Facebook Messenger Chatbot That Learns From Your Documents

Go to Workflow
91 views
Built by Tristan V Tristan V
Created on June 07, 2026

Description

Quickstart Guide: Facebook Messenger Chatbot with Pinecone RAG

Step-by-step instructions to get this workflow running in n8n.

Prerequisites

Self-hosted n8n instance (v1.113.0+ recommended)
Community nodes enabled in n8n
Facebook Page (you must be an admin)
OpenAI account
Pinecone account (free Starter plan works)

Workflow Architecture

This workflow uses two webhooks with the same URL path but different HTTP methods:

| Webhook | Method | Purpose |
|---------|--------|---------|
| Facebook Verification Webhook | GET | Handles Facebook's webhook verification |
| Facebook Message Webhook | POST | Receives incoming messages |

Both webhooks share the same URL: https://your-n8n.com/webhook/facebook-messenger-webhook

n8n automatically routes requests based on HTTP method:
GET** requests → Verification flow
POST** requests → Message processing flow

RAG Enhancement

User Message → Batching → AI Agent ─┬─ OpenAI Chat Model
├─ Conversation Memory
└─ Pinecone Assistant Tool (RAG)


Search your documents


Answer with citations

Step 1: Install the Pinecone Assistant Community Node

The Pinecone Assistant node is a community node that must be installed separately.

In n8n, go to Settings → Community Nodes
Click Install a community node
Enter: @pinecone-database/n8n-nodes-pinecone-assistant
Click Install
Restart n8n if prompted

> Note: Community nodes must be enabled in your n8n instance. For Docker, set N8N_COMMUNITY_PACKAGES_ALLOW_INSTALL=true.

Step 2: Create Pinecone Account & Assistant

2.1 Create Pinecone Account

Go to Pinecone
Sign up for a free account (Starter plan includes 100 files per assistant)
Complete the onboarding

2.2 Create an Assistant

In the Pinecone console, go to Assistants
Click Create Assistant
Name it n8n-assistant (or choose your own name)
Select your preferred region
Click Create

2.3 Upload Your Documents

Click on your newly created assistant
Go to the Files tab
Click Upload Files
Upload your documents (PDFs, text files, etc.)
Wait for processing to complete

2.4 Get Your Pinecone API Key

In the Pinecone console, click on your profile/account
Go to API Keys
Copy your API key (or create a new one)

Step 3: Get Your OpenAI API Key

Go to OpenAI Platform
Sign in with your OpenAI account
Click Create new secret key
Copy and save the API key

Step 4: Create Facebook App & Get Page Access Token

4.1 Create Facebook App

Go to Facebook Developers
Click My Apps → Create App
Select Other → Next
Select Business → Next
Enter app name and contact email
Click Create App

4.2 Add Messenger Product

In your app dashboard, scroll to Add products to your app
Find Messenger and click Set up

4.3 Connect Your Facebook Page

In Messenger settings, find Access Tokens section
Click Add or Remove Pages
Select your Facebook Page and grant permissions
Click Done

4.4 Generate Page Access Token

Back in Messenger settings, find your page in the list
Click Generate Token
Copy and save the token

Step 5: Create Your Verify Token

The verify token is a secret string for Facebook webhook verification.

Create a random string (e.g., my-secret-token-12345)
Save this value - you'll need it in Steps 7 and 10

Step 6: Create n8n Credentials

6.1 Pinecone Credential

In n8n, go to Credentials → Add Credential
Search for "Pinecone" (or "Pinecone API")
Configure:
Name: Pinecone API
API Key: Paste your Pinecone API key from Step 2.4
Click Save

6.2 OpenAI API Credential

In n8n, go to Credentials → Add Credential
Search for "OpenAI API"
Configure:
Name: OpenAI API
API Key: Paste your OpenAI API key from Step 3
Click Save

6.3 Facebook Graph API Credential

In n8n, go to Credentials → Add Credential
Search for "Facebook Graph API"
Configure:
Name: Facebook Page Access Token
Access Token: Paste your Page Access Token from Step 4.4
Click Save

Step 7: Import the Workflow

In n8n, click Add Workflow → Import from File
Select the workflow.json file from this folder
The workflow will open in the editor

