Chat with PDF Word & PowerPoint via n8n, Apify & OpenAI

Chat with your PDF, Word, and PowerPoint documents—inside n8n

This n8n automation workflow lets you submit document links through an n8n form, converts them into searchable Markdown using Apify, and then answers your questions using OpenAI—with file + page citations pulled directly from your content.

What this workflow does

• Receives document URLs submitted via an n8n form trigger.
• Converts documents to RAG-ready chunks by sending URLs to Apify’s “Document to Markdown for AI & RAG” actor, which extracts text and splits it into heading-based Markdown chunks with page numbers. (It can optionally run OCR for scanned PDFs.)
• Cleans and reshapes chunk metadata into text plus metadata such as source file, page, and heading.
• Embeds chunks for semantic search using OpenAI embeddings, stored in an in-memory vector store.
• Chats via an n8n chat trigger: when you ask a question, the workflow finds the most relevant chunks and uses an OpenAI chat model to answer strictly from the documents, citing the source file and page.

Use cases

• Knowledge base Q&A for manuals, specs, and internal docs across PDF and Office formats.
• Support and sales enablement: answer customer questions with citations to the exact document pages.
• Research assistants that summarize and quote relevant sections from uploaded document links.

Technical details

• Apify: “Document to Markdown for AI & RAG” (requires an Apify API token via HTTP header auth: Authorization: Bearer <YOUR_TOKEN>).
• OpenAI: embeddings for vector search + chat model for document-grounded answers.
• n8n nodes / components shown in the workflow: form trigger, HTTP request, code, set, and n8nn8n-nodes-langchainagent.
• Configuration: ensure documents are accessible via public/shared URLs; optionally adjust chunk size and OCR behavior in the Settings step.