Created using Colaboratory

gradient-embeders-and-generators-for-notion-rag.ipynb

@@ -0,0 +1,265 @@
+{
+ "nbformat": 4,
+ "nbformat_minor": 0,
+ "metadata": {
+ "colab": {
+ "provenance": [],
+ "authorship_tag": "ABX9TyP6v9IBjf1C1AJvH3peU6XG"
+ },
+ "kernelspec": {
+ "name": "python3",
+ "display_name": "Python 3"
+ },
+ "language_info": {
+ "name": "python"
+ }
+ },
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "source": [
+ "# Use Gradient Models for Notion RAG\n",
+ "\n",
+ "In this Colab, we will:\n",
+ "- Creating a custom Haystack component called `NotionExporter`\n",
+ "- Building an indexing pipeline to write our Notion pages into an `InMemoryDocumentStore` with embeddings\n",
+ "- Build a custom RAG pipeline to do question answering on our Notion pages"
+ ],
+ "metadata": {
+ "id": "_coq_qCuItbN"
+ }
+ },
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "id": "prmChq6T_T4l"
+ },
+ "outputs": [],
+ "source": [
+ "!pip install haystack-ai gradient-haystack notion-haystack\n",
+ "!pip install nest-asyncio\n",
+ "\n",
+ "import nest_asyncio\n",
+ "\n",
+ "nest_asyncio.apply()"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "import getpass\n",
+ "import os\n",
+ "\n",
+ "notion_api_key = getpass.getpass(\"Enter Notion API key:\")\n",
+ "gradient_access_token = getpass.getpass(\"Gradient Access token:\")"
+ ],
+ "metadata": {
+ "id": "JRogtDXaAMIF"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "source": [
+ "### Test the NotionExporter\n",
+ "\n",
+ "- You can follow the steps outlined in the Notion [documentation](https://developers.notion.com/docs/create-a-notion-integration#create-your-integration-in-notion) to create a new Notion integration, connect it to your pages, and obtain your API token.\n",
+ "- Page IDs in Notion are the tailing numbers at the end of the page URL, separated by a '-' at 8-4-4-4-12 digits"
+ ],
+ "metadata": {
+ "id": "aPXd4RjEKzBG"
+ }
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "from notion_haystack import NotionExporter\n",
+ "\n",
+ "exporter = NotionExporter(api_token=notion_api_key)"
+ ],
+ "metadata": {
+ "id": "3_blVmFsAGSf"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "exporter.run(page_ids=[\"6f98e9a6-a880-40e9-b191-1c4f41efec87\"])"
+ ],
+ "metadata": {
+ "id": "cK8VTkJlAfLs"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "source": [
+ "## Build an Indexing Pipeline to Write Notion Pages to a Document Store\n",
+ "\n",
+ "- Documentation on [`GradientDocumentEmbedder`](https://haystack.deepset.ai/integrations/gradient#usage)\n",
+ "- Documentation on [`DocumentSplitter`](https://docs.haystack.deepset.ai/v2.0/docs/documentsplitter)\n",
+ "- Documentation on [`DocumentWriter`](https://docs.haystack.deepset.ai/v2.0/docs/documentwriter)"
+ ],
+ "metadata": {
+ "id": "KlMkLSVjJVoW"
+ }
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "from haystack.components.preprocessors import DocumentSplitter\n",
+ "from gradient_haystack.embedders.gradient_document_embedder import GradientDocumentEmbedder\n",
+ "from haystack.components.writers import DocumentWriter\n",
+ "from haystack.document_stores import InMemoryDocumentStore\n",
+ "\n",
+ "\n",
+ "document_store = InMemoryDocumentStore()\n",
+ "exporter = NotionExporter(api_token=notion_api_key)\n",
+ "splitter = DocumentSplitter()\n",
+ "document_embedder = GradientDocumentEmbedder(access_token=gradient_access_token, workspace_id=\"9ee7071c-2fa9-4155-8edd-94ed371f1750_workspace\")\n",
+ "writer = DocumentWriter(document_store=document_store)\n"
+ ],
+ "metadata": {
+ "id": "bZefXI2cBRME"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "from haystack import Pipeline\n",
+ "\n",
+ "indexing_pipeline = Pipeline()\n",
