📧 Email Recap Tool (Local / Offline)

Automatically reads .eml files from a folder, analyzes each email using a local Qwen3 model, and exports a clean summary to Excel.

100% offline — no data leaves your machine.

What It Does

• Parses .eml files (sender, subject, date, body)
• Uses a local Qwen3 model via LM Studio to extract keywords and write a short summary
• Exports everything to a formatted .xlsx file

Output Columns

Column	Description
Mailbox	Email address of the sender
Subject	Email subject
Date	Date sent
Keywords	Important keywords extracted by AI
Summary	Short 1–2 sentence summary of the email

---

Requirements

• macOS (tested on MacBook Air M2)
• Python 3.10+
• LM Studio installed with a Qwen3 model downloaded

---

Setup

1. Install LM Studio & download the model

1. Download LM Studio
2. Open LM Studio → Search for Qwen3-8B (recommended for M2)
3. Download the Q4_K_M variant (good balance of speed vs quality on M2)
4. Load the model → go to Local Server tab → click Start Server

LM Studio will serve the model at http://localhost:1234

2. Set the model name in the code

Open src/ai_analyzer.py and update this line to match the model name shown in LM Studio:

MODEL_NAME = "qwen3-8b"  # ← change this to match your LM Studio model name

3. Install Python dependencies

cd email-recap
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

---

How To Run

Step 1 — Put your .eml files into data/emails/

data/
└── emails/
    ├── invoice_june.eml
    ├── meeting_notes.eml
    └── ...

Step 2 — Make sure LM Studio server is running

LM Studio → Local Server tab → Start Server (green button)

Step 3 — Run the script

python src/main.py

Step 4 — Open the output

outputs/email_recap.xlsx

---

Project Structure

email-recap/
├── README.md               # This file
├── .gitignore
├── requirements.txt        # openai, openpyxl
├── src/
│   ├── main.py             # Entry point — runs the full pipeline
│   ├── eml_reader.py       # Parses .eml files into structured data
│   ├── ai_analyzer.py      # Calls local Qwen model for keywords + summary
│   └── excel_writer.py     # Writes results to formatted .xlsx
└── data/
    └── emails/             # Drop your .eml files here

---

Example Output

Mailbox	Subject	Date	Keywords	Summary
vendor@supplier.com	Invoice #1042	2024-06-01	invoice, payment, due date	Vendor sent invoice #1042 for $2,400 due by June 15.
hr@company.com	Annual Leave Policy	2024-06-03	leave, policy, HR, approval	HR updated the annual leave policy, requiring 2-week prior approval.

---

Notes

• Recommended model for M2: qwen3.5-9b* (~5GB RAM, fast enough for batch processing)
• Email body is trimmed to 3000 characters before sending to the model (saves processing time)
• Qwen3's "thinking mode" output (<think>...</think>) is automatically stripped
• No API key, no internet connection needed