Google Gemini + SerpAPI + Supabase for Electronic Q&A

Instant answers from electronic datasheets—plus automatic web updates when your knowledge is missing

This n8n workflow lets you ask electronic component questions in chat and get answers grounded in indexed PDF datasheets using Google Gemini and a Supabase vector store. If the needed information isn’t found, it automatically searches the web with SerpAPI, downloads a public datasheet PDF, indexes it, and answers again from the updated knowledge base.

What this workflow does

• Receives your chat message and starts a Google Gemini-powered agent with conversation memory.
• Retrieval from Supabase: searches a Supabase vector store of embedded datasheet content to find the most relevant passages for your question.
• Fallback to web search: when retrieved context is insufficient, it uses SerpAPI (Google Search) to locate a public datasheet PDF URL.
• Triggers ingestion: sends the PDF URL to a separate “Update Database” workflow/tool to ingest the new document.
• Updates the vector knowledge base: downloads the PDF, extracts text, splits it into chunks, generates Google Gemini embeddings, and inserts chunks into the Supabase datasheets vector table.
• Re-runs retrieval and answers: queries Supabase again and returns the final response using only the retrieved datasheet information.

Use cases

• Answering capacitor, resistor, regulator, and IC datasheet questions (e.g., parameters, pin functions, operating conditions).
• Reducing manual time spent searching PDFs when a team asks component-specific questions.
• Keeping an internal electronic Q&A knowledge base current by automatically indexing newly found public datasheets.
• Building an automation layer for SaaS operators or engineering teams that rely on accurate, source-backed component answers.

Technical details

• Integrations: Google Gemini (PaLM) for chat and embeddings, Supabase vector store (with vector extension and match_documents RPC), SerpAPI for Google Search.
• Nodes/tools used: sticky note, http request, n8nn8n-nodes-langchainchat, n8nn8n-nodes-langchainagent, extract from file, http request tool.
• Workflow components: chat agent with memory, retrieval pipeline, PDF URL fallback, and a dedicated “Update Database” ingestion path.