Ideas Home

Simple web site about complex ideas.

View the Project on GitHub akrafts-gpt/simple-static-site

Email to LLM Pipeline

Inbox → Insight: Email Forwarding & LLM-Powered Event Extraction

Transforming scattered event notifications into structured suggestions is at the heart of our discovery engine. This article dives into how a simple email-forwarding workflow and an LLM-based parsing pipeline deliver privacy-safe, high-quality event data.

Forward-to-Service Workflow

  1. User Action:
    • The user forwards any event-related email (from newsletters, ticket platforms, venue announcements) to a dedicated service address.
    • No full inbox access is required, ensuring user privacy.
  2. Privacy & Permissions:
    • Forwarded emails are processed only for event extraction and then discarded.
    • Users can revoke forwarding at any time, and no personal metadata beyond the event content is retained.

LLM Parsing Pipeline

Once an email arrives, our AI pipeline performs:

Schema Mapping & Data Normalization

All extracted fields flow into a consistent event schema:

Field Example
title “Summer Jazz Festival”
start_time 2025-08-15T19:00:00+02:00
end_time 2025-08-15T22:30:00+02:00
location “Stadsteatern, Stockholm, SE”
categories [“music”, “festival”]
image_url https://…/jazzfest-thumb.jpg

This normalized structure feeds directly into our calendar UI and recommendation engine.

Example: Email → Structured Event

Raw Email Snippet:
“Join us for the Summer Jazz Festival on August 15 at Stadsteatern. Doors open at 7 PM…”

Parsed Output:

{
  "title": "Summer Jazz Festival",
  "start_time": "2025-08-15T19:00:00+02:00",
  "end_time": "2025-08-15T22:30:00+02:00",
  "location": "Stadsteatern, Stockholm, SE",
  "categories": ["music", "festival"],
  "image_url": "https://…/jazzfest-thumb.jpg"
}

Next Steps: Scaling & Integration


← Back to Overview