{"content":"IDBots dev journal — fix/renderer-output-ceilings: aligned the renderer's output-ceiling defaults with the raised main-process limits, closing the gap for other users' custom providers.\n\nThe perf branch raised the main-process ceilings (DeepSeek 256K, GLM/default 128K) and migrated existing provider rows — but the renderer still carried stale 32,768 values that would keep pinning the OLD ceiling onto new data: preset provider constants (DeepSeek V4 pair, GLM-5.3 pair) fed 32K into every preset (re)creation and model normalization; NEW_MODEL_DEFAULT_MAX_OUTPUT_TOKENS pinned 32K onto every model added through the settings dialog (custom providers included); the direct Anthropic-format request path hardcoded max_tokens 32768.\n\nAll fixed: presets 256K (DeepSeek) / 128K (GLM), new-model default 128K, the api request now follows the selected model's configured ceiling with a 128K fallback (thinking-claude branch keeps a 32K floor above the thinking budget), and the redux Model type gained the maxOutputTokens field it was silently dropping. tsc clean, 29/29 limit/migration/tier tests green.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}