Email Extraction Demo

Click the pencil on a row to edit its Title and Text as Markdown. Click the pencil again (or elsewhere) to view the rendered result — links are clickable in view mode. The triangle expands the screenshot. Click Save (Ctrl+S) to write back to a21_Email-Extraction-Demo.json.
Phase 2 and 3 Tasks

The document defines the Tasks for Phase 2 and Phase 3 of the project. Only a sample is shown below. Phase 1 was defined in a Gemini browser chat "discovery" session. The session was then saved as a markdown file and given to Amazon Q (now Kiro) to write a Plan, Spec and this Tasks document.

Phase 4 Specs

This document defines the specs for the final phase of the project. You can read the full document here if you'd like. It was written by Amazon Q based on the Plan document it wrote for this phase.

Phase 4 Tasks

This document defines the Tasks for Phase 4 of the project. The Specs and Tasks documents were then read by Amazon Q as a guide to writing the code for the client and server apps. 

GMail PGE folders

The process starts by me searching for PGE emails in my GMail app (browser or desktop) and moving them into a PGE folder. The server app that Q wrote queries my emails and folder via a Google API.

Run App 21

This screen shot show how we start the server app. Normally we have a separate client app, but the one HTML page is delivered by the server app.  Lots of local /api endpoints exist for the server to connect with Google's GMail and Drive APIs, as well as an Openrouter model.  We put all non-secret variables in _config.js. You can view it here. 

Client Control App

Once the server app is running, this simple dashboard allows me to authenticate my Google account, and start the sync process that looks for new GMail messages in a PGE folder. It the parses them and saves EML, IMG, HTML and PDFs or other attachment files to a folder in Google Drive.

I can then extract metadata into a SQLite DB, update a detailed Google Sheet and view a summary Sheet.

SQLite DB Records

Below is a detailed spreadsheet listing the data records saved in a local SQL database. The AI Model has extracted 6 types of metadata from the emails:

  • Disputes: Summaries of each email
  • Parties: From, To or CC'd Names in the emails
  • Events: Meetings, Phone calls, Court Dates, etc. in the emails
  • Emails: Subjects of all emails in the email thread
  • Attachments: All Attachments in the emails
Google Drive Folder

The primary folder contains HTML versions of the emails for easy viewing. The other three types of files are saved in their own folders:

  • Messages: .eml
  • Attachments: .pdfs and other attachments
  • Assets: .png, .jpg
Saved HTML Email

The same emails can appear in the threads of many emails. Each HTML version of an email has it original header information with the entire thread from there. They along with attachments are de-duped in the last consolidate step.

Summary Spreadsheet

This Consolidated Google sheet was actually created by Hermes. It has collapsible rows for each of the five types of metadata data. It provides a history of everything related to the PGE legal case.