7.1 Configure the Verify Token

Find the "Is Token Valid?" node
Click on the node to open its settings
In the conditions, find Value 2 that shows YOUR_VERIFY_TOKEN_HERE
Replace it with your verify token from Step 5

7.2 Configure the Pinecone Assistant Name

Find the "Get context snippets in Pinecone Assistant" node
Click on the node to open its settings
Change Assistant Name from n8n-assistant to your actual assistant name (from Step 2.2)

Step 8: Connect Credentials to Nodes

8.1 Connect Facebook Credential

Update these 3 nodes with your Facebook credential:

Click on "Send Seen Indicator" → Select your Facebook Page Access Token credential
Click on "Send Typing Indicator" → Select your Facebook Page Access Token credential
Click on "Send Response to User" → Select your Facebook Page Access Token credential

8.2 Connect OpenAI Credential

Click on "OpenAI Chat Model" → Select your OpenAI API credential

8.3 Connect Pinecone Credential

Click on "Get context snippets in Pinecone Assistant" → Select your Pinecone API credential

Step 9: Publish the Workflow

Click Save to save the workflow
Click the Publish button to make the workflow live
Copy the webhook URL (e.g., https://your-n8n.com/webhook/facebook-messenger-webhook)

Step 10: Configure Facebook Webhook

Go to Facebook Developers → Your App → Messenger Settings
Find Webhooks section
Click Add Callback URL
Enter:
Callback URL: Your n8n webhook URL from Step 9
Verify Token: Same value from Step 5
Click Verify and Save
After verification, subscribe to webhook fields:
messages (required)
messaging_postbacks (recommended)

Step 11: Test Your Chatbot

11.1 Add Test Users (if needed)

With Standard Access, only users with app roles can message the bot:

Go to your Facebook App → App Roles → Roles
Add users as Testers
Those users must accept the invitation

11.2 Send a Test Message

Open Facebook Messenger
Search for your Facebook Page
Try these test messages:
"Hello!" - Should get a friendly greeting
"What information do you have?" - Should search your documents
"Tell me about [topic in your docs]" - Should return relevant information with context

How Pinecone RAG Works

User asks: "What are your return policies?"


┌─────────────────────────┐
│ AI Agent receives msg │
└─────────┬───────────────┘


┌─────────────────────────┐
│ Calls Pinecone Tool │ → Searches your documents
└─────────┬───────────────┘


┌─────────────────────────┐
│ Gets relevant snippets │ ← "Return Policy.pdf: Items can be..."
└─────────┬───────────────┘


┌─────────────────────────┐
│ AI formulates answer │ → "According to our policy, items can be..."
└─────────────────────────┘

Troubleshooting

| Problem | Solution |
|---------|----------|
| "Pinecone Assistant Tool" not found | Ensure community node is installed (Step 1) |
| "No relevant information found" | Upload more documents to your Pinecone Assistant |
| Webhook verification fails | Check verify token matches in n8n and Facebook |
| No response from bot | Check n8n execution logs for errors |
| "Error validating access token" | Regenerate Page Access Token in Facebook |
| AI Agent not using Pinecone tool | After import, open the "AI Agent1" node, make a small edit to the system message (e.g., add a space), and save. This re-initializes the tool bindings. |

Customization

Change the AI Behavior

Edit the "AI Agent1" node's system message to:
Adjust how it cites sources
Change personality/tone
Add specific instructions for your use case

Change the Assistant

Update the "Get context snippets in Pinecone Assistant" node to use a different assistant name.

Adjust Response Length

The workflow truncates responses to 1900 characters for Messenger. Edit the "Format Response" node to change this.

Resources

Pinecone Assistant Documentation
Pinecone Assistant n8n Node (GitHub)
Facebook Messenger Platform
OpenAI API Documentation
n8n Documentation

Nodes Used (5)

AI Agent
@n8n/n8n-nodes-langchain.agent
Code
n8n-nodes-base.code
HTTP Request
n8n-nodes-base.httpRequest
OpenAI Chat Model
@n8n/n8n-nodes-langchain.lmChatOpenAi
Simple Memory
@n8n/n8n-nodes-langchain.memoryBufferWindow