January 22, 2026

Google Drive + Gemini: ask your docs, get answers

Lisa Granqvist Partner Workflow Automation Expert

Get a free AI assessment → ⬇️ Use template

You know the feeling: the answer is somewhere in your Google Drive, but finding it means skimming five PDFs, two Docs, and that one slide deck you forgot existed. It’s not hard work. It’s just endless work.

This Drive Gemini Q&A setup hits students hardest during exam weeks. But it’s also a quiet tax on marketers pulling messaging from past campaigns, and on consultants who keep “just one more” client folder in Drive. You get a chat where you ask normal questions and receive grounded answers pulled from your own documents.

Below you’ll see how the automation indexes Drive folders into a searchable knowledge base, then uses Gemini to answer with context, memory, and document retrieval.

How This Automation Works

The full n8n workflow, from trigger to final output:

n8n Workflow Template: Google Drive + Gemini: ask your docs, get answers

Click to explore

flowchart LR

    subgraph sg0["When Executed by Another Workflow Flow"]
        direction LR
        n8@{ icon: "mdi:vector-polygon", form: "rounded", label: "Embeddings Google Gemini4", pos: "b", h: 48 }
        n9@{ icon: "mdi:robot", form: "rounded", label: "Default Data Loader2", pos: "b", h: 48 }
        n10["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/postgres.svg' width='40' height='40' /></div><br/>Execute a SQL query"]
        n11["<div style='background:#f5f5f5;padding:10px;border-radius:8px;display:inline-block;border:1px solid #e0e0e0'><img src='https://flowpast.com/wp-content/uploads/n8n-workflow-icons/code.svg' width='40' height='40' /></div><br/>Code in JavaScript"]
        n12@{ icon: "mdi:play-circle", form: "rounded", label: "When Executed by Another Wor..", pos: "b", h: 48 }
        n13@{ icon: "mdi:swap-vertical", form: "rounded", label: "Loop Over Items", pos: "b", h: 48 }
        n14@{ icon: "mdi:cog", form: "rounded", label: "Search files and folders", pos: "b", h: 48 }
        n15@{ icon: "mdi:cube-outline", form: "rounded", label: "Insert into Supabase Vectors..", pos: "b", h: 48 }
        n16@{ icon: "mdi:cog", form: "rounded", label: "Download File", pos: "b", h: 48 }
        n16 --> n15
        n13 --> n16
        n11 --> n10
        n10 --> n14
        n9 -.-> n15
        n14 --> n13
        n8 -.-> n15
        n15 --> n13
        n12 --> n11
    end

    subgraph sg1["When chat message received Flow"]
        direction LR
        n0@{ icon: "mdi:vector-polygon", form: "rounded", label: "Embeddings Google Gemini", pos: "b", h: 48 }
        n1@{ icon: "mdi:cube-outline", form: "rounded", label: "Supabase Vector Store", pos: "b", h: 48 }
        n2@{ icon: "mdi:memory", form: "rounded", label: "Postgres Chat Memory", pos: "b", h: 48 }
        n3@{ icon: "mdi:wrench", form: "rounded", label: "Calculator", pos: "b", h: 48 }
        n4@{ icon: "mdi:play-circle", form: "rounded", label: "When chat message received", pos: "b", h: 48 }
        n5@{ icon: "mdi:brain", form: "rounded", label: "Google Gemini Chat Model", pos: "b", h: 48 }
        n6@{ icon: "mdi:wrench", form: "rounded", label: "Folder all file to vector", pos: "b", h: 48 }
        n7@{ icon: "mdi:robot", form: "rounded", label: "Study Agent", pos: "b", h: 48 }
        n3 -.-> n7
        n2 -.-> n7
        n1 -.-> n7
        n0 -.-> n1
        n5 -.-> n7
        n6 -.-> n7
        n4 --> n7
    end

