"notes":"* repeat penalty for both gptj and llama models\n* scroll the context window when conversation reaches context limit\n* persistent thread count setting\n* new default template\n* new settings for model path, repeat penalty\n* bugfix for settings dialog onEditingFinished\n* new tab based settings dialog format\n* bugfix for datalake when conversation contains forbidden json chars\n* new C library API and split the backend into own separate lib for bindings\n* apple signed/notarized dmg installer\n* update llama.cpp submodule to latest\n* bugfix for too large of a prompt\n* support for opt-in only anonymous usage and statistics\n* bugfixes for the model downloader and improve performance\n* various UI bugfixes and enhancements including the send message textarea automatically wrapping by word\n* new startup dialog on first start of a new release displaying release notes and opt-in buttons\n* new logo and icons\n* fixed apple installer so there is now a symlink in the applications folder\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Matthieu Talbot\n* Tim Jobbins\n* chad (eachadea)\n* Community (beta testers, bug reporters)"
"notes":"* repeat penalty for both gptj and llama models\n* scroll the context window when conversation reaches context limit\n* persistent thread count setting\n* new default template\n* new settings for model path, repeat penalty\n* bugfix for settings dialog onEditingFinished\n* new tab based settings dialog format\n* bugfix for datalake when conversation contains forbidden json chars\n* new C library API and split the backend into own separate lib for bindings\n* apple signed/notarized dmg installer\n* update llama.cpp submodule to latest\n* bugfix for too large of a prompt\n* support for opt-in only anonymous usage and statistics\n* bugfixes for the model downloader and improve performance\n* various UI bugfixes and enhancements including the send message textarea automatically wrapping by word\n* new startup dialog on first start of a new release displaying release notes and opt-in buttons\n* new logo and icons\n* fixed apple installer so there is now a symlink in the applications folder\n* fixed bug with versions\n* fixed optout marking\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Matthieu Talbot\n* Tim Jobbins\n* chad (eachadea)\n* Community (beta testers, bug reporters)"
"notes":"* reverse prompt for both llama and gptj models which should help stop them from repeating the prompt template\n* resumable downloads for models\n* chat list in the drawer drop down\n* add/remove/rename chats\n* persist chats to disk and restore them with full context (WARNING: the average size of each chat on disk is ~1.5GB)\n* NOTE: to turn on the persistent chats feature you need to do so via the settings dialog as it is off by default\n* automatically rename chats using the AI after the first prompt/response pair\n* new usage statistics including more detailed hardware info to help debug problems on older hardware\n* fix dialog sizes for those with smaller displays\n* add support for persistent contexts and internal model state to the C api\n* add a confirm button for deletion of chats\n* bugfix for blocking the gui when changing models\n* datalake now captures all conversations when network opt-in is turned on\n* new much shorter prompt template by default\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Community (beta testers, bug reporters)"
"notes":"* compress persistent chats and save order of magnitude disk space on some small chats\n* persistent chat files are now stored in same folder as models\n* use a thread for deserializing chats on startup so the gui shows window faster\n* fail gracefully and early when we detect incompatible hardware\n* repeat penalty restore default bugfix\n* new mpt backend for mosaic ml's new base model and chat model\n* add mpt chat and base model to downloads\n* lower memory required for gptj models by using f16 for kv cache\n* better error handling for when a model is deleted by user and persistent chat remains\n* add a user default model setting so the users preferred model comes up on startup\n",
