Notion Knowledge Base Assistant

• AI-powered Search: Uses GPT-4o (or other OpenAI models) to understand user queries and generate responses.
• Notion Integration: Connects to a specified Notion database for knowledge retrieval.
• Dynamic Tooling: Leverages Langchain tools (toolHttpRequest) to query Notion's API for searching databases and retrieving page content.
• Contextual Memory: Employs Window Buffer Memory to maintain conversation context for more coherent interactions.
• Structured Data Handling: Extracts and formats data from Notion, including database details and tags, to inform the AI.
• Customizable AI Behavior: Configurable system message for the AI Agent, defining its role, behavior, error handling, and output format.
• Chat Interface: Provides a simple chat interface for user interaction via the When chat message received trigger.

Notion Knowledge Base Assistant Setup Guide

This guide outlines how to set up and use the Notion Knowledge Base Assistant workflow.

Prerequisites:

1. n8n Account: You need a running instance of n8n.
2. Notion Account: Access to a Notion workspace.
3. OpenAI Account: An API key for OpenAI models.

Setup Steps:

1. Add Notion Credentials:

   • Navigate to Credentials in your n8n workspace.
   • Click Add credential and select Notion API.
   • Follow the official Notion guide to create an integration and obtain your API key.
   • Give your credential a recognizable name (e.g., max-bot).

2. Add OpenAI Credentials:

   • Navigate to Credentials in your n8n workspace.
   • Click Add credential and select OpenAI API.
   • Enter your OpenAI API key.
   • Give your credential a recognizable name (e.g., Max's DevRel OpenAI account).

3. Configure Get database details Node:

   • Drag and drop the Get database details node onto your canvas.
   • In the Database field, select your Notion database. If you've recently created an integration, you might need to manually enter the databaseId or click the refresh/add button to link it. The workflow snippet uses 7ea9697d-4875-441e-b262-1105337d232e which is likely the ID of the "StarLens Company Knowledge Base".
   • Ensure the Notion API credential is set to the one you created in Step 1.

4. Configure Search notion database and Search inside database record Nodes:

   • These nodes use the toolHttpRequest type.
   • Ensure the Notion API credential is set to the one you created in Step 1 for both nodes.
   • The url parameter in Search notion database uses {{ $json.notionID }}. This notionID is dynamically fetched by the Format schema node from the Get database details node.
   • The placeholderDefinitions in Search notion database define how keywords and tags will be passed to the tool.
   • The url parameter in Search inside database record expects a page_id which will be dynamically provided by the AI Agent based on search results.

5. Configure OpenAI Chat Model Node:

   • Ensure the OpenAI API credential is set to the one you created in Step 2.
   • The model is set to gpt-4o. You can change this to other compatible models if needed.

6. Configure AI Agent Node:

   • The text field is set to ={{ $json.chatInput }}, which comes from the When chat message received trigger.
   • The systemMessage is crucial for defining the AI's behavior. It instructs the AI on its role, how to query the Notion database, desired output format, and error handling. It also dynamically includes the {{ $json.databaseName }}.

7. Configure When chat message received Node:

   • This node acts as the chat trigger.
   • Ensure the public option is enabled if you want to share the chat URL.
   • The initialMessages provide a greeting when the chat starts.
   • The webhookId (b76d02c0-b406-4d21-b6bf-8ad2c623def3) should be unique to your instance.

8. Workflow Connections:

   • The connections in the JSON define the flow: When chat message received triggers the process, Get database details fetches database info, Format schema prepares variables, and then the AI Agent uses OpenAI Chat Model, Search notion database, Search inside database record, and Window Buffer Memory to process the request.

9. Test the Workflow:

   • Click the Chat button at the bottom of the When chat message received node to test.
   • Alternatively, activate the workflow and copy the Chat URL provided by the When chat message received node to interact with the assistant.

Customization and Notes:

• Notion Database: Ensure the Notion database you are connecting to has fields like 'question' (rich text) and 'tags' (multi-select) for the search functionality to work effectively.
• Tag Options: The tagsOptions variable is dynamically generated from the Get database details node. If your Notion database structure changes, you might need to adjust how this is fetched.
• Error Handling: The Sticky Note1 provides important troubleshooting tips, particularly regarding the Get database details node and potential resource not found errors.
• Performance: The Get database details node fetches tag options, which adds a small delay. Refer to the detailed video mentioned in Sticky Note1 if you wish to optimize this.
• AI Model: You can switch the OpenAI Chat Model to other models if desired, ensuring compatibility with Langchain.