Skip to content

EXP (Experimental Workflow)

An experimental WhatsApp-based educational quiz system that uses AI to evaluate student responses and provide personalized feedback. This workflow serves as a testing ground for new features before they're moved to production, including AI agent configurations, tool integrations, and prototype functionalities.

Purpose

No business context provided yet — add a context.md to enrich this documentation.

This experimental workflow appears to be designed for testing educational features, specifically: - Interactive WhatsApp-based learning experiences - AI-powered response evaluation using OpenAI - Student progress tracking through Airtable - Automated feedback delivery via Twilio

⚠️ Warning: This workflow shares the same webhook path as production PBA workflows and should only be activated during testing phases to avoid conflicts.

How It Works

  1. User Registration: When a new WhatsApp message arrives, the system checks if the user exists in Airtable
  2. New User Setup: For new users, creates a user record with default settings (Lesson 16, Question 1)
  3. Question Delivery: Retrieves the current question from Airtable and sends both text and audio versions via WhatsApp
  4. Response Collection: Stores a resume URL in the user record and waits for the student's response
  5. AI Evaluation: When a response arrives, an OpenAI-powered AI agent evaluates the answer against predefined criteria
  6. Feedback Delivery: Sends personalized feedback and scoring information back to the student
  7. Progress Tracking: Updates user progress in Airtable and determines next steps
  8. Continuation: Either moves to the next question or completes the lesson based on progress

Workflow Diagram

graph TD
    A[WhatsApp Webhook] --> B[Query User Table]
    B --> C{Is New User?}
    C -->|Yes| D[Create New User Record]
    C -->|No| E[Get Session State]
    D --> F[Get Current Question]
    E --> G[AI Agent]
    F --> H[Send Question]
    H --> I[Send Audio]
    I --> J[Respond to Webhook]
    J --> K[Store Resume URL]
    K --> L[Wait for Response]
    L --> M[POST to Resume URL]
    G --> N{Evaluation Logic}
    N --> O[Send Feedback]
    O --> P[Update Progress]
    P --> Q{Last Question?}
    Q -->|Yes| R[Send Completion]
    Q -->|No| S[Continue to Next]

Trigger

  • Type: Webhook (HTTP POST)
  • Path: /webhook/33898c22-f5f0-45f0-9983-3fd19c2daebb
  • Source: WhatsApp messages via Twilio
  • Method: POST
  • Status: Currently disabled (experimental use only)

Nodes Used

Node Type Purpose
Webhook Receives WhatsApp messages from Twilio
HTTP Request Queries and updates Airtable records
Airtable Manages user data, questions, and progress
Twilio Sends WhatsApp messages (text and audio)
AI Agent Evaluates student responses using OpenAI
OpenAI Chat Model Provides AI language processing capabilities
Wait Pauses execution for user responses
If Conditional logic for user flow control
Set Data transformation and field mapping
Memory Buffer Maintains conversation context
Respond to Webhook Returns responses to webhook calls

External Services & Credentials Required

Required Services

  • Twilio: WhatsApp Business API integration
  • OpenAI: GPT model for response evaluation
  • Airtable: Database for users, questions, and progress tracking

Credentials Needed

  • Twilio Account SID and Auth Token
  • OpenAI API Key
  • Airtable Personal Access Token
  • Airtable Base ID: appzp49MGzOjgteaz

Airtable Tables

  • Users: Student profiles and current progress
  • Questions: Quiz questions with evaluation criteria
  • User Progress: Detailed progress tracking
  • User Responses: Historical response data

Environment Variables

No specific environment variables are configured in this workflow. All credentials are managed through n8n's credential system.

Data Flow

Input

  • WhatsApp message data from Twilio webhook
  • User phone number (WaId)
  • Message body (student response)
  • Profile name

Processing

  • User lookup and creation
  • Question retrieval and delivery
  • AI-powered response evaluation
  • Progress calculation and updates

Output

  • WhatsApp messages (questions, audio, feedback)
  • Updated Airtable records
  • Resume URLs for conversation continuity
  • Webhook responses for flow control

Error Handling

The workflow includes basic error handling through: - Conditional logic for new vs. existing users - Disabled nodes to prevent execution conflicts - Resume URL mechanism for conversation recovery

However, comprehensive error handling appears limited in this experimental version.

Known Limitations

  • Webhook Conflict: Shares webhook path with production workflows
  • Experimental Status: Not intended for production use
  • Limited Error Handling: Basic error recovery mechanisms
  • Single Lesson Focus: Appears hardcoded to Lesson 16
  • Manual Activation Required: Must be manually enabled/disabled for testing
  • PBA Workflows: Production workflows that share the same webhook endpoint
  • Tool Workflows: Referenced as dependencies for the AI agent functionality

Setup Instructions

1. Import Workflow

  1. Copy the workflow JSON
  2. Import into your n8n instance
  3. Ensure all required nodes are available

2. Configure Credentials

  1. Set up Twilio credentials with Account SID and Auth Token
  2. Add OpenAI API key for GPT model access
  3. Configure Airtable credentials with Personal Access Token
  4. Verify Airtable base access to appzp49MGzOjgteaz

3. Airtable Setup

Ensure these tables exist with proper fields: - Users: Phone Number, Name, Current Lesson, Current Question Index, Resume URL - Questions: Question Number, Question Text, Intro Audio URL, evaluation criteria - User Progress: User links, scores, completion status

4. Testing Protocol

  1. Deactivate all production PBA workflows
  2. Activate this experimental workflow
  3. Test features thoroughly
  4. Document findings and issues
  5. Deactivate this workflow
  6. Reactivate production workflows

5. Webhook Configuration

  • Configure Twilio to send WhatsApp messages to the webhook URL
  • Ensure webhook path doesn't conflict with active workflows
  • Test webhook connectivity before activation

6. AI Agent Configuration

  • Verify OpenAI model selection (currently gpt-4.1-mini)
  • Test AI evaluation prompts with sample responses
  • Adjust scoring criteria as needed

⚠️ Important: Keep this workflow disabled unless actively testing to prevent conflicts with production systems.