"contributors":"* Adam Treat (Nomic AI)\n* Zach Nussbaum (Nomic AI)\n* Aaron Miller\n* Community (beta testers, bug reporters)"
"notes":"* add webserver feature that offers mirror api to chatgpt on localhost:4891\n* add chatgpt models installed using openai key to chat client gui\n* fixup the memory handling when switching between chats/models to decrease RAM load across the board\n* fix bug in thread safety for mpt model and de-duplicated code\n* uses compact json format for network\n* add remove model option in download dialog\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Community (beta testers, bug reporters)"
"notes":"* add webserver feature that offers mirror api to chatgpt on localhost:4891\n* add chatgpt models installed using openai key to chat client gui\n* fixup the memory handling when switching between chats/models to decrease RAM load across the board\n* fix bug in thread safety for mpt model and de-duplicated code\n* uses compact json format for network\n* add remove model option in download dialog\n* remove text-davinci-003 as it is not a chat model\n* fix installers on mac and linux to include libllmodel versions\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Community (beta testers, bug reporters)"
"notes":"* fix buffer overrun in backend\n* bugfix for browse for model directory\n* dedup of qml code\n* revamp settings dialog UI\n* add localdocs plugin (beta) feature allowing scanning of local docs\n* various other bugfixes and performance improvements\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller\n* Juuso Alasuutari\n* Justin Wang\n* Community (beta testers, bug reporters)"
"notes":"* bugfix for model download remove\n* bugfix for blocking on regenerate\n* lots of various ui improvements enhancements\n* big new change that brings us up2date with llama.cpp/ggml support for latest models\n* advanced avx detection allowing us to fold the two installers into one\n* new logging mechanism that allows for bug reports to have more detail\n* make localdocs work with server mode\n* localdocs fix for stale references after we regenerate\n* fix so that browse to dialog on linux\n* fix so that you can also just add a path to the textfield\n* bugfix for chatgpt and resetting context\n* move models.json to github repo so people can pr suggested new models\n* allow for new models to be directly downloaded from huggingface in said prs\n* better ui for localdocs settings\n* better error handling when model fails to load\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Konstantin Gukov\n* Joseph Mearman\n* Nandakumar\n* Chase McDougall\n* mvenditto\n* Andriy Mulyar (Nomic AI)\n* FoivosC\n* Ettore Di Giacinto\n* Tim Miller\n* Peter Gagarinov\n* Community (beta testers, bug reporters)"
"notes":"* bugfix for model download remove\n* bugfix for blocking on regenerate\n* lots of various ui improvements enhancements\n* big new change that brings us up2date with llama.cpp/ggml support for latest models\n* advanced avx detection allowing us to fold the two installers into one\n* new logging mechanism that allows for bug reports to have more detail\n* make localdocs work with server mode\n* localdocs fix for stale references after we regenerate\n* fix so that browse to dialog on linux\n* fix so that you can also just add a path to the textfield\n* bugfix for chatgpt and resetting context\n* move models.json to github repo so people can pr suggested new models\n* allow for new models to be directly downloaded from huggingface in said prs\n* better ui for localdocs settings\n* better error handling when model fails to load\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Konstantin Gukov\n* Joseph Mearman\n* Nandakumar\n* Chase McDougall\n* mvenditto\n* Andriy Mulyar (Nomic AI)\n* FoivosC\n* Ettore Di Giacinto\n* Tim Miller\n* Peter Gagarinov\n* Community (beta testers, bug reporters)"