    %% Styling
    classDef trigger fill:#e8f5e9,stroke:#388e3c,stroke-width:2px
    classDef ai fill:#e3f2fd,stroke:#1976d2,stroke-width:2px
    classDef aiModel fill:#e8eaf6,stroke:#3f51b5,stroke-width:2px
    classDef decision fill:#fff8e1,stroke:#f9a825,stroke-width:2px
    classDef database fill:#fce4ec,stroke:#c2185b,stroke-width:2px
    classDef api fill:#fff3e0,stroke:#e65100,stroke-width:2px
    classDef code fill:#f3e5f5,stroke:#7b1fa2,stroke-width:2px
    classDef disabled stroke-dasharray: 5 5,opacity: 0.5
    class n12,n4 trigger
    class n9,n7 ai
    class n5 aiModel
    class n3,n6 ai
    class n2 ai
    class n15,n1 ai
    class n8,n0 ai
    class n10 database
    class n11 code
    classDef customIcon fill:none,stroke:none
    class n10,n11 customIcon

The Problem: Your Knowledge Is Trapped in Drive

Google Drive is great at storage. It’s not great at answering questions. Search helps when you remember the exact phrase, but it falls apart when you only remember the idea, the topic, or the “thing they said about pricing in that proposal.” Then you start opening files one by one, scrolling, doing Ctrl+F, and hoping the right paragraph jumps out. Multiply that by a few folders and a few projects, and you’re losing real time plus focus, which honestly is the expensive part.

The friction compounds. Here’s where it breaks down in day-to-day work.

You reread the same documents repeatedly because there’s no fast way to extract just the relevant passage.
Important details get missed when you’re scanning quickly, so answers end up “close enough” instead of correct.
Sharing folders doesn’t solve it, because access is not the same thing as comprehension.
When you try to use an AI chat without retrieval, it guesses, which means you still have to verify everything manually.

The Solution: Turn Drive Folders Into a Chatty Knowledge Base

This workflow gives you a conversational assistant that actually reads from your Google Drive files. When you share a Google Drive folder link (or a Drive link inside a chat), the automation fetches the files, converts them into plain text, and creates semantic “fingerprints” (embeddings) using Google Gemini’s embedding model. Those embeddings get stored in a Supabase Postgres database with vector search, so the system can later retrieve the most relevant passages even when your question doesn’t match exact wording. Then, when you ask a question in chat, the agent searches your indexed docs, pulls the best snippets, and responds naturally using Gemini 2.5 Pro while keeping recent conversation history.

The workflow starts from a chat message trigger, then routes to an AI agent that can call tools. One tool indexes Drive folders into Supabase, another retrieves relevant passages, and a built-in calculator handles math without derailing the conversation. The final output is a grounded answer that stays anchored to your files, not generic web knowledge.

What You Get: Automation vs. Results

What This Workflow Automates

Results You’ll Get

It detects Google Drive links shared in chat and kicks off indexing automatically.
Files are pulled from a Drive folder, downloaded, and converted to plain text.
Gemini generates 768-dimension embeddings and stores them in Supabase (pgvector) for semantic search.
The agent retrieves the best-matching passages and answers with chat memory kept in Postgres.

What used to take about 30 minutes of skimming becomes a few chat questions.
You get faster, more consistent answers because the assistant searches by meaning, not keywords.
Context carries over between sessions with a 10-message memory window, so you don’t repeat yourself as much.
Teams spend less time “looking for the doc” and more time using what they already wrote.
Answers stay grounded in your Drive content, which makes review and verification much easier.

Example: What This Looks Like

Say you have one Drive folder with 30 study PDFs and lecture notes. Before this, you might open 6 files and spend maybe 10 minutes each to find the one paragraph you need, so about an hour per question-heavy session. With the workflow: you paste the folder link once, wait about 2 minutes for indexing, then ask questions in chat and get passages back immediately. Even if you do this three times a week, that’s a few hours you stop donating to search.

What You’ll Need

n8n instance (try n8n Cloud free)
Self-hosting option if you prefer (Hostinger works well)
Google Drive to fetch and download your documents.
Google Gemini API for embeddings and chat responses.
Supabase (Postgres + pgvector) to store vectors and run semantic search.
Google Drive OAuth2 credentials (create in Google Cloud Console).
Gemini API key (get it from Google AI Studio).

Skill level: Intermediate. You’ll connect credentials, paste a folder URL, and confirm a few nodes are pointing at the right services.

