Answer Code of Conduct Questions in Slack with GPT-4 & RAG Technology

📘 Code of Conduct Q&A Slack Chatbot with RAG Powered

> Empower employees to instantly access and understand the company’s Code of Conduct via a Slack chatbot, powered by Retrieval-Augmented Generation (RAG) and LLMs.

🧑‍💼 Who’s it for

This workflow is designed for: HR and compliance teams** to automate policy-related inquiries Employees** who want quick answers to Code of Conduct questions directly inside Slack Startups or enterprises** that need internal compliance self-service tools powered by AI

⚙️ How it works / What it does

This RAG-powered Slack chatbot answers user questions based on your uploaded Code of Conduct PDF using GPT-4 and embedded document chunks. Here's the flow:

Receive Message from Slack: A webhook triggers when a message is posted in Slack. Check if it’s a valid query: Filters out non-user messages (e.g., bot mentions). Run Agent with RAG: Uses GPT-4 with Query Data Tool to retrieve relevant document chunks. Returns a well-formatted, context-aware answer. Send Response to Slack: Fetches user info and posts the answer back in the same channel. Document Upload Flow: HR can upload the PDF Code of Conduct file. It’s parsed, chunked, embedded using OpenAI, and stored for future query retrieval. A backup copy is saved to Google Drive.

🛠️ How to set up

Prepare your environment: Slack Bot token & webhook configured (Sample slack app manifest: https://wisestackai.s3.ap-southeast-1.amazonaws.com/slack_bot_manifest.json) OpenAI API key (for GPT-4 & embedding) Google Drive credentials (optional for backup)

Upload the Code of Conduct PDF: Use the designated node to upload your document (Sample file: https://wisestackai.s3.ap-southeast-1.amazonaws.com/20220419-ingrs-code-of-conduct-policy-en.pdf) This triggers chunking → embedding → data store.

Deploy the chatbot: Host the webhook and connect it to your Slack app. Share the command format with employees (e.g., @CodeBot Can I accept gifts from partners?)

Monitor and iterate: Improve chunk size or embed model if queries aren’t accurate. Review unanswered queries to enhance coverage.

📋 Requirements

n8n (Self-hosted or Cloud) Slack App (with chat:write, users:read, commands) OpenAI account (embedding + GPT-4 access) Google Drive integration (for backups) Uploaded Code of Conduct in PDF format

🧩 How to customize the workflow

| What to Customize | How to Do It | |-----------------------------|------------------------------------------------------------------------------| | 🔤 Prompt style | Edit the System & User prompts inside the Code Of Conduct Agent node | | 📄 Document types | Upload additional policy PDFs and tag them differently in metadata | | 🤖 Agent behavior | Tune GPT temperature or replace with different LLM | | 💬 Slack interaction | Customize message formats or trigger phrases | | 📁 Data Store engine | Swap to Pinecone, Weaviate, Supabase, etc. depending on use case | | 🌐 Multilingual support | Preprocess text and support locale detection via Slack metadata |