Dr. Dovid Berry AI Parenting Agent Workflow

July 22, 2025 · View on GitHub

alt text

An n8n workflow that creates an AI-powered parenting advice system using multiple language models, PDF generation, and automated email delivery.

Overview

Dr. Dovid Berry is an AI parenting research assistant that provides thoroughly researched, evidence-based parenting advice. The system accepts questions through a web form, processes them through AI models, generates PDF reports, and delivers personalized responses via email.

Workflow Architecture

The workflow consists of 8 interconnected nodes:

Webhook - Receives form submissions from Tally.so
Code Node - Processes form data and extracts recipient information
AI Agent - Generates structured parenting advice using language models
OpenRouter Chat Model - Primary AI model (Claude Sonnet 4)
Google Gemini Chat Model - Alternative AI model
Structured Output Parser - Ensures consistent JSON response format
HTTP Request (PDFShift) - Converts HTML responses to PDF
Google Drive Upload - Saves PDFs to cloud storage
Gmail Send - Delivers formatted email responses

Key Features

AI-Powered Responses

Uses Claude Sonnet 4 via OpenRouter as the primary model
Google Gemini 2.0 Flash as backup/alternative
Structured output ensures consistent formatting
Contextual advice for parents in Jerusalem, Israel

PDF Generation

PDFShift API converts HTML to professional PDFs
Automatic file naming based on question summaries
Print-optimized formatting

Automated Delivery

Email templates with professional styling
Google Drive integration for file storage
Multiple recipient support
Branded email design with Dr. Dovid Berry persona

Form Integration

Tally.so form for question submission
Dynamic recipient selection
Question preprocessing and validation

Technical Implementation

Language Models

Primary: anthropic/claude-sonnet-4 (via OpenRouter)
Secondary: models/gemini-2.0-flash (via Google AI)
Structured JSON output with predefined schema

PDF Processing

Service: PDFShift API
Format: A4, portrait orientation
Features: Print optimization, CSS styling

Cloud Storage

Platform: Google Drive
Organization: Automatic folder structure
Access: Shareable links in email responses

Email System

Provider: Gmail API
Template: Custom HTML with inline CSS
Features: Responsive design, branded styling

Response Structure

The AI generates responses in this JSON format:

{
  "summary_file_safe": "question-topic-in-lowercase-with-hyphens",
  "summary_human_readable": "Question Topic In Title Case",
  "summarised_prompt": "Brief summary of the user's question",
  "html_response": "<p>Email-ready HTML content</p>",
  "html_full_document": "<!DOCTYPE html>...Complete HTML document..."
}

Setup Requirements

API Keys and Credentials

OpenRouter API - For Claude Sonnet 4 access
Google AI API - For Gemini model access
PDFShift API - For PDF generation
Google Drive OAuth - For file storage
Gmail OAuth - For email sending

Form Configuration

Tally.so form with specific field structure:
- "Your question for Dr Dovid Berry" (text field)
- "Send answer to" (multiple choice for recipients)

Webhook Setup

n8n webhook endpoint configured to receive Tally.so submissions
POST method with JSON payload processing

File Structure

├── config.json                 # n8n workflow configuration
├── email-template.html         # Email template for responses
├── images/
│   └── 1.png                   # Workflow diagram
└── README.md                   # This documentation