n8n Workflow: Index Web Pages to Supabase with Jina + Gemini

Index your web pages into a Supabase vector store—automatically with Jina + Google Gemini (n8n workflow)

This n8n workflow fetches readable text from web pages using jina.ai, cleans and restructures it with Google Gemini, generates embeddings, and stores the results in your Supabase vector table—while tracking each URL’s indexing status in an n8n Data Table.

What this workflow does

• Triggers either from the built-in n8n form (enter a URL and type) or by manually running the workflow to process records in the Data Table marked as not indexed.
• De-duplicates by checking whether the submitted URL already exists in the n8n Data Table; if it does, the workflow deletes matching documents in Supabase via the Supabase REST API.
• Fetches raw page content using jina.ai to extract the page’s readable text.
• Cleans and restructures the extracted text with Google Gemini (without rewriting—focused on cleaning/reformatting).
• Chunks, embeds, and inserts the cleaned content into Supabase: it splits text into chunks, attaches metadata (including url and type), generates Google Gemini embeddings, and inserts documents into the Supabase vector store table.
• Updates indexing status in the n8n Data Table with the cleaned text and sets the status to indexed, then waits briefly before continuing batch processing.

Use cases

• Build a SaaS knowledge base from product pages or documentation URLs and keep it updated.
• Run batch indexing for multiple URLs stored in an n8n Data Table until everything is marked indexed.
• Refresh embeddings for specific pages by re-running with a URL and type.

Technical details

• Integrations/Credentials: Google Gemini (PaLM) API and Supabase
• n8n components: if, set, wait, switch, data table, form trigger
• Supabase REST API: delete request (replace YOUR_PROJECT_REF and ensure the REST endpoint can delete rows)

Perfect for n8n users and automation engineers who want a reliable web page indexing to Supabase pipeline using jina.ai + Gemini embeddings.