Don’t want to set this up yourself? Talk to an automation expert (free 15-minute consultation).

How It Works

A chat message kicks things off. The workflow uses a chat trigger (often paired with Telegram in setups like this) so you can paste a Drive link and ask questions in the same place.

Drive links are recognized and prepared. A parsing step extracts the folder or file reference, then the system initializes the vector database side so storage and retrieval are ready for that content.

Documents get indexed in batches. n8n retrieves the Drive items, downloads each file, converts it to text, generates embeddings with Gemini, and stores vectors in Supabase. It loops through items in controlled batches, which keeps runs stable even when folders are large.

Answers come from retrieval, not guessing. When you ask a question, the agent searches Supabase for the most similar passages, uses Postgres chat memory to keep context, and responds via Gemini 2.5 Pro. If your question involves calculations, it can call the calculator tool as well.

You can easily modify the “what counts as a source” (folders, file types, or how much text gets chunked) to fit your needs. See the full implementation guide below for customization options.

Step-by-Step Implementation Guide

Step 1: Configure the Chat Trigger

Set up the entry points for both real-time chat interactions and the sub-workflow ingestion process.

Add and open Chat Message Trigger and set Public to true so users can send messages directly.
Add and open Subworkflow Trigger Start and set Input Source to jsonExample.
In Subworkflow Trigger Start, set JSON Example to { "Drive_Folder_link": "https://drive.google.com/drive/folders/example" } to document the expected input format.

You can manually run Subworkflow Trigger Start with a real Drive URL to validate ingestion before enabling the chat experience.

Step 2: Connect Google Drive and Parse Links

Configure Google Drive access and convert user-provided links into Drive IDs for listing and downloading files.

Open Parse Drive Link and keep the JavaScript Code as provided to extract folder or file IDs from Drive URLs.
Open Retrieve Drive Items and set Resource to fileFolder.
In Retrieve Drive Items, set Folder ID to {{ $('Parse Drive Link').item.json.folderId }}.
Credential Required: Connect your googleDriveOAuth2Api credentials to Retrieve Drive Items.
Open Download Drive File and set Operation to download.
In Download Drive File, set File ID to {{ $json.id }} and keep the conversion option to text/plain.
Credential Required: Connect your googleDriveOAuth2Api credentials to Download Drive File.

⚠️ Common Pitfall: If the Drive link points to a file rather than a folder, ensure the file is accessible to the connected Google account or the list operation will return empty results.

Step 3: Configure Vector Database Initialization and Ingestion

Initialize the Postgres vector table, iterate through files, load documents, embed them, and store the vectors in Supabase.

Open Initialize Vector DB and set Operation to executeQuery with the provided SQL that creates the documents table and match_documents function.
Credential Required: Connect your postgres credentials to Initialize Vector DB.
Open Batch Item Iterator and keep defaults to iterate through Drive items in batches.
Open Binary Document Loader and set Data Type to binary so downloaded files are parsed as documents.
Open Gemini Embedding Generator and ensure it remains connected as the embedding provider for Store Vectors in Supabase.
Credential Required: Connect your googlePalmApi credentials to Gemini Embedding Generator.
Open Store Vectors in Supabase and set Mode to insert with Table Name set to documents and Query Name set to match_documents.
Credential Required: Connect your supabaseApi credentials to Store Vectors in Supabase.

⚠️ Common Pitfall: The SQL in Initialize Vector DB drops the documents table every run. Remove the DROP TABLE line if you need to preserve existing embeddings.

Step 4: Set Up the AI Assistant and Retrieval Tools

Configure the agent, language model, tools, and memory so chat responses leverage stored study material.

