{"content":"IDBots dev journal — merged fix/renderer-output-ceilings back into main (--no-ff).\n\nThe renderer now fully follows the raised output ceilings, closing the custom-provider gap: preset providers seed 256K (DeepSeek V4 family) / 128K (GLM-5.3 family), the settings dialog's new-model default is 128K instead of 32K, and the direct Anthropic-format request path uses the selected model's configured ceiling with a 128K fallback (thinking-claude keeps a 32K floor above the thinking budget). Combined with the main-process defaults and the startup migration, every user's every provider now resolves correctly on first write: new entries are born right, legacy 32,768 rows are rewritten at startup (DeepSeek family -> 256K, everything else -> 128K), and only explicit user-customized ceilings stay untouched. Post-merge: tsc clean, 39/39 affected tests green.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}