Local UI to run and train LLMs and diffusion models. Supports GGUF, MLX, Qwen3.8, DeepSeek-V4.1, MiniMax-H3, Gemma 4, FLUX and more.
Python agentaichatgptdeepseekfine-tuninggemmaimage-generationllamallmllmsopenaipythonqwenreinforcement-learningself-hostedstable-diffusiontext-to-speechttsuiunsloth
Re-measure the Studio startup budget after the audio selector and thinking merges (#13640)
Daniel Hancb6cbe85462f
diff --git a/studio/frontend/scripts/check-bundle-budget.ts b/studio/frontend/scripts/check-bundle-budget.tsindex 6b29382fa25..ed1cfd6923b 100644--- a/studio/frontend/scripts/check-bundle-budget.ts+++ b/studio/frontend/scripts/check-bundle-budget.ts@@ -75,8 +75,13 @@ export const BUDGET = {// is the token activity Chart view (#13558, +0.7 KB raw / +0.3 KB transfer): its persisted store// pulls zustand/middleware into the startup set as a 99th chunk. Raised by that growth so main's// margin stays where it was.- transferBytes: 2_023_500,- rawBytes: 6_734_000,+ // Six merges since add 2.9 KB raw / 1.4 KB transfer (6,570.7 -> 6,573.6 KB raw, 1,975.5 -> 1,976.9 KB+ // transfer from b8d76e1696 to 8af12324b0, one machine and build), with the chunk count unchanged and+ // no single import doing it. The largest step is the audio model selector fix (#13517, +2.9 KB raw /+ // +0.9 KB transfer); thinking traces (#13545), MLX precision folders (#13565) and vLLM thinking+ // effort (#13571) add the rest. Raised by that growth so main's margin stays where it was.+ transferBytes: 2_025_000,+ rawBytes: 6_737_000,};// The chunk count is reported but not budgeted. Splitting a page out of the entry raises it while lowering the