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
Commits on main
Studio Images: another account's loaded model no longer offers Generate (#16200)
7d5c612 Debojyoti Banerjee committed 19 days ago
Studio: keep <function=...> tool-call parsing linear on long output (#16105)
cc9525a Daniel Han committed 19 days ago
Studio: keep transcript listing rows while their files are unchanged (#16363)
aae8e4b Daniel Han committed 19 days ago
Studio: long chats stop re-rendering source badges per token and tool cards per run (#16277)
edde292 Daniel Han committed 19 days ago
Studio: Images and Video strips stop re-rendering every card per thumbnail and page (#16311)
a9f51af Daniel Han committed 19 days ago
Studio: read each gallery PNG's recipe once while it is unchanged and resolve only symlinked clips (#16355)
649cd14 Daniel Han committed 19 days ago
Give Claude its shell back when the local agent can't run a command (#16211)
36464c2 Nilay committed 19 days ago
fix(studio): report why a model's security checks blocked it (#16292)
9684fb3 Long Yixing committed 19 days ago
Studio: reconnect a chat's MCP session after the server restarts (#16209)
8af61f2 Nilay committed 19 days ago
Tests: stub the new num_proc helpers in the no-torch venv, and read with utf-8 in the metadata cache test (#16386)
277f84d Daniel Han committed 19 days ago
Studio CI: run the web tools overlap timings serially (#16383)
c58bf7f Daniel Han committed 19 days ago
Studio: reuse each repo's partial verdict in the local model listing (#16199)
ea69b9c Daniel Han committed 19 days ago
Studio: Appearance > Code with a live preview, line numbers, word wrap and ligatures (#16299)
6abeedc Michael Han committed 19 days ago
Studio: greeting toggle in Appearance, and the composer stays put when the greeting is off (#16301)
fa39948 Michael Han committed 19 days ago
Studio: choose where the background image shows (new chat page, chats, or all pages) (#16297)
576951b Michael Han committed 19 days ago
Studio: run diffusion dataset upload file work on a worker thread (#16361)
d148c97 Daniel Han committed 19 days ago
Studio: list the audio gallery from one folder scan and reuse unchanged sidecars (#16367)
0b2ea47 Daniel Han committed 19 days ago
Studio: reuse the BOS probe encoding instead of tokenizing a long prompt twice (#16289)
a88db1e Daniel Han committed 19 days ago
Studio: decode external provider images off the event loop (#16362)
9d1a085 Daniel Han committed 19 days ago
Studio export: save MLX merged and base pushes once into a reused folder (#16278)
a3446e8 Daniel Han committed 19 days ago
Studio: stop re-stripping the whole turn per token once tool markup is open (#16118)
c45c140 Daniel Han committed 19 days ago
Studio video: decode reference audio once per request and stage H3 reference frames faster (#16324)
e023bac Daniel Han committed 19 days ago
Studio: cut repeated work in the local model listing (#16146)
d8ae3f9 Daniel Han committed 19 days ago
Studio model picker: On Device rows re-render only when their own data changes (#16316)
c802f93 Daniel Han committed 19 days ago
Studio diffusion: change LoRA weights without reloading the adapter files (#16310)
cd14d76 Daniel Han committed 19 days ago
Studio: background fit and position controls, outline strength slider, composer setting in Appearance, audio and library polish (#16287)
c510fc1 Michael Han committed 19 days ago
Studio: copy tool schemas without deepcopy in the tool loop controller (#16306)
446155b Daniel Han committed 19 days ago
Lean varlen: ask torch 2.15's backend selector instead of the removed _can_use_cudnn (#16288)
f3d5b72 Daniel Han committed 19 days ago
FP8: float32 dequant before torch 2.14, no direct path under mismatched autocast (#15927)
b068981 Daniel Han committed 19 days ago