Post LinkedIn content from a Supabase knowledge base using Google Gemini

Description

Quick overview
Automates your LinkedIn content creation pipeline. It syncs Google Drive files directly into a Supabase knowledge base, uses a Google Gemini and Tavily AI Agent to perform real-time web research, and automatically generates, designs, and publishes daily structured posts with custom AI images.

How it works
New files added to a watched Google Drive folder are automatically downloaded, text-extracted, chunked, and embedded with Google Gemini before being stored in a Supabase vector table.
When an existing file is updated, its previous Supabase rows are deleted first, then the refreshed content is re-downloaded, re-chunked, and re-embedded to keep the knowledge base current.
Every morning at 8am, a schedule trigger kicks off an AI Content Creation Agent built with Google Gemini.
The agent queries the Supabase knowledge base via a vector-store tool to pull a relevant company concept, service, or methodology.
It then searches the web with Tavily to find recent data, news, or trends that support the chosen idea, and generates structured post content (title, body, hashtags) plus an image prompt using a structured output parser.
Google Gemini generates a promotional image from that prompt, and the finished post—with image—is published automatically to a LinkedIn organization page.

Setup
Connect your Google Drive credentials and point the "file created" and "file updated" triggers at the folder that holds your knowledge base source documents.
Add Supabase credentials, create a vector table (e.g. "documents") with a match_documents function, and confirm the file_id metadata field used to identify and delete stale rows on updates.
Add Google Gemini credentials for the chat model, embeddings, and image generation nodes, and Tavily credentials for the web search tool.
Connect your LinkedIn credentials, set the target organization (or personal profile) to publish to, then review the agent's prompt, output schema, and the 8am schedule before activating the workflow.

Requirements
Google Drive account with a folder containing the source knowledge base documents
Supabase project with pgvector enabled and a "documents" table configured for the Supabase vector store node
Google Gemini (PaLM) API credentials for chat, embeddings, and image generation
Tavily API account for web search
LinkedIn account (organization page or personal profile) with publishing permissions

Customization
Adjust the schedule trigger to post at a different time or frequency
Edit the AI agent's system prompt and output schema to match your brand voice, language, or post structure
Change the chunk size/overlap in the text splitter nodes to fit your document types
Swap the image style prompt or color palette to match your visual identity
Update the Supabase table name or metadata schema if your vector store setup differs

Additional info
Uses two parallel Google Drive triggers (created / updated) so the vector store stays in sync automatically without manual re-indexing.