Automation & AI agentsQuality Watchdog — a guard that reads the automations themselves
Quality Watchdog — a guard that reads the automations themselves
A watchdog that reads how the automations are built rather than the run log, and catches what breaks silently.
How it's wired
Drawn from the live workflow — every node and every arrow is really there. The labels are translated from the originals.
What it is
Ordinary monitoring looks at the run log and answers the question “did it run”. But automations have a class of failures the log answers “yes” to: a step that swallowed its own error and carried on; a branch of a condition that has never once fired; a message going to a chat everybody has left. Formally everything is green; in fact the work has not been done for weeks.
So the watchdog leaves the log alone and reads the structure of the scripts themselves: how they are assembled, what is connected to what, where an error handler is empty, where each branch leads. This is structural analysis, and that is what finds the things no single run reveals.
Specifications: Node.js, nine separate detectors, each for its own kind of silent failure; the analysis runs over an export of the scripts from n8n, and the report arrives in Telegram. The very first pass took the urgent problems from twenty-three down to one in a day.
Nearby work
Zero — an offline voice assistantA voice assistant that wakes on its name and thinks on the laptop’s graphics card: no internet, no login, an answer in 0.3–0.6 seconds.My own automation serverMy own server carrying over a hundred automations around the clock: bots, digests, finance, reminders, nightly backups, app synchronisation.Svita — seven personal AI agentsSeven personal AI agents with a shared memory: each has its own Telegram bot and its own role, but what you tell one, the others know.