Build an OpenAI Assistant with Google Drive Integration

Summary

This workflow automates the process of creating an OpenAI Assistant that uses a document stored in Google Drive as its knowledge base. It downloads a Google Doc, converts it to PDF, uploads it to OpenAI, attaches it to a newly created assistant, and finally provides a chat interface to interact with that assistant. Perfect for building custom customer support bots, internal knowledge base assistants, or any conversational agent that relies on company documents.

Step-by-Step Walkthrough

The workflow is split into two main flows: a setup flow triggered manually, and a chat flow triggered by incoming messages.

Setup Flow (Manual Trigger)

• When clicking ‘Test workflow’ (Manual Trigger) — No auth. This node starts the setup process. It executes two parallel branches: creating the assistant and downloading the document.
• Google Drive (Google Drive node) — OAuth2. Downloads a specific Google Document by its file ID and converts it to PDF format (using the googleFileConversion option). The resulting PDF is stored as a binary property (data.pdf).
• OpenAI (OpenAI node) — API Key auth. Creates a new OpenAI Assistant with the following parameters:
  • Name: "Travel with us" Assistant
  • Model: gpt-4o-mini
  • Instructions: A detailed prompt instructing the assistant to only use the attached document and avoid general knowledge.
  • Operation: create
• OpenAI2 (OpenAI node) — API Key auth. Uploads the PDF file from the Google Drive step to OpenAI using the file resource and assistants purpose. The binary data is taken from the data.pdf property.
• OpenAI1 (OpenAI node) — API Key auth. Updates the previously created assistant (by its assistant ID) to include the uploaded file (via the file_ids option). This attaches the document as the assistant’s knowledge source.

Chat Flow (Real-time)

• When chat message received (Chat Trigger) — No auth. This webhook-based trigger listens for user messages (e.g., from an embedded chat widget). It passes each message to the assistant node.
• Window Buffer Memory (Memory Buffer Window) — No auth. Stores the recent conversation history (windowed memory) so the assistant can reference previous messages. Connected to the assistant via the ai_memory connection.
• OpenAI Assistent (OpenAI node) — API Key auth. Uses the same assistant ID as created in the setup flow. It processes each incoming message using the attached document and memory, returning an answer based solely on the document content.

Setup Instructions

To use this workflow, you need accounts for the following services:

1. OpenAI – Create an account at platform.openai.com and generate an API key. In n8n, configure an OpenAI credential with that API key.
2. Google Drive – Enable the Google Drive API in your Google Cloud Console, create an OAuth 2.0 client ID, and set up a Google Drive OAuth2 credential in n8n. Ensure the credential has access to the document you want to use.
3. Document – Identify the file ID of the Google Doc you want to use. You can get it from the URL (e.g., https://docs.google.com/document/d/<FILE_ID>/edit). Replace the fileId parameter in the Google Drive node with your actual file ID.
4. Assistant ID – After running the setup flow once, note the assistant ID created by the OpenAI node. Update the assistantId in both OpenAI1 and OpenAI Assistent nodes with this ID (or use the list mode to select it from your OpenAI account).

Use Cases & Variations

• Customer Support Chatbot – Replace the travel agency instructions with your own business FAQs or product manuals. The assistant will answer only from those documents.
• Internal Knowledge Base – Use a Google Doc containing company policies, onboarding guides, or technical documentation. Employees can query the assistant directly.
• Multiple Documents – Extend the workflow to download and upload several files (e.g., using a loop or multiple Google Drive nodes) and attach them all to the assistant via file_ids.
• Different Document Sources – Replace Google Drive with other storage services (e.g., Dropbox, S3) or fetch documents via HTTP requests.
• Model Customization – Change the model to gpt-4 or gpt-3.5-turbo for different cost/accuracy trade-offs. Adjust the instructions and temperature as needed.

{
  "id": "AjJ7O98qjw8XVirk",
  "meta": {
    "instanceId": "a4bfc93e975ca233ac45ed7c9227d84cf5a2329310525917adaf3312e10d5462",
    "templateCredsSetupCompleted": true
  },
  "name": "Build an OpenAI Assistant with Google Drive Integration",
  "tags": [
    {
      "id": "2VG6RbmUdJ2VZbrj",
      "name": "Google Drive",
      "createdAt": "2024-12-04T16:50:56.177Z",
      "updatedAt": "2024-12-04T16:50:56.177Z"
    },
    {
      "id": "paTcf5QZDJsC2vKY",
      "name": "OpenAI",
      "createdAt": "2024-12-04T16:52:10.768Z",
      "updatedAt": "2024-12-04T16:52:10.768Z"
    }
  ],
  "nodes": [
    {
      "id": "8a00e7b2-8348-47d2-87db-fe40b41a44f1",
      "name": "When clicking ‘Test workflow’",
      "type": "n8n-nodes-base.manualTrigger",
      "position": [
        180,
        260
      ],
      "parameters": {},
      "typeVersion": 1
    },
    {
      "id": "1d8fe39a-c7b9-4c38-9dc6-0fbce63151ba",
      "name": "Google Drive",
      "type": "n8n-nodes-base.googleDrive",
      "position": [
        480,
        380
      ],
      "parameters": {
        "fileId": {
          "__rl": true,
          "mode": "list",
          "value": "1JG7ru_jBcWu5fvgG3ayKjXVXHVy67CTqLwNITqsSwh8",
          "cachedResultUrl": "https://docs.google.com/document/d/1JG7ru_jBcWu5fvgG3ayKjXVXHVy67CTqLwNITqsSwh8/edit?usp=drivesdk",
          "cachedResultName": "[TEST] Assistente Agenzia viaggi"
        },
        "options": {
// ... truncated (copy to see full JSON)