gpt4all/gpt4all-chat/CHANGELOG.md

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog.

3.10.0 - 2025-02-24

Added

• Whitelist Granite (non-MoE) model architecture (by @ThiloteE in #3487)
• Add support for CUDA compute 5.0 GPUs such as the GTX 750 (#3499)
• Add a Remote Providers tab to the Add Model page (#3506)

Changed

• Substitute prettier default templates for OLMoE 7B 0924/0125 and Granite 3.1 3B/8B (by @ThiloteE in #3471)
• Build with LLVM Clang 19 on macOS and Ubuntu (#3500)

Fixed

• Fix several potential crashes (#3465)
• Fix visual spacing issues with deepseek models (#3470)
• Add missing strings to Italian translation (by @Harvester62 in #3496)
• Update Simplified Chinese translation (by @Junior2Ran in #3467)

3.9.0 - 2025-02-04

Added

• Whitelist OLMoE and Granite MoE model architectures (no Vulkan) (by @ThiloteE in #3449)

Fixed

• Fix "index N is not a prompt" when using LocalDocs with reasoning (#3451)
• Work around rendering artifacts on Snapdragon SoCs with Windows (#3450)
• Prevent DeepSeek-R1 reasoning from appearing in chat names and follow-up questions (#3458)
• Fix LocalDocs crash on Windows ARM when reading PDFs (#3460)
• Fix UI freeze when chat template is {# (#3446)

3.8.0 - 2025-01-30

Added

• Support DeepSeek-R1 Qwen models (#3431)
• Support for think tags in the GUI (#3440)
• Support specifying SHA256 hash in models3.json instead of MD5 (#3437)

Changed

• Use minja instead of Jinja2Cpp for significantly improved template compatibility (#3433)

Fixed

• Fix regression while using localdocs with server API (#3410)
• Don't show system messages in server chat view (#3411)
• Fix codesign --verify failure on macOS (#3413)
• Code Interpreter: Fix console.log not accepting a single string after v3.7.0 (#3426)
• Fix Phi 3.1 Mini 128K Instruct template (by @ThiloteE in #3412)
• Don't block the gui thread for reasoning (#3435)
• Fix corruption of unicode in output of reasoning models (#3443)

3.7.0 - 2025-01-21

Added

• Add support for the Windows ARM64 target platform (CPU-only) (#3385)

Changed

• Update from Qt 6.5.1 to 6.8.1 (#3386)

Fixed

• Fix the timeout error in code interpreter (#3369)
• Fix code interpreter console.log not accepting multiple arguments (#3371)
• Remove 'X is defined' checks from templates for better compatibility (#3372)
• Jinja2Cpp: Add 'if' requirement for 'else' parsing to fix crash (#3373)
• Save chats on quit, even if the window isn't closed first (#3387)
• Add chat template replacements for five new models and fix EM German Mistral (#3393)
• Fix crash when entering {{ a["foo"( as chat template (#3394)
• Sign the maintenance tool on macOS to prevent crash on Sequoia (#3391)
• Jinja2Cpp: Fix operator precedence in 'not X is defined' (#3402)

3.6.1 - 2024-12-20

Fixed

• Fix the stop generation button no longer working in v3.6.0 (#3336)
• Fix the copy entire conversation button no longer working in v3.6.0 (#3336)

3.6.0 - 2024-12-19

Added

• Automatically substitute chat templates that are not compatible with Jinja2Cpp in GGUFs (#3327)
• Built-in javascript code interpreter tool plus model (#3173)

Fixed

• Fix remote model template to allow for XML in messages (#3318)
• Fix Jinja2Cpp bug that broke system message detection in chat templates (#3325)
• Fix LocalDocs sources displaying in unconsolidated form after v3.5.0 (#3328)

3.5.3 - 2024-12-16

Fixed

• Fix LocalDocs not using information from sources in v3.5.2 (#3302)

3.5.2 - 2024-12-13

Added

• Create separate download pages for built-in and HuggingFace models (#3269)

Fixed

• Fix API server ignoring assistant messages in history after v3.5.0 (#3256)
• Fix API server replying with incorrect token counts and stop reason after v3.5.0 (#3256)
• Fix API server remembering previous, unrelated conversations after v3.5.0 (#3256)
• Fix mishandling of default chat template and system message of cloned models in v3.5.0 (#3262)
• Fix untranslated text on the startup dialog (#3293)

3.5.1 - 2024-12-10

Fixed

• Fix an incorrect value for currentResponse (#3245)
• Fix the default model button so it works again after 3.5.0 (#3246)
• Fix chat templates for Nous Hermes 2 Mistral, Mistral OpenOrca, Qwen 2, and remote models (#3250)
• Fix chat templates for Llama 3.2 models (#3251)

