What's Changed
- fix(sunset): gate sunset mode on cloud deployment and keep PATs working through the wind-down by @AnishSarkar22 in #1815
- Release 2.0.2: move Latest to 2.x, bridge legacy updaters, permalink downloads by @MODSetter in #1816
- [Docs|Plugins] Plan the plugin system by @CREDO23 in #1817
- feat(local): replace Ollama with a bundled llama.cpp runtime, a curated GGUF catalog, and a hardware-aware fit estimator by @AnishSarkar22 in #1819
- feat(llm): llama.cpp runtime: warm loads, per-layer KV pricing, one Hugging Face consent, manifest order over ranks by @AnishSarkar22 in #1820
- [Docs|Models] Plan the model catalog by @CREDO23 in #1899
- [Docs|Models] Mark what the model catalog replaces by @CREDO23 in #1900
- refactor(llm): one model catalog, starting with the remote side by @CREDO23 in #1901
- feat(llm): rework the local model catalog around pinned builds and one row shape by @AnishSarkar22 in #1902
- feat(local-ui): rebuild model settings as per-type sections with local and server models grouped together by @AnishSarkar22 in #1904
- [LLM|Models] Serve image models from the local catalog by @CREDO23 in #1903
- feat(local): run sd-server in dev and give image models chat's download states by @AnishSarkar22 in #1905
- feat(local-ui): add chat and image model steps to onboarding, with server dialogs and a logo loader by @AnishSarkar22 in #1906
- [LLM|Audio] Voice podcasts on audio.cpp by @CREDO23 in #1907
- feat(local-ui): bring audio models in line with chat and image, and add an audio onboarding step by @AnishSarkar22 in #1908
- [LLM|Models] Ship the local model manifest in the Studio worker by @CREDO23 in #1910
- [LLM|Audio] Ship Kokoro as the default podcast voice by @CREDO23 in #1909
- feat(local): translate the desktop app into Japanese and German, open without a broken chat model, and add a dev app identity by @AnishSarkar22 in #1911
- feat(local): add the chat eval, judge its runs, and fix the compact chat prompt by @MODSetter in #1912
- feat(local): measure retrieval, then rank on both legs by @MODSetter in #1913
- fix(local): keep a word's combining marks when the keyword index splits text by @MODSetter in #1915
- fix(search): score every candidate on its own cosine by @MODSetter in #1916
- [LLM|Models] Add image editing, video and more image models by @CREDO23 in #1914
- ci: split CI into one workflow per product and keep only repo-wide checks shared by @AnishSarkar22 in #1917
- fix(local): renumber the sd.cpp migration to 0018 by @AnishSarkar22 in #1918
- feat(local): reorder onboarding and Settings model steps, and fix clipped UI in longer languages by @AnishSarkar22 in #1919
- [LLM|Chat] Show a thinking model's reasoning while it answers by @CREDO23 in #1920
- [LLM|Studio] Stop the model on cancel and turn thinking off by @CREDO23 in #1921
- refactor(local-ui): move the desktop frontend from Radix UI to Base UI by @AnishSarkar22 in #1922
- feat(local): podcast voice and language pickers, scroll-fade and checkbox polish by @AnishSarkar22 in #1923
- feat(local): add an About section to settings and re-sync UI components to base-nova by @AnishSarkar22 in #1924
- fix(local): make room for local image and podcast generation by @MODSetter in #1925
- feat(local): report issues with this session's log by @MODSetter in #1926
- feat(local): macOS Help menu, update check, window restore and thinking header redesign by @AnishSarkar22 in #1927
- fix(local): fade dialogs out with their content, nest app-wide dialogs, hide servers for audio by @AnishSarkar22 in #1929
- feat(local): make model installs jobs the API owns by @AnishSarkar22 in #1930
- [Local|Packaging] Fail pnpm dist on a missing toolchain, with the install command by @CREDO23 in #1928
- fix(local): read the model manifests as UTF-8 on Windows by @AnishSarkar22 in #1931
- [Local|Release] Release desktop 2.0.3 by @CREDO23 in #1932
Full Changelog: v2.0.2...v2.0.3