"notes":"* replit model support\n* macos metal accelerated support\n* fix markdown for localdocs references\n* inline syntax highlighting for python and cpp with more languages coming\n* synced with upstream llama.cpp\n* ui fixes and default generation settings changes\n* backend bugfixes\n* allow for loading files directly from huggingface via TheBloke without name changes\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Andriy Mulyar (Nomic AI)\n* Ettore Di Giacinto\n* AMOGUS\n* Felix Zaslavskiy\n* Tim Miller\n* Community (beta testers, bug reporters)"
"notes":"* replit model support\n* macos metal accelerated support\n* fix markdown for localdocs references\n* inline syntax highlighting for python and cpp with more languages coming\n* synced with upstream llama.cpp\n* ui fixes and default generation settings changes\n* backend bugfixes\n* allow for loading files directly from huggingface via TheBloke without name changes\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Andriy Mulyar (Nomic AI)\n* Ettore Di Giacinto\n* AMOGUS\n* Felix Zaslavskiy\n* Tim Miller\n* Community (beta testers, bug reporters)"
"notes":"* New GPT4All Falcon model\n* New Orca models\n* Token generation speed is now reported in GUI\n* Bugfix for localdocs references when regenerating\n* General fixes for thread safety\n* Many fixes to UI to add descriptions for error conditions\n* Fixes for saving/reloading chats\n* Complete refactor of the model download dialog with metadata about models available\n* Resume downloads bugfix\n* CORS fix\n* Documentation fixes and typos\n* Latest llama.cpp update\n* Update of replit\n* Force metal setting\n* Fixes for model loading with metal on macOS\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Andriy Mulyar (Nomic AI)\n* cosmic-snow\n* AMOGUS\n* Community (beta testers, bug reporters)"
"notes":"* New GPT4All Falcon model\n* New Orca models\n* Token generation speed is now reported in GUI\n* Bugfix for localdocs references when regenerating\n* General fixes for thread safety\n* Many fixes to UI to add descriptions for error conditions\n* Fixes for saving/reloading chats\n* Complete refactor of the model download dialog with metadata about models available\n* Resume downloads bugfix\n* CORS fix\n* Documentation fixes and typos\n* Latest llama.cpp update\n* Update of replit\n* Force metal setting\n* Fixes for model loading with metal on macOS\n",
"contributors":"* Nils Sauer (Nomic AI)\n* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Richard Guo (Nomic AI)\n* Andriy Mulyar (Nomic AI)\n* cosmic-snow\n* AMOGUS\n* Community (beta testers, bug reporters)"
"notes":"* Per model settings\n* Character settings\n* Adding a system prompt\n* Important bugfix for chatgpt install\n* Complete refactor and revamp of settings dialog\n* New syntax highlighting for java, bash, go\n* Use monospace font for syntax highlighting of codeblocks\n* New setting for turning off references in localdocs\n* Fix memory leaks in falcon model\n* Fix for backend memory handling\n* Server mode bugfix\n* Models.json retrieve bugfix\n* Free metal context bugfix\n* Add a close dialog feature to all chat dialogs\n",
"contributors":"* Lakshay Kansal (Nomic AI)\n* Matthew Gill\n* Brandon Beiler\n* cosmic-snow\n* Felix Zaslavskiy\n* Andriy Mulyar (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters)"
"notes":"* Fix bad bug that was breaking numerous current installs (sorry folks!)\n* Fix bug with 'browse' button in settings dialog\n* Wayland support on linux\n* Reduce template ui size in settings dialog\n",
"contributors":"* Akarshan Biswas\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters)"
"notes":"* Fix bug with prolonging shutdown with generation\n* Fix bug with update model info on deleting chats\n* Fix bug with preventing closing of model download dialog\n* Allows allow closing the model download dialog\n* Fix numerous bugs with download of models.json and provide backup option\n* Add json and c# highlighting\n* Fix bug with chatgpt crashing\n* Fix bug with chatgpt not working for some keys\n* Fix bug with mixpanel opt outs not counting\n* Fix problem with OOM errors causing crash and then repeating on next start\n* Fix default thread setting and provide guardrails\n* Fix tap handler in settings dialog for buttons\n* Fix color of some text fields on macOS for settings dialog\n* Fix problem with startup dialog not closing\n* Provide error dialog for settings file not accessible\n* Try and fix problems with avx-only detection\n* Fix showing error in model downloads unnecessarily\n* Prefer 7b models to load by default\n* Add Wizard v1.1 to download list\n* Rename Orca models to Mini Orca\n* Don't use a system prompt unless model was trained with one by default\n",
