Tom AI now confirms fixes work before claiming victory
Posted by tomspark
SparkBox 1.6.639 fixes a real problem I've been tracking: Tom AI was running fixes that didn't apply to your box, burning your fix count and reporting false positives. I rewired it to measure first, then act only when the measurement points at that exact issue. More importantly, after it runs a fix — especially the qBittorrent password re-sync — it now asks Sonarr, Radarr, and your other apps to test the login themselves. If they all pass, it says "fixed". If it can't verify, it says "unconfirmed", never guessed. When Tom AI can't solve it, it now files a support case for our team with everything it measured: your checkup findings, app health, the download queue, how your media is wired. You get a case number to reference. The case auto-deletes after 90 days and your conversation history never leaves your box. I also gave Tom AI better eyes into why your download clients fail — it can now read Sonarr's and Radarr's own health messages instead of guessing — and locked down the checkup so it can only claim facts it actually measured from your dashboard. Highlights: • Fixes only run when measurement proves they apply to your box • Tom AI confirms app logins actually work after the fix • Support cases are auto-filed when Tom AI can't solve it, with full context • Better diagnostics: Tom AI reads app health messages directly • Fleet-wide tracking of which AI model ran each fix, so we measure if fixes are getting better Check the Updates tab in your dashboard — it'll show you the version and any pending fixes. If something goes wrong or doesn't work as described, post here in d/sparkbox. --- Update from your SparkBox dashboard's Updates tab, or see the full release notes on the Releases page.