RAG Multi-Agent System (n8n + Pinecone)
109-node n8n: Google Drive PDF → Pinecone vector store → Cohere embeddings → Ollama AI Agent → Airtop browser scraping → Apify actors. 5 sub-workflows. Full RAG + conversation memory.
Multi-agent RAG: PDF ingestion + vector search + web scraping + multi-tool LLM
Massive 109-node n8n multi-agent system combining RAG, web scraping, and AI orchestration.
5 Sub-Workflows
1. Google Drive PDF OCR Drive trigger → download → OCR API → field extraction → Google Sheets append
2. RAG Pipeline PDF → Recursive CharTextSplitter (1K chars, 200 overlap) → Cohere embed-english-v3.0 → Pinecone upsert → AI Agent query
3. Airtop Web Scraping Headless browser: navigate → wait → click → type → extract → structured output (JS-rendered pages)
4. Competitive Research (Apify) Apify actors (Amazon/LinkedIn scrapers) → dataset → Ollama synthesis → structured report
5. Chat Interface n8n Chat → Ollama Llama3.1 AI Agent → conversation memory → multi-tool calling (vector search + scraper + DB)
Technology Stack
| Service | Role |
|---|---|
| Pinecone | Vector store (1536-d) |
| Cohere | Embeddings (embed-english-v3.0) |
| Ollama | Local LLM (Llama3.1, zero API cost) |
| Airtop | Headless browser for JS-rendered pages |
| Apify | Web scraping actors |
| Google Workspace | Drive + Sheets + Gmail |
Why 109 Nodes? Each workflow has error handling branches, conditional routing, and validation steps. Explicit flow makes the system auditable and production-reliable.