3.5.0 - 2024-12-09

Changed

• Update Italian translation (by @Harvester62 in #3236)
• Update Romanian translation (by @SINAPSA-IC in #3232)

Fixed

• Fix a few more problems with the Jinja changes (#3239)

3.5.0-rc2 - 2024-12-06

Changed

• Fade messages out with an animation when they are removed from the chat view (#3227)
• Tweak wording of edit/redo confirmation dialogs (#3228)
• Make edit/redo buttons disabled instead of invisible when they are temporarily unavailable (#3228)

3.5.0-rc1 - 2024-12-04

Added

• Add ability to attach text, markdown, and rst files to chat (#3135)
• Add feature to minimize to system tray (by @bgallois in #3109)
• Basic cache for faster prefill when the input shares a prefix with previous context (#3073)
• Add ability to edit prompts and regenerate any response (#3147)

Changed

• Implement Qt 6.8 compatibility (#3121)
• Use Jinja for chat templates instead of per-message QString.arg-style templates (#3147)
• API server: Use system message(s) from client instead of settings (#3147)
• API server: Accept messages in any order supported by the model instead of requiring user/assistant pairs (#3147)
• Remote models: Pass system message with "system" role instead of joining with user message (#3147)

Removed

• Remove option to save binary model state to disk (#3147)

Fixed

• Fix bug in GUI when localdocs encounters binary data (#3137)
• Fix LocalDocs bugs that prevented some docx files from fully chunking (#3140)
• Fix missing softmax that was causing crashes and effectively infinite temperature since 3.4.0 (#3202)

3.4.2 - 2024-10-16

Fixed

• Limit bm25 retrieval to only specified collections (#3083)
• Fix bug removing documents because of a wrong case sensitive file suffix check (#3083)
• Fix bug with hybrid localdocs search where database would get out of sync (#3083)
• Fix GUI bug where the localdocs embedding device appears blank (#3083)
• Prevent LocalDocs from not making progress in certain cases (#3094)

3.4.1 - 2024-10-11

Fixed

• Improve the Italian translation (#3048)
• Fix models.json cache location (#3052)
• Fix LocalDocs regressions caused by docx change (#3079)
• Fix Go code being highlighted as Java (#3080)

3.4.0 - 2024-10-08

Added

• Add bm25 hybrid search to localdocs (#2969)
• LocalDocs support for .docx files (#2986)
• Add support for attaching Excel spreadsheet to chat (#3007, #3028)

Changed

• Rebase llama.cpp on latest upstream as of September 26th (#2998)
• Change the error message when a message is too long (#3004)
• Simplify chatmodel to get rid of unnecessary field and bump chat version (#3016)
• Allow ChatLLM to have direct access to ChatModel for restoring state from text (#3018)
• Improvements to XLSX conversion and UI fix (#3022)

Fixed

• Fix a crash when attempting to continue a chat loaded from disk (#2995)
• Fix the local server rejecting min_p/top_p less than 1 (#2996)
• Fix "regenerate" always forgetting the most recent message (#3011)
• Fix loaded chats forgetting context when there is a system prompt (#3015)
• Make it possible to downgrade and keep some chats, and avoid crash for some model types (#3030)
• Fix scroll positition being reset in model view, and attempt a better fix for the clone issue (#3042)

3.3.1 - 2024-09-27 (v3.3.y)

Fixed

• Fix a crash when attempting to continue a chat loaded from disk (#2995)
• Fix the local server rejecting min_p/top_p less than 1 (#2996)

3.3.0 - 2024-09-20

Added

• Use greedy sampling when temperature is set to zero (#2854)
• Use configured system prompt in server mode and ignore system messages (#2921, #2924)
• Add more system information to anonymous usage stats (#2939)
• Check for unsupported Ubuntu and macOS versions at install time (#2940)

Changed

• The offline update button now directs users to the offline installer releases page. (by @3Simplex in #2888)
• Change the website link on the home page to point to the new URL (#2915)
• Smaller default window size, dynamic minimum size, and scaling tweaks (#2904)
• Only allow a single instance of program to be run at a time (#2923)

Fixed

• Bring back "Auto" option for Embeddings Device as "Application default," which went missing in v3.1.0 (#2873)
• Correct a few strings in the Italian translation (by @Harvester62 in #2872 and #2909)
• Correct typos in Traditional Chinese translation (by @supersonictw in #2852)
• Set the window icon on Linux (#2880)
• Corrections to the Romanian translation (by @SINAPSA-IC in #2890)
• Fix singular/plural forms of LocalDocs "x Sources" (by @cosmic-snow in #2885)
• Fix a typo in Model Settings (by @3Simplex in #2916)
• Fix the antenna icon tooltip when using the local server (#2922)
• Fix a few issues with locating files and handling errors when loading remote models on startup (#2875)
• Significantly improve API server request parsing and response correctness (#2929)
• Remove unnecessary dependency on Qt WaylandCompositor module (#2949)
• Update translations (#2970)
• Fix macOS installer and remove extra installed copy of Nomic Embed (#2973)