"contributors":"* Lakshay Kansal (Nomic AI)\n* Aaron Miller (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters)"
"notes":"* Add starcoder model support\n* Add ability to switch between light mode/dark mode\n* Increase the size of fonts in monospace code blocks a bit\n",
"contributors":"* Lakshay Kansal (Nomic AI)\n* Adam Treat (Nomic AI)"
"notes":"* Bugfix for properly falling back to CPU when GPU can't be used\n* Report the actual device we're using\n* Fix context bugs for GPU accelerated models\n",
"contributors":"* Adam Treat (Nomic AI)\n* Aaron Miller (Nomic AI)"
"notes":"* Bugfix for devices to show up in the settings combobox on application start and not just on model load\n* Send information on requested device and actual device on model load to help assess which model/gpu/os combos are working\n",
"notes":"* Major new release supports GGUF models only!\n* New models like Mistral Instruct, Replit 1.5, Rift Coder and more\n* All previous version of ggml-based models are no longer supported\n* Extensive changes to vulkan support\n* Better GPU error messages\n* Prompt processing on the GPU\n* Save chats now saves to text (less harddrive space)\n* Many more changes\n",
"contributors":"* Aaron Miller (Nomic AI)\n* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Accessibility fixes\n* Bugfix for crasher on Windows\n",
"contributors":"* Aaron Miller (Nomic AI)\n* Jared Van Bortel (Nomic AI)\n* Victor Tsaran <vtsaran@yahoo.com>\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Support for GGUF v3 models\n* Important fixes for AMD GPUs\n* Don't start recalculating context immediately for saved chats\n* UI fixes for chat name generation\n* UI fixes for leading whitespaces in chat generation\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Major feature update for localdocs!\n* Localdocs now uses an embedding model for retrieval augmented generation\n* Localdocs can now search while your collections are indexing\n* You're guaranteed to get hits from localdocs for every prompt you enter\n* Fix: AMD gpu fixes\n* Fix: Better error messages\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Major bugfix release with new models!\n* Model: Recently released Orca 2 model which does exceptionally well on reasoning tasks\n* Fix: System prompt was not always being honored\n* Fix: Download network retry on cloudflare errors\n",
"contributors":"* Adam Treat (Nomic AI)\n* Jared Van Bortel (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Update to newer llama.cpp\n* Implemented configurable context length\n* Bugfixes for localdocs\n* Bugfixes for serialization to disk\n* Bugfixes for AVX\n* Bugfixes for Windows builds\n* Bugfixes for context retention and clearing\n* Add a button to collections dialog\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Update to latest llama.cpp\n* Update to newly merged vulkan backend\n* Partial GPU offloading support\n* New localdocs speed increases and features\n* New GUI settings option for configuring how many layers to put on GPU\n* New lightmode theme, darkmode theme and legacy theme\n* Lots of UI updates and enhancements\n* Scores of bugfixes for stability and usability\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Karthik Nair\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Add support for twelve new model architectures\n* Including Baichuan, BLOOM, CodeShell, GPT-2, Orion, Persimmon, Phi and Phi-2, Plamo, Qwen, Qwen2, Refact, and StableLM\n* Fix for progress bar colors on legacy theme\n* Fix sizing for model download dialog elements\n* Fix dialog sizes to use more screen realestate where