RAG Chatbot for Procurement: Supabase + TogetherAI + Openrouter

This workflow implements a Retrieval-Augmented Generation (RAG) chatbot tailored for commerce and procurement. The first part (run once) fetches content from Google Docs, splits it into 1000-char chunks using a code node, generates embeddings with TogetherAI's API, and stores them in a Supabase 'embed' table for persistent vector storage.

The second part activates on chat messages: it embeds the user query with TogetherAI, performs a vector similarity search via Supabase RPC to retrieve top 5 relevant chunks, aggregates results, and likely generates responses (integrating Openrouter for LLM inference, though preview cuts off). Telegram trigger is available but disabled.

Benefits include cost-effective, accurate responses grounded in your procurement docs, reducing hallucinations in AI chats. Ideal for e-commerce teams needing instant Q&A on suppliers, pricing, or policies without manual searches. Saves hours on customer support and internal queries, scalable with Supabase.