Production-grade AI automation workflows built with n8n. Each project solves a real business problem using LLMs, multi-agent architectures, and third-party integrations.
Built by @ivancidev — Full Stack & AI Automation Engineer
| # | Project | Description | Stack | Nodes |
|---|---|---|---|---|
| 01 | RAG System | Auto-ingests Google Drive docs into PGVector and exposes a chat interface | Gemini, PGVector, PostgreSQL | 31 |
| 02 | AI Proposal Generator | Multi-agent system: requirement → architecture → SCRUM estimate → UI mockups → Excel + PPTX | Claude, Tavily, SharePoint | 68 |
| 03 | WhatsApp AI Bot | Multimodal bot: handles text, images (GPT-4o Vision) and voice (Whisper + TTS) | GPT-4o, Whisper, WhatsApp API | 28 |
| 04 | OCR Invoice Processor | Watches a Drive folder, OCR-extracts invoices via Vision API, logs to Sheets | Google Vision, Drive, Sheets | 11 |
| 05 | Personal Assistant Agent | Conversational agent that manages Gmail (read/draft/send) and calendar via chat | Gemini, Gmail, MCP | 10 |
| 06 | Web Intelligence Scraper | Scrapes websites with Firecrawl, AI-extracts structured data, saves to Google Docs + Sheets | Gemini, Firecrawl, Google Docs | 17 |
| 07 | Google Maps Scraper | Lead gen pipeline: Maps → Apify scrape → website contact extraction via Firecrawl → Sheets | Apify, Firecrawl, Sheets | 15 |
/01-rag-system
Automated RAG pipeline that monitors a Google Drive folder, detects new or modified files (via MD5 checksum), splits and embeds them into PostgreSQL PGVector using Gemini embeddings, and exposes a conversational chat interface.
Key design decisions:
- Incremental ingestion — only re-processes files that changed (MD5 diff)
- Cleans old vectors before re-embedding to avoid duplicates
- Handles nested folder structures recursively
- Legal assistant prompt pre-configured for document Q&A with source citations
Drive folder → Schedule → MD5 check → Gemini embed → PGVector
│
Chat interface + AI Agent
Stack: Google Drive · Google Gemini · PostgreSQL PGVector (Neon) · n8n AI Agent
/02-ai-proposal-generator
The most complex workflow: a supervisor multi-agent system that turns a client requirement into a complete technical and commercial proposal with zero manual work.
Pipeline stages:
- Functional Agent — analyzes the requirement and produces a structured spec
- Technical Supervisor — delegates to 5 specialist agents (Frontend, Backend, Cloud, DevOps, Automation) and consolidates their outputs
- Human approval loop — form with change request feedback at each stage
- UI Agent — generates HTML mockups → converts to PNG for client preview
- SCRUM Estimator — story points + sprint timeline
- Cost Calculator — uses Tavily to search real market rates, builds Excel file
- Gantt Generator — HTML → PNG → Excel
- PPT Generator — corporate PowerPoint, uploaded to SharePoint
Stack: Claude (Anthropic) · Tavily · SharePoint · n8n multi-agent toolWorkflow
/03-whatsapp-ai-bot
Multimodal WhatsApp bot with per-user memory. Routes incoming messages by type and handles each with the right AI model.
| Message type | Processing |
|---|---|
| Text | GPT-4.1-mini + conversation memory |
| Image | GPT-4o Vision analysis → GPT-4.1-mini response |
| Image document | GPT-4o Vision → GPT-4.1-mini response |
| Voice | Whisper STT → GPT-4.1-mini → TTS opus reply |
Stack: WhatsApp Business API · OpenAI GPT-4o · Whisper · TTS · n8n
/05-personal-assistant-agent
Conversational AI agent for personal productivity. Manages Gmail and calendar through natural language chat. Built-in business rules enforce email drafting, working hours, and priority classification automatically.
Key features: Gmail read/draft/send, MCP tool integration, calendar enforcement, per-session memory, priority-based email classification.
Stack: Google Gemini · Gmail OAuth2 · n8n MCP Client · Buffer Memory
/06-course-web-scraper
Reads a list of URLs from Google Sheets, scrapes them (HTTP + Firecrawl for JS-heavy sites), converts to Markdown, analyzes with Gemini AI, and saves structured summaries to Google Docs + updates the Sheet row.
Key features: Dual-scraper strategy, Gemini AI extraction, Google Docs output, Sheets as