Files

2.7 KiB

DRE Mail Poller Pattern

Session: July 8, 2026

Overview

Lightweight no-LLM IMAP poller that checks multiple DRE mailboxes on MXroute, parses headers + body, and writes a rolling JSON database for a web portal.

Key Files

File Purpose
/root/.hermes/scripts/dre-mail-poller.py Main poller script
/var/www/internal/data/dre-mails.json Output JSON database
/var/log/dre-mail-poller.log Log file
/etc/cron.d/dre-mail-poller Cron config (every minute)

IMAP Server

  • Host: heracles.mxrouting.net:993
  • Mailboxes:
    • dre@debtrecoveryexperts.com / M8ke.Money.Honey!
    • collections@debtrecoveryexperts.com / M8ke.Money.Honey!

The password is shared across both mailboxes — the MXroute cPanel config uses the same password for all mailboxes under that domain.

Script Architecture

MAILBOXES = [
    {"label": "dre", "user": "dre@debtrecoveryexperts.com", "pass": "M8ke.Money.Honey!"},
    {"label": "collections", "user": "collections@debtrecoveryexperts.com", "pass": "M8ke.Money.Honey!"},
]

Each mailbox iterates independently. A connection failure on one doesn't block the other.

Deduplication

Composite key: {label}:{IMAP_UID}. UIDs are stable per-mailbox in IMAP. Existing IDs loaded from JSON file into a set() on each run; new messages appended only when id not in set.

Output schema

{
  "id": "dre:12345",
  "mailbox": "dre",
  "from": "John Doe <john@example.com>",
  "to": "dre@debtrecoveryexperts.com",
  "subject": "Invoice #123",
  "date": "2025-07-08T09:38:00+00:00",
  "body_preview": "First 500 chars...",
  "body": "Full plain-text body...",
  "is_read": false,
  "claim_match": "DRE-2025-0001"
}

claim_match is extracted via re.search(r'DRE-\d{4}-\d{4}', subject).

Caddy integration

No Caddy changes needed — the existing internal.debtrecoveryexperts.com block already has root * /var/www/internal and file_server. The JSON is served at /data/dre-mails.json without any extra config.

Pitfalls

  • Connection timeouts: MXroute can be slow (1-2s per connect). Script uses imaplib.IMAP4_SSL with default timeout which is usually sufficient.
  • Large bodies: Some emails have megabytes of HTML. Body extraction with stdlib is fast but the rolling DB cap (500 entries) prevents unbounded growth.
  • UNSEEN only: The poller fetches UNSEEN and never marks messages as SEEN. This means the same mail appears on every poll until a real email client reads it. The dedup set prevents duplicate entries in the JSON.
  • Body extraction order: text/plaintext/html → any text/*. HTML-only messages (common from Apple Mail replies) are handled by the HTML fallback.