|
854 | 854 | "notes":
|
855 | 855 | "
|
856 | 856 | <b>— What's New —</b>
|
857 |
| -* LocalDocs Experience: Completely redesigned LocalDocs UI and backend |
858 |
| -* New User Interface: Modern design making it easier to focus on what matters most - getting things done |
| 857 | +* Complete UI overhaul (PR #2396) |
| 858 | +* LocalDocs improvements (PR #2396) |
| 859 | + * Use nomic-embed-text-v1.5 as local model instead of SBert |
| 860 | + * Ship local model with application instead of downloading afterwards |
| 861 | + * Store embeddings flat in SQLite DB instead of in hnswlib index |
| 862 | + * Do exact KNN search with usearch instead of approximate KNN search with hnswlib |
| 863 | +* Markdown support (PR #2476) |
| 864 | +* Support CUDA/Metal device option for embeddings (PR #2477) |
| 865 | + |
| 866 | +<b>— Fixes —</b> |
| 867 | +* Fix embedding tokenization after PR #2310 (PR #2381) |
| 868 | +* Fix a crash when loading certain models with \"code\" in their name (PR #2382) |
| 869 | +* Fix an embedding crash with large chunk sizes after PR #2310 (PR #2383) |
| 870 | +* Fix inability to load models with non-ASCII path on Windows (PR #2388) |
| 871 | +* CUDA: Do not show non-fatal DLL errors on Windows (PR #2389) |
| 872 | +* LocalDocs fixes (PR #2396) |
| 873 | + * Always use requested number of snippets even if there are better matches in unselected collections |
| 874 | + * Check for deleted files on startup |
| 875 | +* CUDA: Fix PTX errors with some GPT4All builds (PR #2421) |
| 876 | +* Fix blank device in UI after model switch and improve usage stats (PR #2409) |
| 877 | +* Use CPU instead of CUDA backend when GPU loading fails the first time (ngl=0 is not enough) (PR #2477) |
| 878 | +* Fix crash when sending a message greater than n\_ctx tokens after PR #1970 (PR #2498) |
859 | 879 | ",
|
860 | 880 | "contributors":
|
861 | 881 | "
|
|
0 commit comments