Container Runtime? \n\nIt provides a flexible environment to build and operationalize a variety of workloads, especially hashtag#ML that require Python packages from multiple sources and powerful compute — both CPUs and GPUs. With this Snowflake-native experience, you can also train models, perform hyperparameter tuning as well as batch inference.\n\nHere are some resources for you to get started:\n\n👉 [Documentation](https://docs.snowflake.com/en/user-guide/ui-snowsight/notebooks-on-spcs) includes quickstart guides\n\n👉 [Github repo](https://github.com/Snowflake-Labs/snowflake-demo-notebooks) of tutorials and example notebooks\n\n👉 [YouTube](https://www.youtube.com/playlist?list=PLavJpcg8cl1Efw8x_fBKmfA2AMwjUaeBI) demo playlist"
+  },
+  {
+   "cell_type": "code",
+   "id": "3775908f-ca36-4846-8f38-5adca39217f2",
+   "metadata": {
+    "language": "python",
+    "name": "Install_twelvelabs",
+    "collapsed": false
+   },
+   "source": "!pip install twelvelabs",
+   "execution_count": null,
+   "outputs": []
+  },
+  {
+   "cell_type": "code",
+   "id": "2dee16f1-240c-4424-905a-e4d45569596f",
+   "metadata": {
+    "language": "python",
+    "name": "Generate_video_embeddings",
+    "collapsed": false,
+    "codeCollapsed": false
+   },
+   "outputs": [],
+   "source": "from twelvelabs import TwelveLabs\nfrom twelvelabs.models.embed import EmbeddingsTask\nfrom snowflake.snowpark.context import get_active_session\nsession = get_active_session()\nTWELVE_LABS_API_KEY =\"tlk_2VAE3AA2XGRHKXXXXXXXXXXXX\"\n# Initialize the Twelve Labs client\ntwelvelabs_client = TwelveLabs(api_key=TWELVE_LABS_API_KEY)\ndef get_video_embeddings(url: str) -> float:\n    twelvelabs_client = TwelveLabs(api_key=TWELVE_LABS_API_KEY)\n    task = twelvelabs_client.embed.task.create(\n        engine_name=\"Marengo-retrieval-2.6\",\n        video_url=url)\n    task.wait_for_done()\n    task_result = twelvelabs_client.embed.task.retrieve(task.id)\n    embeddings = []\n    for v in task_result.video_embeddings:\n        embeddings.append({\n            'embedding': v.embedding.float,\n            'start_offset_sec': v.start_offset_sec,\n            'end_offset_sec': v.end_offset_sec,\n            'embedding_scope': v.embedding_scope\n        })\n    return embeddings if embeddings else None\n\nvideo_url = 'https://videos.s3.us-west-1.amazonaws.com/misc/videos/demo.mp4'\nembeddings = get_video_embeddings(video_url)",
+   "execution_count": null
+  },
+  {
+   "cell_type": "code",
+   "id": "73487b72-beb7-4542-b29b-e0b3fd51a1ff",
+   "metadata": {
+    "language": "python",
+    "name": "Find_text_to_video_similarity_score",
+    "collapsed": false
+   },
+   "outputs": [],
+   "source": "def get_similar_videos(text_query):\n    # Twelve Labs Embed API supports text-to-embedding  \n    text_query_embeddings = twelvelabs_client.embed.create(\n      engine_name=\"Marengo-retrieval-2.6\",\n      text=text_query,\n      text_truncate=\"start\"\n    ).text_embedding.float\n\n    return session.sql(f\"\"\"SELECT \n        round(VECTOR_COSINE_SIMILARITY({embeddings[0]['embedding']}::VECTOR(FLOAT, 1024),\n        {text_query_embeddings}::VECTOR(FLOAT, 1024)),2) as similarity\"\"\") \\\n    .to_pandas().iloc[0]['SIMILARITY']\n\nget_similar_videos('snowflake notebooks')",
+   "execution_count": null
+  }
+ ]
+}