Open Learning Assistant Agent and set Text to {{ $json.chatInput }}.
In Learning Assistant Agent, keep the provided System Message to guide study interactions and Drive link handling.
Open Gemini Chat Language Model and set Model Name to models/gemini-2.5-pro.
Credential Required: Connect your googlePalmApi credentials to Gemini Chat Language Model.
Open Supabase Vector Retrieval and set Mode to retrieve-as-tool with the provided Tool Description.
Credential Required: Connect your supabaseApi credentials to Supabase Vector Retrieval.
Open Gemini Embedding Builder and leave parameters default so it supplies embeddings for Supabase Vector Retrieval.
Credential Required: Connect your googlePalmApi credentials to Gemini Embedding Builder.
Open Postgres Conversation Memory and set Session Key to {{ $json.sessionId }} with Session ID Type set to customKey and Context Window Length set to 10.
Credential Required: Connect your postgres credentials to Postgres Conversation Memory.
Open Run Sub-Workflow Config Req and keep the Drive_Folder_link mapping set to {{ /*n8n-auto-generated-fromAI-override*/ $fromAI('Drive_Folder_link', ``, 'string') }}.
Keep Math Utility Tool connected for on-the-fly calculations inside responses.

Ensure all AI tool nodes are connected to Learning Assistant Agent as shown so the agent can call retrieval, memory, and workflow tools during chat.

Step 5: Confirm Execution Flow and Looping

Verify the ingestion chain and chat flow are wired exactly as intended.

Confirm the ingestion path runs in this order: Subworkflow Trigger Start → Parse Drive Link → Initialize Vector DB → Retrieve Drive Items → Batch Item Iterator → Download Drive File → Store Vectors in Supabase → back to Batch Item Iterator for the next item.
Confirm the chat path runs: Chat Message Trigger → Learning Assistant Agent.
Verify Binary Document Loader is connected to Store Vectors in Supabase via the document input.

Step 6: Test and Activate Your Workflow

Run end-to-end tests for both ingestion and chat, then activate the workflow for production use.

Click Execute Workflow on Subworkflow Trigger Start and pass a real Drive folder link to verify items are retrieved and vectors are stored in Supabase.
Manually trigger Chat Message Trigger with a question about your materials and confirm Learning Assistant Agent responds with relevant content.
Successful execution should show documents inserted by Store Vectors in Supabase and a chat response that references stored data.
Toggle the workflow to Active to begin handling live chat messages and Drive ingestion in production.

🔒

Unlock Full Step-by-Step Guide

Get the complete implementation guide + downloadable template

Common Gotchas

Google Drive credentials can expire or need specific permissions. If things break, check the Drive node’s OAuth status in n8n credentials first, then confirm the folder is shared with the connected account.
If you’re using Wait nodes or external rendering, processing times vary. Bump up the wait duration if downstream nodes fail on empty responses.
Default prompts in AI nodes are generic. Add your brand voice early or you’ll be editing outputs forever.

Frequently Asked Questions

How long does it take to set up this Drive Gemini Q&A automation?

About 15 minutes if your credentials are ready.

Do I need coding skills to automate Drive Gemini Q&A?

No. You’ll mainly connect accounts and paste in keys. The only “code” piece is already included for parsing Drive links.

Is n8n free to use for this Drive Gemini Q&A workflow?

Yes. n8n has a free self-hosted option and a free trial on n8n Cloud. Cloud plans start at $20/month for higher volume. You’ll also need to factor in Gemini API usage (often just a few dollars a month unless you index huge libraries).

Where can I host n8n to run this automation?

Two options: n8n Cloud (managed, easiest setup) or self-hosting on a VPS. For self-hosting, Hostinger VPS is affordable and handles n8n well. Self-hosting gives you unlimited executions but requires basic server management.

Can I customize this Drive Gemini Q&A workflow for Slack instead of chat/Telegram?

Yes, but you’ll swap the entry point. Replace the Chat Message Trigger with a Slack trigger, then pass the message text into the Learning Assistant Agent the same way. Common tweaks include restricting indexing to approved folders, changing the chunk size used before embedding, and adjusting the retrieval count so answers quote fewer or more passages.

Why is my Google Drive connection failing in this workflow?

Most of the time it’s OAuth. Reconnect the Google Drive OAuth2 credential in n8n and make sure the connected Google account can actually open the folder you’re indexing. Also check that your Google Cloud project has Drive API enabled, because it’s easy to forget after creating credentials. If failures only happen on some files, the document type can be the culprit (scanned PDFs and weird exports sometimes produce empty text). Finally, watch for rate limits when indexing large folders back-to-back.

How many documents can this Drive Gemini Q&A automation handle?

