POPFile

Bayesian email classifier written in Perl. Acts as a proxy between mail clients and mail servers (POP3, SMTP, NNTP), inserting an X-Text-Classification: header with the predicted category. Users correct misclassifications through the web UI, which trains the classifier over time.

Prerequisites

• Docker + Compose

Run

docker compose up

Local (with perlbrew + Carton)

carton install
carton exec perl popfile.pl

The web UI is available at http://localhost:8080 by default.

POPFILE_ROOT overrides the root directory for config, database, and message cache (default: ./).

CLI utilities

# Classify a message and show word scores
carton exec perl bayes.pl <message-file>

# Train a message into a bucket
carton exec perl insert.pl <bucket-name> <message-file>

Svelte UI

The frontend lives in ui/ and builds to public/.

cd ui && npm install

# Development (hot-reload, proxies /api to localhost:8080)
npm run dev

# Production build → public/
npm run build

Tests

carton exec perl -I. t/mailparse.t

Runtime files (gitignored)

Path	Description
popfile.cfg	Generated configuration
popfile.db	SQLite database (corpus + history)
messages/	Cached message files
*.pid, *.log	Runtime state