Answer code of conduct questions in Slack with GPT-4 & RAG technology

Name: Answer code of conduct questions in Slack with GPT-4 & RAG technology
Availability: InStock
Rating: 4.5 (12 reviews)
Author: Trung Tran

$20/month : Unlimited workflows

2500 executions/month

Try free

THE #1 IN WEB SCRAPING

Scrape any website without limits

Try free

HOSTINGER 🎉 Early Black Friday Deal
DISCOUNT 20%

Self-hosted n8n

Unlimited workflows - from $4.99/mo

Try free

#1 hub for scraping, AI & automation

6000+ actors - $5 credits/mo

Try free

📘 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

Trung Tran

0 workflows

Nodes

set gmail telegram agent google-gemini

Complexity

advanced

Published 22 Aug 2025

Likes 0

View on n8n.io Download Workflow

✨

Share Your Workflow

Have a great workflow to share? Join the n8n Creator Hub and help the community!

Submit Your Template How to Submit

Related Workflows

Draft and manage academic research papers with GPT-4 and Pinecone

## How It Works This workflow automates academic research processing by routing queries through specialized AI models while maintaining contextual memory. Designed for researchers, faculty, and graduate students, it solves the challenge of managing multiple AI models for different research tasks while preserving conversation context across sessions. The system accepts research queries via webhook, stores them in vector databases for semantic search, and intelligently routes requests to appropriate AI models (OpenAI, Anthropic Claude, or NVIDIA NIM). Results are consolidated, formatted, and delivered via email with full citation tracking. The workflow maintains conversation history using Pinecone vector storage, enabling follow-up queries that reference previous interactions. This eliminates manual model switching, context loss, and repetitive credential management—streamlining research workflows from literature review to hypothesis generation. ## Setup Steps 1. Configure Pinecone credentials 2. Add OpenAI API key for GPT-4 access and embeddings 3. Set up Anthropic Claude API credentials for advanced reasoning 4. Configure NVIDIA NIM API key for specialized academic models 5. Connect Google Sheets for query logging and result tracking 6. Set Gmail OAuth credentials for automated result delivery 7. Configure webhook URL for query submission endpoint ## Prerequisites Active accounts and API keys for Pinecone, OpenAI ## Use Cases Literature review automation with semantic paper discovery. ## Customization Modify AI model selection logic for domain-specific optimization. ## Benefits Reduces research processing time by 60% through automated routing.

View

Qualify and email literary agents with GPT‑4.1, Gmail and Google Sheets

## Inspiration & Notes This workflow was born out of a very real problem. While writing a book, I found the process of discovering suitable literary agents and managing outreach to be manual, and surprisingly difficult to scale. Researching agents, checking submission rules, personalizing emails, tracking submissions, and staying organized quickly became a full-time job on its own. So instead of doing it manually, I automated it. I built this entire workflow in **3 days** — and the goal of publishing it is to show that you can do the same. With the right structure and intent, complex sales and marketing workflows don’t have to take months to build. --- ## Contact & Collaboration If you have questions, business inquiries, or would like help setting up automation workflows, feel free to reach out: 📩 **[email protected]** I genuinely enjoy designing workflows and automation systems, especially when they support meaningful projects. I work primarily from interest and impact rather than purely financial motivation. Whether I take on a project for **FREE** or paid for the following reasons: - I **LOVE** setting up workflows and automation. - I work for **meaningfulness**, not for money. - **I may do the work for free**, depending on how meaningful the project is. If the problem statement matters, the motivation follows. - **It also depends on the value I bring to the table** -- If I can contribute significant value through system design, I’m more inclined to get involved. If you’re building something thoughtful and need help automating it, I’m always happy to have a conversation. Enjoy~! --- # 0. Overview Automates the end-to-end literary agent outreach pipeline, from data ingestion and eligibility filtering to deep agent research, personalized email generation, submission tracking, and analytics. ## Architecture The system is organized into four logical domains: The system is modular and is divided into four domains: --> Data Engineering --> Marketing & Research --> Sales (Outreach) --> Data Analysis Each domain operates independently and passes structured data downstream. --- ## 1. Data Engineering **Purpose:** Ingest and normalize agent data from multiple sources into a single source of truth. **Inputs** - Google BigQuery - Azure Blob Storage - AWS S3 - Google Sheets - (Optional) HTTP sources **Key Steps** - Scheduled ingestion trigger - Merge and normalize heterogeneous data formats (CSV, tables) - Deduplication and validation - AI-assisted enrichment for missing metadata - Append-only writes to a central Google Sheet **Output** - Clean, normalized agent records ready for eligibility evaluation --- ## 2. Marketing & Research **Purpose:** Decide *who* to contact and *how* to personalize outreach. ### Eligibility Evaluation An AI agent evaluates each record against strict rules: - Email submissions enabled - Not QueryTracker-only or QueryManager-only - Genre fit (e.g. Memoir, Spiritual, Self-help, Psychology, Relationships, Family) **Outputs** - `send_email` (boolean) - `reason` (auditable explanation) ### Deep Research For eligible agents only: - Public research from agency sites, interviews, Manuscript Wish List, and LinkedIn (if public) - Extracts: - Professional background - Editorial interests - Genres represented - Notable clients/books (if publicly listed) - Public statements - Source-backed personalization angles **Strict Rule:** All claims must be explicitly cited; no inference or hallucination is allowed. --- ## 3. Sales (Outreach) **Purpose:** Execute personalized email outreach and maintain clean submission tracking. **Steps** - AI generates agent-specific email copy - Copy is normalized for tone and clarity - Email is sent (e.g. Gmail) - Submission metadata is logged: - `Submission Completed` - `Submission Timestamp` - Channel used **Result** - Consistent, traceable outreach with CRM-style hygiene --- ## 4. Data Analysis **Purpose:** Measure pipeline health and outreach effectiveness. **Features** - Append-only decision and submission logs - QuickChart visualizations for fast validation (e.g. TRUE vs FALSE completion rates) - Optional integration with: - Power BI - Google Analytics 4 **Supports** - Completion rate analysis - Funnel tracking - Source/platform performance - Decision auditing --- ## Design Principles - **Separation of concerns** (ingestion ≠ decision ≠ outreach ≠ analytics) - **AI with hard guardrails** (strict schemas, source-only facts) - **Append-only logging** (analytics-safe, debuggable) - **Modular & extensible** (plug-and-play data sources) - **Human-readable + machine-usable outputs** --- ## Constraints & Notes - Only public, professional information is used - No private or speculative data - HTTP scraping avoided unless necessary - Power BI Embedded is not required - Workflow designed and implemented end-to-end in ~3 days --- ## Use Cases ### Marketing - Audience discovery - Agent segmentation - Personalization at scale - Campaign readiness - Funnel automation ### Sales - Lead qualification - Deduplication - Outreach execution - Status tracking - Pipeline hygiene --- ## Tech Stack - **Automation:** n8n - **AI:** OpenAI (GPT) - **Scripting:** JavaScript - **Data Stores:** Google Sheets - **Email:** Gmail - **Visualization:** QuickChart - **BI (optional):** Power BI, Google Analytics 4 - **Cloud Sources:** AWS S3, Azure Blob, BigQuery --- ## Status This workflow is production-ready, modular, and designed for extension into other sales or marketing domains beyond literary outreach. ---