A lot, as long as your database and API limits can keep up. On n8n Cloud Starter, you’re limited by monthly executions, so big libraries may push you toward a higher plan. If you self-host, there’s no execution cap, and scale is mostly about server size and how quickly you want indexing to finish. In practice, most small teams start with a few hundred documents, then expand once they’re happy with the answer quality.

Is this Drive Gemini Q&A automation better than using Zapier or Make?

For this workflow, n8n has a few advantages: more complex logic with unlimited branching at no extra cost, a self-hosting option for unlimited executions, and native agent + vector retrieval components that are awkward (or pricey) to replicate elsewhere. Zapier or Make can work if you only need “upload file → summarize → send message.” But once you want semantic search, memory, and an agent that can call tools, you’ll feel the limits fast. If you’re torn, Talk to an automation expert and describe your volume and document types.

Once this is running, Drive stops being a filing cabinet and starts acting like a searchable brain. Set it up once, then use the time you get back for work that actually moves things forward.

Need Help Setting This Up?

Our automation experts can build and customize this workflow for your specific needs. Free 15-minute consultation—no commitment required.

Lisa Granqvist

Workflow Automation Expert

Expert in workflow automation and no-code tools.

{
"id": null,
"name": "Automated Study Assistant Workflow",
"versionId": null,
"meta": {
"instanceId": "c93fb84905ad020fe5b833be67636a66ee046ea04ecfa471f62fca75f0127c48",
"templateId": null,
"templateCredsSetupCompleted": null
},
"nodes": [
{
"id": "flowpast-topbar-10653",
"name": "Flowpast Branding",
"type": "n8n-nodes-base.stickyNote",
"position": [
0,
20
],
"parameters": {
"color": 7,
"width": 2224,
"height": 80,
"content": "## Flowpast.com | Automation Workflow Library\n**\ud83d\udcd6 Full tutorial & setup guide:** flowpast.com"
},
"typeVersion": 1
},
{
"id": "0db018d8-693f-4e47-be62-4b34d7b8d77f",
"name": "Gemini Embedding Builder",
"type": "@n8n/n8n-nodes-langchain.embeddingsGoogleGemini",
"position": [
372,
125
],
"parameters": [],
"credentials": {
"googlePalmApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "edf2e17e-a730-486b-8e2a-8acaef9e84a3",
"name": "Supabase Vector Retrieval",
"type": "@n8n/n8n-nodes-langchain.vectorStoreSupabase",
"position": [
798,
120
],
"parameters": {
"mode": "retrieve-as-tool",
"options": [],
"tableName": {
"__rl": true,
"mode": "list",
"value": "documents",
"cachedResultName": "documents"
},
"toolDescription": "Use this tool to search and retrieve relevant information from the user's study materials stored in the vector database. Query the documents to answer user questions accurately."
},
"credentials": {
"supabaseApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1.3
},
{
"id": "1a55495f-44be-4c71-9a9d-f4886a8980a8",
"name": "Postgres Conversation Memory",
"type": "@n8n/n8n-nodes-langchain.memoryPostgresChat",
"position": [
1074,
375
],
"parameters": {
"sessionKey": "={{ $json.sessionId }}",
"sessionIdType": "customKey",
"contextWindowLength": 10
},
"credentials": {
"postgres": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1.3
},
{
"id": "39d7a9b3-66d8-41fb-8454-6a80885131d1",
"name": "Math Utility Tool",
"type": "@n8n/n8n-nodes-langchain.toolCalculator",
"position": [
720,
380
],
"parameters": [],
"typeVersion": 1
},
{
"id": "d943532d-4ae7-4829-a381-191cf84ea622",
"name": "Chat Message Trigger",
"type": "@n8n/n8n-nodes-langchain.chatTrigger",
"position": [
2124,
160
],
