Get Job Alerts
One email when a teaching platform opens hiring or a new AI training project drops. No weekly filler.
Job alerts
Get Job Alerts
One email when a platform on this site starts hiring, checked by hand. No spam, leave any time.
Clients with growing document libraries have no way to search them intelligently — files pile up in Google Drive with zero semantic retrieval. This AI automation solves that by continuously indexing Drive files into a Supabase vector store, making every document searchable through RAG pipelines.
What This Automation Does
- Watches a Google Drive folder and automatically triggers re-indexing whenever a file is created or updated
- Routes each file to the correct content extractor based on type — PDF, Word document, Excel spreadsheet, or CSV
- Converts document text into OpenAI embeddings and stores them in a Supabase vector store for semantic search
- Stores tabular data from spreadsheets and CSVs as structured JSONB records alongside vector embeddings for hybrid retrieval
Tools Used
- n8n
- Google Drive
- Supabase
- OpenAI
Where to Get Hired for This Skill
On Contra, top freelancers across this stack have earned 145 combined verified reviews from real client projects.
Source: Contra freelancer search · refreshed 30 May 2026
Start Earning as a Freelancer on Contra
Contra is a commission-free professional network for independents. Browse live AI automation work and keep what you earn.
Join Contra Free →How To Build It
Provision the Supabase database schema
A one-time setup routine executes SQL to create the required Postgres tables for documents, metadata, and row-level records, plus the vector similarity match function that powers semantic search queries downstream.
Watch the client's Drive folder for changes
Two separate Google Drive listeners cover both file creation and file update events in the target folder, ensuring the index stays current without requiring any manual re-runs or scheduled polling.
Clear stale embeddings before re-indexing
Before any file is processed, existing vector embeddings and document rows for that file are deleted from Supabase so duplicate or outdated chunks never pollute search results.
Extract and route content by file type
A conditional routing step directs each downloaded file to the appropriate content extractor — PDFs and Word documents are chunked into text passages using a character text splitter, while Excel and CSV files are parsed into structured JSONB row records and stored separately.
Embed text chunks and upsert into the vector store
Extracted text passages are sent to OpenAI's embeddings API, and the resulting vectors are upserted into Supabase alongside document metadata — title, Drive URL, and file ID — making every chunk retrievable by semantic similarity.
Pitfalls
- Google Drive OAuth tokens expire or lose scope after a permission change, silently stopping the folder watchers — clients need to be warned to re-authorise credentials whenever Drive sharing settings are modified.
- OpenAI's embeddings API has rate limits that cause batch failures when a client's folder contains hundreds of large files triggered simultaneously; the workflow needs retry logic or a deliberate delay between embedding requests to stay within limits.
- Supabase free-tier projects pause after periods of inactivity, which means the vector store becomes unavailable mid-client-demo — always confirm the client is on a paid plan or build in a keep-alive ping before delivery.
FAQ
Can I build this without coding?
Almost entirely. The one exception is a short SQL script that runs once to set up the Supabase tables and vector similarity function — this is copy-paste SQL provided in the setup step, not custom programming. Everything else is configured through visual workflow tools.
How long does it take?
Expect 4 to 6 hours to build, test, and document a clean handoff for a new client. If you are reusing the workflow for a second client, setup drops to under 2 hours since only the Drive folder path and Supabase credentials need to change.
What can I charge?
Pricing depends on the client's document volume and whether they want a one-time delivery or ongoing maintenance. A one-time RAG infrastructure setup is typically scoped as a fixed project, with an optional monthly retainer covering credential refreshes, schema changes, and new file-type support.
Which tool is required vs optional?
Google Drive, Supabase, and OpenAI are all required — the entire pipeline depends on them working together. The file-type routing handles PDFs, Word docs, Excel, and CSV out of the box, but you can drop extractors for formats the client does not use to simplify the build.
This is original DigiNo analysis. The underlying automation pattern is a community workflow template – view the original on n8n.
Get Job Alerts
One email when a teaching platform opens hiring or a new AI training project drops. No weekly filler.
Job alerts
Get Job Alerts
One email when a platform on this site starts hiring, checked by hand. No spam, leave any time.
Build a WhatsApp AI assistant for voice, images, and PDFs