View

Create Bosta shipping orders from Odoo invoices using OpenAI GPT models

## What Problem Does It Solve? - **Manual Data Entry Bottlenecks:** Moving shipping data from Odoo to Bosta manually is slow and prone to errors, especially during high-volume periods. - **Address Mismatches:** Odoo stores addresses as unstructured text, while Bosta requires strict Zone/District IDs. Mismatches lead to failed deliveries and returns. - **Messy Labels:** Long ERP product names look unprofessional on shipping labels. - **This workflow solves these by:** - Instantly creating the shipping order in Bosta when an Odoo invoice is confirmed. - Using an **AI Agent** to intelligently parse raw addresses and map them to the exact Bosta ID. - Ensuring the high operational standards required by automating data cleaning and COD rounding. ## How to Configure It ### 1. Odoo Setup - Create an Automation Rule in Odoo that sends a POST request to this workflow's Webhook URL when an invoice state changes to "Confirmed". ### 2. Credentials - Connect your **Odoo**, **OpenAI**, and **Telegram** accounts in the respective n8n nodes. - Add your **Bosta API Key** in the Header parameters of the `Create Bosta Order` node. ### 3. Product Mapping - Open the `Summarize Items` code node and update the `NAME_MAP` object to link your Odoo product names to short shipping labels. ### 4. Data Table - Ensure the `Fetch Zones` node is connected to your Bosta Zones/Districts data table in n8n. ## How It Works - **Trigger:** The workflow starts automatically when an invoice is confirmed in Odoo. - **Fetch & Process:** It pulls customer details and invoice items, then aggregates quantities (e.g., turning 3 lines of "Shampoo" into "Shampoo (3)"). - **AI Analysis:** The AI Agent cross-references the raw address with the official Bosta zones list to strictly select the correct IDs. - **Execution:** The order is created in Bosta. If successful, the process is complete. - **Error Handling:** If any step fails, a Telegram message is sent immediately with the invoice number to alert the **operations team**. ## Customization Ideas - **Write Back:** Add a node to update the Odoo invoice with the generated Bosta tracking number. - **Multi-Courier:** Add a switch node to route orders to different couriers (e.g., Aramex, Mylerz) based on the city. - **Campaign Logging:** Log successful shipments to a spreadsheet to track fulfillment metrics. - **Notification Channels:** Change the error alert from Telegram to Slack or Email. If you need any help [Get in Touch](https://www.linkedin.com/in/abdallaelshikh0/)

View

👨‍💻

Need Custom Automation?

N8N Automation Expert

Specialized in N8N automation, I design custom workflows that connect your tools and automate your processes.