AI Agent for Notion Database with n8n

Building an AI Agent for Notion Database

• An AI agent is built for a Notion database using n8n, a workflow automation tool.
• The agent answers questions about the database by retrieving information.
• A connection between Notion and n8n is established.
• Key n8n nodes include: "Search Notion Database" (searches by keywords and tags), and "Search Inside Database Record" (retrieves info from a specific record).
• GPT-4 0-mini chat model powers the AI agent's capabilities.
• The prompt guides the agent, detailing database structure and desired output.

Setting Up the Connection

• Connect Notion to n8n by navigating to Notion settings, connections, and selecting "develop or manage integrations".
• An integration is created associating the Notion workspace, naming it, and configuring integration settings.
• The integration secret is copied and pasted into n8n's credentials.
• A successful connection is verified.

Building the Workflow

• The workflow starts with the "On Message Received" trigger.
• The "Get the Database" node fetches database details, avoiding the "simplify" option for complete data.
• The "Set" node formats output: extracting session ID, action, chat input, Notion ID, database name, and tag options.
• It converts tag options into a comma-separated string.
• The "AI Agent" node uses the GPT-4 0-mini chat model.
• The prompt is customized to instruct the AI on database operations, clarity, structure, and response formatting.
• Window buffer memory temporarily stores data.

Search Inside the Database

• The "Search Notion Database" node is an HTTP request tool.
• The database name is automatically retrieved using expressions.
• The URL is built using the database ID from the "Get the Database" node.
• Authentication happens via predefined credentials.
• The request body is a JSON object filtering by keyword and tag (using "or" condition).
• Placeholders are set for "keyword" and "tag" input.

Retrieving Page Content

• The "Search Inside Database Record" node uses HTTP requests.
• The node retrieves information from a particular page identified by its ID.
• The URL includes the page ID from the previous node.
• Only relevant Notion page fields are returned.
• Selected fields: ID, Type, Paragraph, Heading 1/2/3, Link Preview, Bulleted/Number List, and Text.

Conclusion

• The agent handles user queries, fetching and linking relevant information from the database.
• Individual page content retrieval increases the agent's efficiency.
• The process demonstrates how AI and automation streamline information retrieval.

AI Agent Configuration

• Empty fields can be left in the AI agent configuration, but selecting all fields might negatively impact accuracy.
• Renaming modules/nodes is recommended, though not mandatory.
• The example focuses on searching within a database record.

AI Agent Interaction Example

• The agent interacts in a dialogue format.
• The agent can locate and link to a Notion page containing latest sales pitches.
• The agent confirms the existence of sales pitches when queried.
• It retrieves sales pitch links from the Notion page when prompted.
• It effectively navigates and extracts only the sales pitch links.

Benefits of AI Agents

• AI agents streamline information retrieval through dialogue, saving time.
• They efficiently manage complex data workflows by handling multiple, nested queries.
• The example highlights how an AI agent can access data from a Notion database.

Resources

• A collaboration/project contact link and a template download link (possibly on Gumroad or Google Drive) are offered.