Telegram RAG Chatbot: Sync Google Drive to Supabase (n8n)

Turn your Google Drive PDFs into a Telegram RAG chatbot—fully synced to Supabase

This n8n workflow lets you chat in Telegram using Retrieval-Augmented Generation (RAG) over PDFs stored in Google Drive. It automatically indexes new and updated PDFs into a Supabase vector store—and removes knowledge when files move to a Drive “trash” folder.

What this workflow does

• Telegram → RAG answer: When a user sends a Telegram message, n8n passes the text to a LangChain AI Agent.
• Context retrieval from Supabase: The agent uses a Supabase vector store (configured for vector search with OpenAI embeddings) to find relevant document chunks, then generates a response.
• Chat response back to Telegram: The generated answer is sent to the user in Telegram.
• Google Drive → Vector indexing: When a new PDF is created in a watched Google Drive folder, the workflow downloads it, extracts text, chunks it, generates embeddings, and inserts vectors into the Supabase documents table.
• PDF updates → re-index: When a PDF is updated in another watched Google Drive folder, the workflow deletes existing vectors that match the file name, then re-downloads, re-extracts, re-embeds, and re-inserts.
• Trash handling → knowledge removal: If a file appears in a designated Google Drive “trash” folder, the workflow deletes matching vectors from Supabase and deletes the file from Google Drive.

Use cases

• Let teams ask questions about internal PDFs (policies, manuals, proposals) directly in Telegram.
• Keep a SaaS support knowledge base continuously updated from Google Drive without manual reindexing.
• Maintain accurate RAG results by automatically removing outdated documents when they’re moved to trash.

Technical details

• n8n nodes: telegram, set, supabase, google drive, sticky note, and n8n-nodes-langchainagent.
• AI & data: OpenRouter chat model, Postgres chat memory, and Supabase vector search (pgvector) using OpenAI embeddings.