Queue-Based Note Processing Architecture | Generated by AI
Done. Here’s the corrected architecture:
Note Plugin — Queue-Based Architecture
What changed: The /note slash command no longer processes the response directly. Instead it enqueues the text content, and the separate ww note watch daemon picks it up.
Flow:
/note (Hermes CLI)
└→ enqueue text → ~/.config/ww/note_queue.json
│
┌───────────┴───────────┐
│ │
ww note watch cron: every 30m
(daemon, polls 3s) (safety net)
│ │
└───────────┬───────────┘
↓
ww note process
├─ LLM title gen
├─ write markdown file
├─ fix MathJax + tables
├─ git commit + push
└─ clear done from queue
What’s running:
ww note watch— daemon pid 133147 (started Jul 3, still active)- Hermes cron
ww-note-process— every 30m as safety net if daemon dies
Plugin (~/.hermes/plugins/note/__init__.py):
/note→_enqueue(text, "note")→ writes to~/.config/ww/note_queue.json/note 3→ enqueues 3rd response back/note --title "Foo"→ stores title hint in queue entry- No LLM calls, no file writes, no git in the plugin itself — just fast enqueue
Dependencies: ww package installed in Hermes venv, pyperclip also installed.