How josepozuelo/n8naiworkflow works
Project type and purpose
• This is an open‑source educational tool/demo (“n8n AI workflow maker”) that takes a natural‑language description of an automation workflow, uses GPT to plan and generate a JSON representation of that workflow, and then lets you paste it into an n8n canvas.
• Key features:
• A set of n8n JSON workflows under n8n_AI_workflows
• Scripts/workflows to build auxiliary data stores (vector store in Pinecone, parameter DB in Firestore) under n8n_data_workflows
• A small REPL Python script (create_embeddings.py) to populate the Pinecone index
File structure
• Top‑level files:
• README.md – project overview & instructions
• n8n_node_list_full.json – raw list of all n8n nodes
• n8n_AI_workflows/ – JSON definitions for the main GPT‑driven workflow + sub‑workflows (“if node parameters,” “action nodes parameters,” etc.)
• n8n_data_workflows/ – JSON workflows that scrape/generate parameter metadata for each n8n node into Firestore
• Replit embedding script (mentioned but lives externally as create_embeddings.py)
README for additional insights
• Three logical parts:
a) Main n8n JSON workflows for generating the user’s requested automation
b) Embedding store creation (Python REPL → Pinecone vector DB)
c) Node‑parameter metadata generation (n8n_data_workflows → Firestore)
• External dependencies:
• OpenAI API (GPT calls inside n8n HTTP Request nodes)
• Pinecone (vector similarity store)
• Firestore (parameter metadata store)
• n8n cloud or self‑hosted instance (runtime & editor)
Detail without overthinking
• Every JSON workflow and external store as a separate component
• How each data store is populated and consumed