available\n* Fix for vram leak when model loading fails\n* Fix for making the collection dialog progress bar more readable\n* Fix for smaller minimum size for main screen\n* Fix for mistral crash\n* Fix for mistral openorca prompt template to ChatLM\n* Fix for excluding non-text documents from localdoc scanning\n* Fix for scrollbar missing on main conversation\n* Fix accessibility issues for screen readers\n* Fix for not showing the download button when not online\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Update to latest llama.cpp with support for Google Gemma\n* Gemma, Phi and Phi-2, Qwen2, and StableLM are now all GPU accelerated\n* Large revamp of the model loading to support explicit unload/reload\n* Bugfixes for ChatML and improved version of Mistral OpenOrca\n* We no longer load a model by default on application start\n* We no longer load a model by default on chat context switch\n* Fixes for visual artifacts in update reminder dialog\n* Blacklist Intel GPU's for now as we don't support yet\n* Fixes for binary save/restore of chat\n* Save and restore of window geometry across application starts\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* New support for model search/discovery using huggingface search in downloads\n* Support for more model architectures for GPU acceleration\n* Three different crash fixes for corner case settings\n* Add a minp sampling parameter\n* Bert layoer norm epsilon value\n* Fix problem with blank lines between reply and next prompt\n",
"contributors":"* Christopher Barrera\n* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"* Fix for network reachability unknown\n* Fix undefined behavior with resetContext\n* Fix ChatGPT which was broken with previous release\n* Fix for clean up of chat llm thread destruction\n* Display of model loading warnings\n* Fix for issue 2080 where the GUI appears to hang when a chat is deleted\n* Fix for issue 2077 better responsiveness of model download dialog when download is taking place\n* Fix for issue 2092 don't include models that are disabled for GUI in application default model list\n* Fix for issue 2087 where cloned modelds were lost and listed in download dialog erroneously\n* Fix for MPT models without duplicated token embd weight\n* New feature with api server port setting\n* Fix for issue 2024 where combobox for model settings uses currently used model by default\n* Clean up settings properly for removed models and don't list stale model settings in download dialog\n* Fix for issue 2105 where the cancel button was not working for discovered model downloads\n",
"contributors":"* Christopher Barrera\n* Daniel Alencar\n* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Add a right-click menu to the chat (by @kryotek777 in PR #2108)\n* Change the left sidebar to stay open (PR #2117)\n* Limit the width of text in the chat (PR #2118)\n* Move to llama.cpp's SBert implementation (PR #2086)\n* Support models provided by the Mistral AI API (by @Olyxz16 in PR #2053)\n* Models List: Add Ghost 7B v0.9.1 (by @lh0x00 in PR #2127)\n* Add Documentation and FAQ links to the New Chat page (by @3Simplex in PR #2183)\n* Models List: Simplify Mistral OpenOrca system prompt (PR #2220)\n* Models List: Add Llama 3 Instruct (PR #2242)\n* Models List: Add Phi-3 Mini Instruct (PR #2252)\n* Improve accuracy of anonymous usage statistics (PR #2238)\n\n<b>— Fixes —</b>\n* Detect unsupported CPUs correctly on Windows (PR #2141)\n* Fix the colors used by the server chat (PR #2150)\n* Fix startup issues when encountering non-Latin characters in paths (PR #2162)\n* Fix issues causing LocalDocs context links to not work sometimes (PR #2218)\n* Fix incorrect display of certain code block syntax in the chat (PR #2232)\n* Fix an issue causing unnecessary indexing of document collections on startup (PR #2236)\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Lam Hieu (`@lh0x00`)\n* 3Simplex (`@3Simplex`)\n* Kryotek (`@kryotek777`)\n* Olyxz16 (`@Olyxz16`)\n* Robin Verduijn (`@robinverduijn`)\n* Tim453 (`@Tim453`)\n* Xu Zhen (`@xuzhen`)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Improve