+ "indexing_pipeline.add_component(instance=exporter, name=\"exporter\")\n",
+ "indexing_pipeline.add_component(instance=splitter, name=\"splitter\")\n",
+ "indexing_pipeline.add_component(instance=document_embedder, name=\"document_embedder\")\n",
+ "indexing_pipeline.add_component(instance=writer, name=\"writer\")"
+ ],
+ "metadata": {
+ "id": "JZVzXyHdDxwg"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "indexing_pipeline.connect(\"exporter.documents\", \"splitter.documents\")\n",
+ "indexing_pipeline.connect(\"splitter.documents\", \"document_embedder.documents\")\n",
+ "indexing_pipeline.connect(\"document_embedder.documents\", \"writer.documents\")"
+ ],
+ "metadata": {
+ "id": "ht3U0JNJEFW6"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "indexing_pipeline.run(data={\"exporter\":{\"page_ids\": [\"6f98e9a6-a880-40e9-b191-1c4f41efec87\"]}})"
+ ],
+ "metadata": {
+ "id": "mdn9N0XkEZBH"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "markdown",
+ "source": [
+ "## Build a RAG Pipeline with Cohere\n",
+ "\n",
+ "- Documentation on [`GradientTextEmbedder`](https://haystack.deepset.ai/integrations/gradient#usage)\n",
+ "- Documentation on [`PromptBuilder`](https://docs.haystack.deepset.ai/v2.0/docs/promptbuilder)\n",
+ "- Documentation on [`GradientGenerator`](GradientTextEmbedder)"
+ ],
+ "metadata": {
+ "id": "A7iIQrtJJ1-6"
+ }
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "import torch\n",
+ "from haystack.components.retrievers import InMemoryEmbeddingRetriever\n",
+ "from haystack.components.builders import PromptBuilder\n",
+ "from gradient_haystack.embedders.gradient_text_embedder import GradientTextEmbedder\n",
+ "from gradient_haystack.generator.base import GradientGenerator\n",
+ "\n",
+ "prompt = \"\"\" Answer the query, based on the\n",
+ "content in the documents.\n",
+ "\n",
+ "Documents:\n",
+ "{% for doc in documents %}\n",
+ " {{doc.content}}\n",
+ "{% endfor %}\n",
+ "\n",
+ "Query: {{query}}\n",
+ "\"\"\"\n",
+ "text_embedder = GradientTextEmbedder(access_token=gradient_access_token, workspace_id=\"9ee7071c-2fa9-4155-8edd-94ed371f1750_workspace\")\n",
+ "retriever = InMemoryEmbeddingRetriever(document_store=document_store)\n",
+ "prompt_builder = PromptBuilder(template=prompt)\n",
+ "generator = GradientGenerator(access_token=gradient_access_token,\n",
+ " workspace_id=\"9ee7071c-2fa9-4155-8edd-94ed371f1750_workspace\",\n",
+ " model_adapter_id=\"905db818-d031-4378-bd67-ac9804cb0961_model_adapter\",\n",
+ " max_generated_token_count=350)\n"
+ ],
+ "metadata": {
+ "id": "tbX35iA3_ImP"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "rag_pipeline = Pipeline()\n",
+ "\n",
+ "rag_pipeline.add_component(instance=text_embedder, name=\"text_embedder\")\n",
+ "rag_pipeline.add_component(instance=retriever, name=\"retriever\")\n",
+ "rag_pipeline.add_component(instance=prompt_builder, name=\"prompt_builder\")\n",
+ "rag_pipeline.add_component(instance=generator, name=\"generator\")\n",
+ "\n",
+ "rag_pipeline.connect(\"text_embedder\", \"retriever\")\n",
+ "rag_pipeline.connect(\"retriever.documents\", \"prompt_builder.documents\")\n",
+ "rag_pipeline.connect(\"prompt_builder\", \"generator\")"
+ ],
+ "metadata": {
+ "id": "71PMDYsUBOY8"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "question = \"What are the steps for creating a custom component?\"\n",
+ "result = rag_pipeline.run(data={\"text_embedder\":{\"text\": question},\n",
+ " \"prompt_builder\":{\"query\": question}})"
+ ],
+ "metadata": {
+ "id": "zszY4bgzB2M0"
+ },
+ "execution_count": null,
+ "outputs": []
+ },
+ {
+ "cell_type": "code",
+ "source": [
+ "print(result['generator']['replies'][0])"
+ ],
+ "metadata": {
+ "id": "6EbRQ3OPCDs4"
+ },
+ "execution_count": null,
+ "outputs": []
+ }
+ ]
+}