Automation / Personal Project
Resume RAG
Automates resume synchronization from Google Docs into Supabase vector embeddings and drafts tailored job application emails using an n8n LangChain chat agent.
How it works
Built in n8n, this two-part workflow manages document indexing and application email drafting. A Google Drive polling trigger detects resume changes in Google Docs, chunks the text into overlapping paragraph-packed segments, and generates 2048-dimensional embeddings via OpenRouter's Nemotron model before inserting them into Supabase PostgreSQL (resume_documents) and pruning stale chunks. A companion chat trigger loads the stored resume context into an n8n LangChain agent powered by Nemotron-3 Super 120B with automatic Qwen 27B fallback, drafting concise, truthful application emails and answering posting questions with 20-message session buffer memory for iterative adjustments.
Capabilities
- Automated Google Drive trigger synchronizing resume edits from Google Docs
- Paragraph-packed chunking with overlap and 2048-dimensional OpenRouter embeddings
- Supabase vector table management with atomic replacement and stale chunk cleanup
- LangChain conversational chat agent drafting tailored job application emails from postings
- Dual-model redundancy pairing Nemotron-3 Super 120B with automatic Qwen 27B fallback
- Multi-turn session buffer memory for conversational revisions and tone adjustments
Built with
- n8n
- Supabase
- PostgreSQL
- LangChain
- Google Docs
- JavaScript