"webhookId": "6f7911bb-b08c-40ba-b613-a81d3d26ee18",
"parameters": {
"public": true,
"options": []
},
"typeVersion": 1.3
},
{
"id": "f37c1723-0049-4b1d-8354-3acfd5179cb4",
"name": "Gemini Chat Language Model",
"type": "@n8n/n8n-nodes-langchain.lmChatGoogleGemini",
"position": [
1176,
125
],
"parameters": {
"options": [],
"modelName": "models/gemini-2.5-pro"
},
"credentials": {
"googlePalmApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "225fb496-37d1-4dd7-b008-179ebb0880cc",
"name": "Run Sub-Workflow Config Req",
"type": "@n8n/n8n-nodes-langchain.toolWorkflow",
"position": [
1050,
500
],
"parameters": {
"workflowId": {
"__rl": true,
"mode": "list",
"value": "",
"cachedResultUrl": "",
"cachedResultName": ""
},
"workflowInputs": {
"value": {
"Drive_Folder_link": "={{ /*n8n-auto-generated-fromAI-override*/ $fromAI('Drive_Folder_link', ``, 'string') }}"
},
"schema": [
{
"id": "Drive_Folder_link",
"type": "string",
"display": true,
"removed": false,
"required": false,
"displayName": "Drive_Folder_link",
"defaultMatch": false,
"canBeUsedToMatch": true
}
],
"mappingMode": "defineBelow",
"matchingColumns": [
"Drive_Folder_link"
],
"attemptToConvertTypes": false,
"convertFieldsToString": false
}
},
"typeVersion": 2.2
},
{
"id": "2cec8fcc-a3ed-459e-9e30-1fad8a7b6765",
"name": "Learning Assistant Agent",
"type": "@n8n/n8n-nodes-langchain.agent",
"position": [
1494,
285
],
"parameters": {
"text": "={{ $json.chatInput }}",
"options": {
"systemMessage": "You are a Study AI Assistant that helps users interact with their study materials in a natural, conversational way.\n\n## Core Behavior\n\n**Always respond conversationally and helpfully.** You can answer questions, provide information from stored materials, and assist with file uploads - all while maintaining a friendly, natural dialogue.\n\n## Input Handling\n\n### 1. Google Drive Links\nWhen you detect a Google Drive URL (folder or file):\n- **Pattern**: `https://drive.google.com/drive/folders/` or `https://drive.google.com/file/d/`\n- **Action**: Automatically trigger the `DriveFolderToSupabase` workflow\n- **Response**: Confirm the upload is processing: \"I'm uploading your files to the vector store. This will take a moment...\"\n\n### 2. Study Material Queries\nWhen users ask questions about their materials:\n- **Search the vector store** using available retrieval tools\n- **Always check the vector store first** before saying you don't have information\n- Provide clear, helpful answers with citations\n- Include document names, sections, or page numbers when available\n\n### 3. General Conversation\nWhen users engage in general conversation:\n- Respond naturally and helpfully\n- If they're asking about themselves or their materials, **search the vector store**\n- Use context from previous messages in the conversation\n- Be conversational, not robotic\n\n## Critical Rules\n\n1. **Never refuse to search**: If someone asks \"what is in the documents\" or \"tell me about X\", immediately query the vector store with relevant keywords\n2. **Infer intent**: Questions like \"about me\", \"what's my name\", or \"vector database\" should trigger a vector store search for relevant content\n3. **Use broad searches**: When unsure, search with general terms rather than refusing to help\n4. **Acknowledge limitations gracefully**: Only say you can't find information AFTER searching, not before\n5. **Maintain conversation context**: Reference previous exchanges naturally\n\n## Search Strategy\n\nWhen querying the vector store:\n- Use **keywords and concepts** from the user's question\n- Try **multiple related terms** if the first search yields poor results\n- For vague queries like \"tell me what's in the documents\", search with terms like: \"overview\", \"introduction\", \"main topic\", \"summary\"\n- **Always attempt a search** before saying you don't have the information\n\n## Response Format\n\n- **Direct answers** to questions\n- **Cite sources** when providing information from documents\n- **Suggest related topics** when appropriate\n- **Ask clarifying questions** only when absolutely necessary (not as a default)\n\n## Examples\n\n**Bad Response**: \"I need a specific question or topic to search for.