Merge pull request #45 from amd/n8n-playbook

n8n playbook

Author: danielholanda

playbooks/core/n8n-automation-gpt-oss/README.md

@@ -1,3 +1,146 @@
-# Automating Workflows with n8n and gpt-oss-120b
+## Overview
 
-<!-- Playbook content goes here -->
+n8n is a workflow automation platform that lets you connect apps and services using a visual node-based editor. Instead of writing code, you build workflows by dragging, connecting, and configuring nodes.
+
+This playbook teaches you how to set up an AI-powered financial news summarizer that scrapes the Financial Times website, extracts key headlines, and uses a local LLM running on your Ryzen™ AI Halo to generate an investor-focused summary.
+
+## What You'll Learn
+
+- How to install and launch n8n
+- Importing and configuring a pre-built workflow
+- Connecting to Lemonade using the native n8n integration
+- Understanding workflow nodes and data flow
+
+## Why Lemonade?
+
+[Lemonade](https://lemonade-server.ai) is a local LLM serving platform built for AMD hardware. It provides an OpenAI-compatible API that runs entirely on your machine—your data never leaves your device.
+
+In this playbook, we use Lemonade to serve a local LLM that n8n connects to for AI-powered tasks. Unlike cloud-based LLM providers, Lemonade runs on your Ryzen™ AI Halo GPU, giving you full control over your models with zero API costs.
+
+n8n includes a **native Lemonade node** (`Lemonade Chat Model`) that provides a first-class integration—no need to configure OpenAI-compatible endpoints manually. This makes connecting your local LLM to automation workflows straightforward.
+
+## Prerequisites
+
+<!-- @require:lemonade,nodejs -->
+
+## Installing n8n
+
+Your STX Halo has Node.js pre-installed. Install n8n globally using npm:
+
+```bash
+npm install -g n8n
+```
+
+## Launching n8n
+
+Start n8n from the terminal:
+
+```bash
+n8n start
+```
+
+n8n starts a local web server. Open your browser to `http://localhost:5678` to access the editor.
+
+> **Tip**: Keep the terminal window open while using n8n. Closing it will stop the server.
+
+## Setting Up the Workflow
+
+### Step 1: Sign Up or Log In to n8n
+
+When you first open n8n, you'll be prompted to create an account or log in:
+
+1. Open `http://localhost:5678` in your browser
+2. Create a new account with your email, or log in if you already have one
+3. Once logged in, you'll see the n8n dashboard
+
+### Step 2: Import the Workflow
+
+We've provided a pre-built workflow that you can import directly:
+
+1. Download the workflow file: [financial-news-workflow.json](assets/financial-news-workflow.json)
+2. If this is your first workflow, click **Start from Scratch** to open the workflow editor. Otherwise, click **Add workflow** in the top right.
+3. Click the **...** menu (three dots) in the top bar and select **Import from file**
+4. Select the downloaded `financial-news-workflow.json` file
+5. The workflow will appear on the canvas
+
+### Step 3: Understanding the Workflow
+
+The imported workflow contains 7 connected nodes:
+
+<p align="center">
+  <img src="assets/workflow-overview.png" alt="n8n Financial News Workflow" width="800"/>
+</p>
+
+| Node | Purpose |
+|------|---------|
+| **When clicking 'Execute workflow'** | Manual trigger to start the workflow |
+| **Fetch Financial News Webpage** | HTTP GET request to `https://www.ft.com/` |
+| **Delay to Ensure Page Load** | Wait node to ensure page content is fully loaded |
+| **Extract News Headlines & Text** | HTML node that extracts headlines, editor's picks, top stories, and regional news using CSS selectors |
+| **Clean Extracted News Data** | Set node that combines all extracted data into a single text field |
+| **AI Financial News Summarizer** | AI Agent that processes the news with a financial analyst system prompt |
+| **Lemonade Chat Model** | Connects to your local Lemonade server running the LLM |
+| **Structured Output Parser** | Formats the AI output as structured JSON |
+| **Convert to File** | Converts the summary to a downloadable file |
+
+### Step 4: Configure Lemonade Credentials
+
+Before running the workflow, you need to connect it to your local Lemonade server:
+
+1. Double click the **Lemonade Chat Model** node
+2. In the dropdown menu **Credential to connect with** select **Create New Credential**
+3. Enter the following settings:
+
+| Field | Value |
+|-------|-------|
+| **Base URL** | `http://localhost:8000/api/v1` |
+| **API Key** | `lemonade` |
+
+4. Click **Save**
+
+> **Note**: Ensure Lemonade is running before testing. The workflow uses `gpt-oss-120b-mxfp4-GGUF` by default—you can change this in the Lemonade Chat Model node settings.
+
+### Step 5: Test the Workflow
+
+1. Ensure Lemonade is running with a model loaded
+2. Click **Execute workflow** at the bottom center of the canvas
+3. Watch each node execute from left to right—they turn green when complete
+4. Click the **AI Financial News Summarizer** node to see the generated summary
+
+## Understanding the AI Agent
+
+The AI Financial News Summarizer uses a system prompt designed for financial analysis:
+
+```
+You are an AI financial analyst. Your role is to read, understand, and
+summarize key financial news from today. The goal is to provide investors
+with a clear and concise market overview to support better investment decisions.
+
+Investor Outlook
+Today's news points to [bullish/bearish/neutral] sentiment. Watch for
+[economic event/earnings report] tomorrow, which could influence market direction.
+```
+
+The agent receives the cleaned news data and outputs a structured summary with market sentiment.
+
+## Saving Your Workflow
+
+Click the workflow name at the top and rename it if desired. Workflows auto-save as you work.
+
+## Next Steps
+
+- **Schedule automation**: Replace Manual Trigger with a **Schedule Trigger** to run daily
+- **Send notifications**: Add a **Discord**, **Slack**, or **Email** node to receive summaries
+- **Try different models**: Change the model in the Lemonade Chat Model node to experiment with different LLMs
+- **Customize extraction**: Modify the HTML Extract node's CSS selectors to target different news sections
+- **Try different backends**: n8n also supports [Ollama](https://n8n.io/workflows/?integrations=Ollama+Chat+Model), LM Studio, and other local LLM backends
+
+### Explore n8n Templates
+
+n8n has hundreds of pre-built workflow templates. Browse the official template library at:
+
+**[https://n8n.io/workflows/](https://n8n.io/workflows/)**
+
+Search for "AI", "LLM", or "automation" to find workflows you can import and customize.
+
+For more information, check out the [n8n Documentation](https://docs.n8n.io/).