{"content":"IDBots development journal — commit cd392672\n\nBranch: feat/studio-mode-optimization\nCommit: perf: reduce Studio idle memory and director load\n\nReduced Studio's contribution to host memory pressure and auxiliary-model load. The host was using about 10 GB of swap during investigation, so operating-system memory pressure is explicitly not claimed to be solved by this patch.\n\nIdle DSH runtime reaping is now memory-adaptive: healthy desktop sessions retain the normal 30-minute warm TTL (server mode retains its own default), while less than 1 GB of free system memory shortens the idle-slot TTL to two minutes. In-flight turns and runtimes with recent kernel activity remain protected, so this only accelerates cleanup of genuinely idle subprocesses and their automation resources. Reap logs now include the effective TTL and free-memory reading.\n\nGenerative Live Stage Director work is capped at four calls per turn instead of eight. After two consecutive failures it stops requesting for the rest of the turn and keeps deterministic scenes instead of amplifying an outage. Projection still cancels queued Director work as implemented in the preceding projection-state fix.\n\nVerification: 54 Studio tests passed; runtime drain and live DSH provider/model-switching integration passed; TypeScript and full production builds passed; targeted ESLint, DSH dependency checks, and all 11 kernel patch fingerprint checks passed. The branch remains local and unmerged, with no remote push.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}