3.2.1 - 2024-08-13

Fixed

• Do not initialize Vulkan driver when only using CPU (#2843)
• Fix a potential crash on exit when using only CPU on Linux with NVIDIA (does not affect X11) (#2843)
• Fix default CUDA architecture list after #2802 (#2855)

3.2.0 - 2024-08-12

Added

• Add Qwen2-1.5B-Instruct to models3.json (by @ThiloteE in #2759)
• Enable translation feature for seven languages: English, Spanish, Italian, Portuguese, Chinese Simplified, Chinese Traditional, Romanian (#2830)

Changed

• Add missing entries to Italian transltation (by @Harvester62 in #2783)
• Use llama_kv_cache ops to shift context faster (#2781)
• Don't stop generating at end of context (#2781)

Fixed

• Case-insensitive LocalDocs source icon detection (by @cosmic-snow in #2761)
• Fix comparison of pre- and post-release versions for update check and models3.json (#2762, #2772)
• Fix several backend issues (#2778)
  • Restore leading space removal logic that was incorrectly removed in #2694
  • CUDA: Cherry-pick llama.cpp DMMV cols requirement fix that caused a crash with long conversations since #2694
• Make reverse prompt detection work more reliably and prevent it from breaking output (#2781)
• Disallow context shift for chat name and follow-up generation to prevent bugs (#2781)
• Explicitly target macOS 12.6 in CI to fix Metal compatibility on older macOS (#2846)

3.1.1 - 2024-07-27

Added

• Add Llama 3.1 8B Instruct to models3.json (by @3Simplex in #2731 and #2732)
• Portuguese (BR) translation (by thiagojramos in #2733)
• Support adding arbitrary OpenAI-compatible models by URL (by @supersonictw in #2683)
• Support Llama 3.1 RoPE scaling (#2758)

Changed

• Add missing entries to Chinese (Simplified) translation (by wuodoo in #2716 and #2749)
• Update translation files and add missing paths to CMakeLists.txt (#2735)

3.1.0 - 2024-07-24

Added

• Generate suggested follow-up questions (#2634, #2723)
  • Also add options for the chat name and follow-up question prompt templates
• Scaffolding for translations (#2612)
• Spanish (MX) translation (by @jstayco in #2654)
• Chinese (Simplified) translation by mikage (#2657)
• Dynamic changes of language and locale at runtime (#2659, #2677)
• Romanian translation by @SINAPSA_IC (#2662)
• Chinese (Traditional) translation (by @supersonictw in #2661)
• Italian translation (by @Harvester62 in #2700)

Changed

• Customize combo boxes and context menus to fit the new style (#2535)
• Improve view bar scaling and Model Settings layout (#2520
• Make the logo spin while the model is generating (#2557)
• Server: Reply to wrong GET/POST method with HTTP 405 instead of 404 (by @cosmic-snow in #2615)
• Update theme for menus (by @3Simplex in #2578)
• Move the "stop" button to the message box (#2561)
• Build with CUDA 11.8 for better compatibility (#2639)
• Make links in latest news section clickable (#2643)
• Support translation of settings choices (#2667, #2690)
• Improve LocalDocs view's error message (by @cosmic-snow in #2679)
• Ignore case of LocalDocs file extensions (#2642, #2684)
• Update llama.cpp to commit 87e397d00 from July 19th (#2694, #2702)
  • Add support for GPT-NeoX, Gemma 2, OpenELM, ChatGLM, and Jais architectures (all with Vulkan support)
  • Add support for DeepSeek-V2 architecture (no Vulkan support)
  • Enable Vulkan support for StarCoder2, XVERSE, Command R, and OLMo
• Show scrollbar in chat collections list as needed (by @cosmic-snow in #2691)

Removed

• Remove support for GPT-J models (#2676, #2693)

Fixed

• Fix placement of thumbs-down and datalake opt-in dialogs (#2540)
• Select the correct folder with the Linux fallback folder dialog (#2541)
• Fix clone button sometimes producing blank model info (#2545)
• Fix jerky chat view scrolling (#2555)
• Fix "reload" showing for chats with missing models (#2520
• Fix property binding loop warning (#2601)
• Fix UI hang with certain chat view content (#2543)
• Fix crash when Kompute falls back to CPU (#2640)
• Fix several Vulkan resource management issues (#2694)
• Fix crash/hang when some models stop generating, by showing special tokens (#2701)