Auto Categorize Outlook Emails with AI

• Email Filtering: Processes only unread, unflagged emails.
• Content Sanitization: Removes HTML tags and extraneous information from email bodies.
• AI-Powered Categorization: Utilizes an LLM to determine the most suitable category (Action, Junk, Receipt, SaaS, Community, Business, Other) and an optional subcategory.
• Dynamic Category Application: Applies generated categories to Outlook emails.
• Automated Folder Management: Moves categorized emails to designated Outlook folders.
• Error Handling: Includes a mechanism to catch and log errors during processing.

1. Trigger: The workflow starts when the 'Test workflow' button is clicked (manual trigger). In a real scenario, this would likely be an Outlook trigger.
2. Email Retrieval: The Microsoft Outlook23 node fetches unread, unflagged emails from a specified folder.
3. Email Sanitization: The Markdown1 and varEmal1 nodes clean up the email body, removing HTML and preparing it for AI processing.
4. Category Assignment Setup: The varID & Category1 node defines potential categories. This node might need customization based on your specific needs.
5. AI Categorization: The AI Agent1 node, connected to an Ollama Chat Model, analyzes the email content and assigns a category and subCategory based on the prompt and system message.
6. JSON Conversion: The varJSON1 node ensures the AI's output is parsed as JSON, with error handling.
7. Category Application & Moving: A series of Microsoft Outlook nodes (100db1cb, 524f6be3, etc.) are used to apply the determined categories and move the emails to specific folders based on the AI's categorization.
8. Read Status Check: The If1 node checks if the email has been read, and if so, moves it to the 'Actioned' folder.

Important Notes:

• Folder IDs: Ensure the folderId values in the Microsoft Outlook nodes are correctly configured for your Outlook account.
• Category Names: The categories defined in varID & Category1 and the system message in AI Agent1 should be consistent.
• AI Model: The Ollama Chat Model1 is configured to use the 'qwen2.5:14b' model. Adjust this if you use a different model.
• Filters: The Microsoft Outlook23 node uses a custom filter: flag/flagStatus eq 'notFlagged' and not categories/any(). You can modify this to adjust which emails are processed.