\"\n**Good Response**: *[Searches vector store]* \"Based on your uploaded materials, I found information about [topic]. Here's what I can tell you...\"\n\n**Bad Response**: \"I don't have access to personal information like your name.\"\n**Good Response**: *[Searches vector store for personal info]* \"I searched your documents and found [relevant information], or if nothing is found: \"I searched your uploaded materials but didn't find personal information stored. What would you like to know about your study content?\"\n"
},
"promptType": "define"
},
"typeVersion": 2.2
},
{
"id": "54a5e290-1ec4-4b97-96ed-d424aaf3c2ca",
"name": "Gemini Embedding Generator",
"type": "@n8n/n8n-nodes-langchain.embeddingsGoogleGemini",
"position": [
0,
795
],
"parameters": [],
"credentials": {
"googlePalmApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "7682b868-5215-452d-b110-ff8007f2d059",
"name": "Binary Document Loader",
"type": "@n8n/n8n-nodes-langchain.documentDefaultDataLoader",
"position": [
90,
980
],
"parameters": {
"options": [],
"dataType": "binary"
},
"typeVersion": 1.1
},
{
"id": "ceacbea3-3c6a-47d1-83a6-386cb1166414",
"name": "Initialize Vector DB",
"type": "n8n-nodes-base.postgres",
"position": [
1452,
600
],
"parameters": {
"query": "DROP TABLE IF EXISTS documents CASCADE;\n\nCREATE EXTENSION IF NOT EXISTS vector;\n\nCREATE TABLE IF NOT EXISTS documents (\n  id bigserial PRIMARY KEY,\n  content text,\n  metadata jsonb,\n  embedding vector(768)\n);\n\nCREATE OR REPLACE FUNCTION match_documents(\n  query_embedding vector(768),\n  match_count int DEFAULT NULL,\n  filter jsonb DEFAULT '{}'::jsonb\n)\nRETURNS TABLE (\n  id bigint,\n  content text,\n  metadata jsonb,\n  similarity double precision\n)\nLANGUAGE sql\nAS $$\n  SELECT\n    d.id,\n    d.content,\n    d.metadata,\n    1 - (d.embedding <=> query_embedding) AS similarity\n  FROM documents d\n  WHERE (filter = '{}'::jsonb OR d.metadata @> filter)\n  ORDER BY d.embedding <=> query_embedding\n  LIMIT match_count;\n$$;\n",
"options": [],
"operation": "executeQuery"
},
"credentials": {
"postgres": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 2.6
},
{
"id": "90883ae9-8d17-4a72-83be-da4dae013343",
"name": "Parse Drive Link",
"type": "n8n-nodes-base.code",
"position": [
1692,
705
],
"parameters": {
"jsCode": "// Get the Drive_Folder_link from the workflow input\nconst driveUrl = $input.first().json.Drive_Folder_link;\n\n// Extract Google Drive folder/file ID from URL\nfunction getDriveId(url) {\n  const folderMatch = url.match(/\\/folders\\/([a-zA-Z0-9_-]+)/);\n  const fileMatch = url.match(/\\/file\\/d\\/([a-zA-Z0-9_-]+)/);\n  return folderMatch ? folderMatch[1] : (fileMatch ? fileMatch[1] : null);\n}\n\n// Process input items\nreturn items.map(item => {\n  const chatInput = item.json.chatInput || driveUrl || '';\n  const driveId = getDriveId(chatInput);\n\n  return {\n    json: {\n      originalInput: chatInput,\n      folderId: driveId,\n      driveId: driveId\n    }\n  };\n});\n"
},
"typeVersion": 2
},
{
"id": "1e5ac5c6-ae2c-400d-b531-a18c823a3d07",
"name": "Subworkflow Trigger Start",
"type": "n8n-nodes-base.executeWorkflowTrigger",