accuracy of anonymous usage statistics (PR #2297, PR #2299)\n\n<b>— Fixes —</b>\n* Fix some issues with anonymous usage statistics (PR #2270, PR #2296)\n* Default to GPU with most VRAM on Windows and Linux, not least (PR #2297)\n* Fix initial failure to generate embeddings with Nomic Embed (PR #2284)\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Context Menu: Replace \"Select All\" on message with \"Copy Message\" (PR #2324)\n* Context Menu: Hide Copy/Cut when nothing is selected (PR #2324)\n* Improve speed of context switch after quickly switching between several chats (PR #2343)\n* New Chat: Always switch to the new chat when the button is clicked (PR #2330)\n* New Chat: Always scroll to the top of the list when the button is clicked (PR #2330)\n* Update to latest llama.cpp as of May 9, 2024 (PR #2310)\n* **Add support for the llama.cpp CUDA backend** (PR #2310, PR #2357)\n * Nomic Vulkan is still used by default, but CUDA devices can now be selected in Settings\n * When in use: Greatly improved prompt processing and generation speed on some devices\n * When in use: GPU support for Q5\\_0, Q5\\_1, Q8\\_0, K-quants, I-quants, and Mixtral\n* Add support for InternLM models (PR #2310)\n\n<b>— Fixes —</b>\n* Do not allow sending a message while the LLM is responding (PR #2323)\n* Fix poor quality of generated chat titles with many models (PR #2322)\n* Set the window icon correctly on Windows (PR #2321)\n* Fix a few memory leaks (PR #2328, PR #2348, PR #2310)\n* Do not crash if a model file has no architecture key (PR #2346)\n* Fix several instances of model loading progress displaying incorrectly (PR #2337, PR #2343)\n* New Chat: Fix the new chat being scrolled above the top of the list on startup (PR #2330)\n* macOS: Show a \"Metal\" device option, and actually use the CPU when \"CPU\" is selected (PR #2310)\n* Remove unsupported Mamba, Persimmon, and PLaMo models from the whitelist (PR #2310)\n* Fix GPT4All.desktop being created by offline installers on macOS (PR #2361)\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Tim453 (`@Tim453`)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Complete UI overhaul (PR #2396)\n* LocalDocs improvements (PR #2396)\n * Use nomic-embed-text-v1.5 as local model instead of SBert\n * Ship local model with application instead of downloading afterwards\n * Store embeddings flat in SQLite DB instead of in hnswlib index\n * Do exact KNN search with usearch instead of approximate KNN search with hnswlib\n* Markdown support (PR #2476)\n* Support CUDA/Metal device option for embeddings (PR #2477)\n\n<b>— Fixes —</b>\n* Fix embedding tokenization after PR #2310 (PR #2381)\n* Fix a crash when loading certain models with \"code\" in their name (PR #2382)\n* Fix an embedding crash with large chunk sizes after PR #2310 (PR #2383)\n* Fix inability to load models with non-ASCII path on Windows (PR #2388)\n* CUDA: Do not show non-fatal DLL errors on Windows (PR #2389)\n* LocalDocs fixes (PR #2396)\n * Always use requested number of snippets even if there are better matches in unselected collections\n * Check for deleted files on startup\n* CUDA: Fix PTX errors with some GPT4All builds (PR #2421)\n* Fix blank device in UI after model switch and improve usage stats (PR #2409)\n* Use CPU instead of CUDA backend when GPU loading fails the first time (ngl=0 is not enough) (PR #2477)\n* Fix crash when sending a message greater than n\\_ctx tokens after PR #1970 (PR #2498)\n",
