# Email triage state file — critical post-migration check

The `~/.hermes/email_triage/state.json` file tracks processed IMAP UIDs. Without it, the IMAP triage cron job treats every inbox message as new on first run.

## Migration pitfall

**Symptom:** After a `~/.hermes/` migration, the hourly IMAP triage cron fails with `RuntimeError: Context length exceeded (X tokens). Cannot compress further.`

**Root cause:** `state.json` was not in the tarball (or was excluded by a custom exclude pattern). On first run, the script processes the entire inbox — often 3000+ messages — which blows the context window.

**Fix:**

1. Check if the old server still has the file:
   ```bash
   scp root@old-server-ip:/root/.hermes/email_triage/state.json /root/.hermes/email_triage/
   ```

2. If the standby box has it, grab it from there:
   ```bash
   scp root@standby-ip:/root/.hermes/email_triage/state.json /root/.hermes/email_triage/
   ```

3. If both copies are gone, initialize state by marking all old UIDs as processed:
   ```python
   import imaplib, ssl, json, datetime
   ctx = ssl.create_default_context()
   m = imaplib.IMAP4_SSL('mail.germainebrown.com', 993, ssl_context=ctx)
   m.login('g@germainebrown.com', open('/root/.config/himalaya/g-germainebrown.pass').read().strip())
   m.select('INBOX')
   status, data = m.uid('search', None, 'ALL')
   uids = data[0].split() if data[0] else []
   processed = {}
   if len(uids) > 50:
       recent = uids[-50:]
       old = uids[:-50]
       for uid in old:
           processed[uid.decode()] = {'summary': 'pre-migration-rebuild', 'decision': 'skip', 'timestamp': '2026-07-05T10:00:00Z'}
   state = {'processed_uids': processed, 'created_at': datetime.datetime.now(datetime.timezone.utc).isoformat()}
   # Write to ~/.hermes/email_triage/state.json
   ```

## Prevention

- Add `email_triage/state.json` and `email_triage/actions.jsonl` to the migration manifest.
- Verify these files exist before turning off the old server.