"position": [
2106,
665
],
"parameters": {
"inputSource": "jsonExample",
"jsonExample": "{\n  \"Drive_Folder_link\": \"https://drive.google.com/drive/folders/example\"\n}"
},
"typeVersion": 1.1
},
{
"id": "472c0470-a590-476a-b23b-77617b042a39",
"name": "Batch Item Iterator",
"type": "n8n-nodes-base.splitInBatches",
"position": [
768,
595
],
"parameters": {
"options": []
},
"typeVersion": 3
},
{
"id": "8e8a66a7-9a2c-4ed9-91b3-80d805b1dbab",
"name": "Retrieve Drive Items",
"type": "n8n-nodes-base.googleDrive",
"position": [
1038,
725
],
"parameters": {
"filter": {
"folderId": {
"__rl": true,
"mode": "id",
"value": "={{ $('Parse Drive Link').item.json.folderId }}"
}
},
"options": [],
"resource": "fileFolder"
},
"credentials": {
"googleDriveOAuth2Api": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 3
},
{
"id": "21559a2e-f0d3-40a1-8809-5f2a31cde811",
"name": "Store Vectors in Supabase",
"type": "@n8n/n8n-nodes-langchain.vectorStoreSupabase",
"position": [
774,
890
],
"parameters": {
"mode": "insert",
"options": {
"queryName": "match_documents"
},
"tableName": {
"__rl": true,
"mode": "list",
"value": "documents",
"cachedResultName": "documents"
}
},
"credentials": {
"supabaseApi": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 1
},
{
"id": "a818d7b0-1c5e-4273-96d1-d72ff2960823",
"name": "Download Drive File",
"type": "n8n-nodes-base.googleDrive",
"position": [
1026,
900
],
"parameters": {
"fileId": {
"__rl": true,
"mode": "id",
"value": "={{ $json.id }}"
},
"options": {
"googleFileConversion": {
"conversion": {
"docsToFormat": "text/plain"
}
}
},
"operation": "download"
},
"credentials": {
"googleDriveOAuth2Api": {
"id": "credential-id",
"name": ""
}
},
"typeVersion": 3
}
],
"pinData": [],
"connections": {
"Math Utility Tool": {
"ai_tool": [
[
{
"node": "Learning Assistant Agent",
"type": "ai_tool",
"index": 0
}
]
]
},
"Download Drive File": {
"main": [
[
{
"node": "Store Vectors in Supabase",
"type": "main",
"index": 0
}
]
]
},
"Batch Item Iterator": {
"main": [
[],
[
{
"node": "Download Drive File",
"type": "main",
"index": 0
}
]
]
},
"Parse Drive Link": {
"main": [
[
{
"node": "Initialize Vector DB",
"type": "main",
"index": 0
}
]
]
},
"Initialize Vector DB": {
"main": [
[
{
"node": "Retrieve Drive Items",
"type": "main",
"index": 0
}
]
]
},
"Binary Document Loader": {
"ai_document": [
[
{
"node": "Store Vectors in Supabase",
"type": "ai_document",
"index": 0
}
]
]
},
"Postgres Conversation Memory": {
"ai_memory": [
[
{
"node": "Learning Assistant Agent",
"type": "ai_memory",
"index": 0
}
]
]
},
"Supabase Vector Retrieval": {
"ai_tool": [
[
{
"node": "Learning Assistant Agent",
"type": "ai_tool",
"index": 0
}
]
]
},
"Gemini Embedding Builder": {
"ai_embedding": [
[
{
"node": "Supabase Vector Retrieval",
"type": "ai_embedding",
"index": 0
}
]
]
},
"Gemini Chat Language Model": {
"ai_languageModel": [
[
{
"node": "Learning Assistant Agent",
"type": "ai_languageModel",
"index": 0
}
]
]
},
"Retrieve Drive Items": {
"main": [
[
{
"node": "Batch Item Iterator",
"type": "main",
"index": 0
}
]
]
},
"Gemini Embedding Generator": {
"ai_embedding": [
[
{
"node": "Store Vectors in Supabase",
"type": "ai_embedding",
"index": 0
}
]
]
},
"Run Sub-Workflow Config Req": {
"ai_tool": [
[
{
"node": "Learning Assistant Agent",
"type": "ai_tool",
"index": 0
}
]
]
},
"Chat Message Trigger": {
"main": [
[
{
"node": "Learning Assistant Agent",
"type": "main",
"index": 0
}
]
]
},
"Store Vectors in Supabase": {
"main": [
[
{
"node": "Batch Item Iterator",
"type": "main",
"index": 0
}
]
]
},
"Subworkflow Trigger Start": {
"main": [
[
{
"node": "Parse Drive Link",
"type": "main",
"index": 0
}
]
]
}
}
}