"contributors":"* Vincent Giardina (Nomic AI)\n* Jared Van Bortel (Nomic AI)\n* John W. Parent (Kitware)\n* Paige Lee (Nomic AI)\n* Max Cembalest (Nomic AI)\n* Andriy Mulyar (Nomic AI)\n* Adam Treat (Nomic AI)\n* cosmic-snow (`@cosmic-snow`)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Generate suggested follow-up questions feature (#2634)\n\n<b>— What's Changed —</b>\n* Customize combo boxes and context menus to fit the new style (#2535)\n* Improve view bar scaling and Model Settings layout (#2520)\n* Make the logo spin while the model is generating (#2557)\n* Server: Reply to wrong GET/POST method with HTTP 405 instead of 404 (#2615)\n* Update theme for menus (#2578)\n* Move the \"stop\" button to the message box (#2561)\n* Build with CUDA 11.8 for better compatibility (#2639)\n* Make links in latest news section clickable (#2643)\n* Support translation of settings choices (#2667), (#2690)\n* Improve LocalDocs view's error message (by @cosmic-snow in #2679)\n* Ignore case of LocalDocs file extensions (#2642), (#2684)\n* Update llama.cpp to commit 87e397d00 from July 19th (#2694)\n * Add support for GPT-NeoX, Gemma 2, OpenELM, ChatGLM, and Jais architectures (all with Vulkan support)\n * Enable Vulkan support for StarCoder2, XVERSE, Command R, and OLMo\n* Show scrollbar in chat collections list as needed (#2691)\n\n<b>— What's Removed —</b>\n* Remove support for GPT-J models (#2676)\n\n<b>— Fixes —</b>\n* Fix placement of thumbs-down and datalake opt-in dialogs (#2540)\n* Select the correct folder with the Linux fallback folder dialog (#2541)\n* Fix clone button sometimes producing blank model info (#2545)\n* Fix jerky chat view scrolling (#2555)\n* Fix \"reload\" showing for chats with missing models (#2520)\n* Fix property binding loop warning (#2601)\n* Fix UI hang with certain chat view content (#2543)\n* Fix crash when Kompute falls back to CPU (#2640)\n* Fix several Vulkan resource management issues (#2694)\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* cosmic-snow (`@cosmic-snow`)\n* 3Simplex (`@3Simplex`)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Ability to add OpenAI compatible remote models (#2683)\n\n<b>— Fixes —</b>\n* Update llama.cpp to cherry-pick Llama 3.1 RoPE fix. (#2758)\n",
"contributors":"* Adam Treat (Nomic AI)\n* Jared Van Bortel (Nomic AI)\n* Shiranui (@supersonictw)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— What's New —</b>\n* Translations for Simplified Chinese, Traditional Chinese, Italian, Portuguese, Romanian, and Spanish\n* Significantly faster context recalculation when context runs out\n* Models no longer stop generating when they run out of context\n* Add Qwen2-1.5B-Instruct to the model list\n\n<b>— Fixes —</b>\n* Fix a CUDA crash with long conversations since v3.1.0\n* Fix \"file(s)\" and \"word(s)\" appearing in UI instead of proper plurals\n* Show the correct icons for LocalDocs sources with uppercase extensions\n* More reliable reverse prompt detection\n* Fix a minor prompting issue introduced in v3.1.0\n* Disallow context shift for chat name and follow-up generation\n* Fix potential incompatibility with macOS 12 and 13\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* Riccardo Giovanetti (`@Harvester62`)\n* Victor Emanuel (`@SINAPSA-IC`)\n* Jeremy Tayco (`@jstayco`)\n* Shiranui (`@supersonictw`)\n* Thiago Ramos (`@thiagojramos`)\n* ThiloteE (`@ThiloteE`)\n* Dominik (`@cosmic-snow`)\n* Jack (`@wuodoo`)\n* Community (beta testers, bug reporters, bindings authors)"
"notes":"<b>— Fixes —</b>\n* Fix a potential Vulkan crash on application exit on some Linux systems\n* Fix a bad CUDA build option that led to gibberish on newer NVIDIA GPUs\n",
"notes":"* **UI Improvements**: The minimum window size now adapts to the font size. A few labels and links have been fixed. The Embeddings Device selection of \"Auto\"/\"Application default\" works again. The window icon is now set on Linux. The antenna icon now displays when the API server is listening.\n* **Single Instance**: Only one instance of GPT4All can be opened at a time. This is now enforced.\n* **Greedy Sampling**: Set temperature to zero to enable greedy sampling.\n* **API Server Changes**: The built-in API server now responds correctly to both legacy completions, and chats with message history. Also, it now uses the system prompt configured in the UI.\n* **Translation Improvements**: The Italian, Romanian, and Traditional Chinese translations have been updated.\n",
"contributors":"* Jared Van Bortel (Nomic AI)\n* Adam Treat (Nomic AI)\n* 3Simplex (`@3Simplex`)\n* Riccardo Giovanetti (`@Harvester62`)\n* Victor Emanuel (`@SINAPSA-IC`)\n* Dominik (`@cosmic-snow`)\n* Shiranui (`@supersonictw`)"
